From 075f7ebb5fc4614b7896b7150f11f4ab4b6b6622 Mon Sep 17 00:00:00 2001 From: YutaSaito <36355491+uc4w6c@users.noreply.github.com> Date: Wed, 14 Jan 2026 21:20:16 +0900 Subject: [PATCH 01/73] feat: contextual gap checks, word-form digits (#18301) Co-authored-by: Krish Dholakia --- .../litellm_content_filter/content_filter.py | 213 ++++++++++++++++-- .../litellm_content_filter/patterns.json | 23 +- .../litellm_content_filter/patterns.py | 22 +- .../content_filter/test_content_filter.py | 1 - 4 files changed, 226 insertions(+), 33 deletions(-) diff --git a/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/content_filter.py b/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/content_filter.py index a04e438f481..c9bd0135a05 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/content_filter.py +++ b/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/content_filter.py @@ -50,8 +50,32 @@ from litellm.types.proxy.guardrails.guardrail_hooks.litellm_content_filter impor ContentFilterDetection, PatternDetection, ) +from .patterns import PATTERN_EXTRA_CONFIG, get_compiled_pattern -from .patterns import get_compiled_pattern +MAX_KEYWORD_VALUE_GAP_WORDS = 1 +GAP_WORD_TOKENIZER = re.compile(r"\b\w+\b") + + +WORD_NUMBER_MAP = { + "zero": "0", + "oh": "0", + "one": "1", + "two": "2", + "three": "3", + "four": "4", + "five": "5", + "six": "6", + "seven": "7", + "eight": "8", + "nine": "9", +} + +WORD_NUMBER_TOKEN_REGEX = "|".join(WORD_NUMBER_MAP.keys()) +WORD_NUMBER_SEQUENCE_PATTERN = re.compile( + rf"(? (category, severity, action) + self.category_keywords: Dict[ + str, Tuple[str, str, ContentFilterAction] + ] = {} # keyword -> (category, severity, action) # Load categories if provided if categories: @@ -170,7 +194,7 @@ class ContentFilterGuardrail(CustomGuardrail): normalized_blocked_words.append(word) # Compile regex patterns - self.compiled_patterns: List[Tuple[Pattern, str, ContentFilterAction]] = [] + self.compiled_patterns: List[Dict[str, Any]] = [] for pattern_config in normalized_patterns: self._add_pattern(pattern_config) @@ -323,11 +347,13 @@ class ContentFilterGuardrail(CustomGuardrail): pattern_config: ContentFilterPattern configuration """ try: + extra_config: Dict[str, Any] = {} if pattern_config.pattern_type == "prebuilt": if not pattern_config.pattern_name: raise ValueError("pattern_name is required for prebuilt patterns") compiled = get_compiled_pattern(pattern_config.pattern_name) pattern_name = pattern_config.pattern_name + extra_config = PATTERN_EXTRA_CONFIG.get(pattern_name, {}) or {} elif pattern_config.pattern_type == "regex": if not pattern_config.pattern: raise ValueError("pattern is required for regex patterns") @@ -336,8 +362,20 @@ class ContentFilterGuardrail(CustomGuardrail): else: raise ValueError(f"Unknown pattern_type: {pattern_config.pattern_type}") + keyword_regex: Optional[Pattern] = None + if extra_config.get("keyword_pattern"): + keyword_regex = re.compile( + extra_config["keyword_pattern"], re.IGNORECASE + ) + self.compiled_patterns.append( - (compiled, pattern_name, pattern_config.action) + { + "regex": compiled, + "pattern_name": pattern_name, + "action": pattern_config.action, + "keyword_regex": keyword_regex, + "allow_word_numbers": bool(extra_config.get("allow_word_numbers")), + } ) verbose_proxy_logger.debug( f"Added pattern: {pattern_name} with action {pattern_config.action}" @@ -395,6 +433,130 @@ class ContentFilterGuardrail(CustomGuardrail): except Exception as e: raise Exception(f"Error loading blocked words file {file_path}: {str(e)}") + def _find_pattern_spans( + self, text: str, pattern_entry: Dict[str, Any] + ) -> List[Tuple[int, int]]: + """Return all match spans for a pattern, applying contextual rules if required.""" + + regex: Pattern = pattern_entry["regex"] + keyword_regex: Optional[Pattern] = pattern_entry.get("keyword_regex") + allow_word_numbers: bool = pattern_entry.get("allow_word_numbers", False) + + keyword_matches: Optional[List[re.Match]] = None + if keyword_regex is not None: + keyword_matches = list(keyword_regex.finditer(text)) + if not keyword_matches: + return [] + + match_spans: List[Tuple[int, int]] = [] + + for match in regex.finditer(text): + if keyword_matches is not None and not self._match_near_keyword( + match.start(), match.end(), keyword_matches, text + ): + continue + match_spans.append((match.start(), match.end())) + + if allow_word_numbers: + for word_match in WORD_NUMBER_SEQUENCE_PATTERN.finditer(text): + digits = self._convert_word_number_sequence(word_match.group()) + if not digits: + continue + if not regex.fullmatch(digits): + continue + if keyword_matches is not None and not self._match_near_keyword( + word_match.start(), word_match.end(), keyword_matches, text + ): + continue + match_spans.append((word_match.start(), word_match.end())) + + return self._merge_spans(match_spans) + + def _match_near_keyword( + self, + value_start: int, + value_end: int, + keyword_matches: List[re.Match], + text: str, + ) -> bool: + """Check if a value is separated from a keyword by an allowed gap.""" + + for keyword_match in keyword_matches: + keyword_start = keyword_match.start() + keyword_end = keyword_match.end() + + if value_start >= keyword_end: + gap_text = text[keyword_end:value_start] + elif keyword_start >= value_end: + gap_text = text[value_end:keyword_start] + else: + return True # overlapping + + if self._gap_text_allowed(gap_text): + return True + return False + + def _gap_text_allowed(self, gap_text: str) -> bool: + """Return True if the gap between keyword and value meets word-count rules.""" + + if not gap_text.strip(): + return True + if any(char.isdigit() for char in gap_text): + return False + + words = GAP_WORD_TOKENIZER.findall(gap_text) + return len(words) <= MAX_KEYWORD_VALUE_GAP_WORDS + + def _merge_spans(self, spans: List[Tuple[int, int]]) -> List[Tuple[int, int]]: + """Merge overlapping spans to avoid double-masking.""" + + if not spans: + return [] + + spans.sort(key=lambda item: item[0]) + merged: List[Tuple[int, int]] = [spans[0]] + + for start, end in spans[1:]: + last_start, last_end = merged[-1] + if start <= last_end: + merged[-1] = (last_start, max(last_end, end)) + else: + merged.append((start, end)) + return merged + + def _mask_spans( + self, text: str, spans: List[Tuple[int, int]], redaction: str + ) -> str: + """Apply masking for the provided spans using the given redaction tag.""" + + if not spans: + return text + + result_parts: List[str] = [] + previous_end = 0 + for start, end in spans: + result_parts.append(text[previous_end:start]) + result_parts.append(redaction) + previous_end = end + result_parts.append(text[previous_end:]) + return "".join(result_parts) + + def _convert_word_number_sequence(self, sequence: str) -> Optional[str]: + """Convert a spelled-out digit sequence (e.g., 'One-Two') into digits.""" + + tokens = WORD_NUMBER_TOKEN_FINDER.findall(sequence) + if not tokens: + return None + + digits: List[str] = [] + for token in tokens: + digit = WORD_NUMBER_MAP.get(token.lower()) + if digit is None: + return None + digits.append(digit) + + return "".join(digits) if digits else None + def _check_patterns( self, text: str ) -> Optional[Tuple[str, str, ContentFilterAction]]: @@ -407,10 +569,13 @@ class ContentFilterGuardrail(CustomGuardrail): Returns: Tuple of (matched_text, pattern_name, action) if match found, None otherwise """ - for compiled_pattern, pattern_name, action in self.compiled_patterns: - match = compiled_pattern.search(text) - if match: - matched_text = match.group(0) + for pattern_entry in self.compiled_patterns: + spans = self._find_pattern_spans(text, pattern_entry) + if spans: + start, end = spans[0] + matched_text = text[start:end] + pattern_name = pattern_entry["pattern_name"] + action = pattern_entry["action"] verbose_proxy_logger.debug( f"Pattern '{pattern_name}' matched: {matched_text[:20]}..." ) @@ -582,11 +747,13 @@ class ContentFilterGuardrail(CustomGuardrail): ) # Check regex patterns - process ALL patterns, not just first match - for compiled_pattern, pattern_name, action in self.compiled_patterns: - match = compiled_pattern.search(text) - if not match: + for pattern_entry in self.compiled_patterns: + spans = self._find_pattern_spans(text, pattern_entry) + if not spans: continue + pattern_name = pattern_entry["pattern_name"] + action = pattern_entry["action"] if detections is not None: # Don't log matched_text to avoid exposing sensitive content (emails, credit cards, etc.) pattern_detection: PatternDetection = { @@ -604,11 +771,10 @@ class ContentFilterGuardrail(CustomGuardrail): detail={"error": error_msg, "pattern": pattern_name}, ) elif action == ContentFilterAction.MASK: - # Replace ALL matches of this pattern with redaction tag redaction_tag = self.pattern_redaction_format.format( pattern_name=pattern_name.upper() ) - text = compiled_pattern.sub(redaction_tag, text) + text = self._mask_spans(text, spans, redaction_tag) verbose_proxy_logger.info( f"Masked all {pattern_name} matches in content" ) @@ -924,19 +1090,28 @@ class ContentFilterGuardrail(CustomGuardrail): if pattern_match: matched_text, pattern_name, action = pattern_match if action == ContentFilterAction.BLOCK: - error_msg = f"Content blocked: {pattern_name} pattern detected" + error_msg = ( + f"Content blocked: {pattern_name} pattern detected" + ) verbose_proxy_logger.warning(error_msg) raise HTTPException( status_code=403, - detail={"error": error_msg, "pattern": pattern_name}, + detail={ + "error": error_msg, + "pattern": pattern_name, + }, ) # Check blocked words - blocked_word_match = self._check_blocked_words(accumulated_content) + blocked_word_match = self._check_blocked_words( + accumulated_content + ) if blocked_word_match: keyword, action, description = blocked_word_match if action == ContentFilterAction.BLOCK: - error_msg = f"Content blocked: keyword '{keyword}' detected" + error_msg = ( + f"Content blocked: keyword '{keyword}' detected" + ) if description: error_msg += f" ({description})" verbose_proxy_logger.warning(error_msg) diff --git a/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.json b/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.json index d8ec22f81a1..f2427b5b920 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.json +++ b/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.json @@ -120,11 +120,11 @@ "description": "Detects URLs (http/https)" }, { - "name": "passport_us", - "display_name": "Passport (US)", - "pattern": "\\b[0-9]{9}\\b", - "category": "PII Patterns", - "description": "US passport numbers (9 digits)" + "name": "passport_us", + "display_name": "Passport (US)", + "pattern": "\\b[0-9]{9}\\b", + "category": "PII Patterns", + "description": "US passport numbers (9 digits)" }, { "name": "passport_uk", @@ -203,7 +203,6 @@ "category": "Protected Class - Fair Lending", "description": "Detects race, ethnicity and national origin terms - protected under ECOA and Fair Housing Act" }, - { "name": "religion", "display_name": "Religion & Creed (Protected Class)", @@ -236,7 +235,7 @@ "name": "military_status", "display_name": "Military Status (Protected Class)", "pattern": "\\b(veteran|military|armed\\s+forces|army|navy|air\\s+force|marine(s|\\s+corps)?|coast\\s+guard|national\\s+guard|reserve(s|ist)?|active\\s+duty|deployment|deployed|enlisted|commissioned|honorable\\s+discharge|dishonorable\\s+discharge|VA\\s+benefits|GI\\s+bill|military\\s+service|service\\s+member|servicemember|SCRA|MLA|military\\s+lending)\\b", - "category": "Protected Class - Fair Lending", + "category": "Protected Class - Fair Lending", "description": "Detects military status terms - protected under SCRA and MLA" }, { @@ -245,7 +244,7 @@ "pattern": "\\b(welfare|public\\s+assistance|food\\s+stamps|SNAP|WIC|TANF|medicaid|section\\s+8|housing\\s+voucher|subsidized\\s+housing|public\\s+housing|government\\s+benefits|social\\s+services|unemployment\\s+(benefits|insurance)|UI\\s+benefits|EBT|benefit\\s+recipient)\\b", "category": "Protected Class - Fair Lending", "description": "Detects public assistance terms - protected under ECOA" - } , + }, { "name": "weapons_firearms", "display_name": "Weapons & Firearms", @@ -313,10 +312,12 @@ { "name": "nl_bsn_contextual", "display_name": "BSN (Dutch Citizen Service Number)", - "pattern": "\\b(?:BSN|B\\.S\\.N\\.|burgerservicenummer|burger\\s*service\\s*nummer|sofi\\s*nummer|sofinummer|persoonsnummer|identificatienummer|citizen\\s*service\\s*number)[:\\s]*[0-9]{9}\\b|\\b[0-9]{9}\\b(?=\\s*(?:BSN|burgerservicenummer|sofinummer))", + "pattern": "\\b[0-9]{9}\\b", "category": "PII Patterns", "action": "MASK", - "description": "Detects Dutch BSN numbers with contextual keywords" + "description": "Detects Dutch BSN numbers with contextual keywords", + "keyword_pattern": "(?:\\b(?:BSN|B\\.S\\.N\\.|burgerservicenummer|burger\\s*service\\s*nummer|sofi\\s*nummer|sofinummer|persoonsnummer|identificatienummer|citizen\\s*service\\s*number)\\b|8\\s*5\\s*\\|\\\\\\|)", + "allow_word_numbers": true }, { "name": "br_cpf", @@ -369,5 +370,3 @@ } ] } - - diff --git a/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.py b/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.py index 776cf5bd8d2..d3a66690a90 100644 --- a/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.py +++ b/litellm/proxy/guardrails/guardrail_hooks/litellm_content_filter/patterns.py @@ -9,7 +9,7 @@ import json import os import re from enum import Enum -from typing import Dict, List, Pattern +from typing import Any, Dict, List, Pattern def _load_patterns_from_json() -> Dict: @@ -41,6 +41,26 @@ PREBUILT_PATTERNS: Dict[str, str] = { } +# Capture any extra configuration declared per pattern (e.g., contextual keywords) +KNOWN_PATTERN_KEYS = { + "name", + "display_name", + "pattern", + "category", + "action", + "description", +} + +PATTERN_EXTRA_CONFIG: Dict[str, Dict[str, Any]] = {} +for pattern_data in _PATTERNS_DATA["patterns"]: + extra_config = { + key: value + for key, value in pattern_data.items() + if key not in KNOWN_PATTERN_KEYS + } + PATTERN_EXTRA_CONFIG[pattern_data["name"]] = extra_config + + def get_compiled_pattern(pattern_name: str) -> Pattern: """ Get a compiled regex pattern by name. diff --git a/tests/test_litellm/proxy/guardrails/guardrail_hooks/content_filter/test_content_filter.py b/tests/test_litellm/proxy/guardrails/guardrail_hooks/content_filter/test_content_filter.py index 474d2a30036..265bc530dc2 100644 --- a/tests/test_litellm/proxy/guardrails/guardrail_hooks/content_filter/test_content_filter.py +++ b/tests/test_litellm/proxy/guardrails/guardrail_hooks/content_filter/test_content_filter.py @@ -14,7 +14,6 @@ sys.path.insert( from fastapi import HTTPException -import litellm from litellm.proxy.guardrails.guardrail_hooks.litellm_content_filter.content_filter import ( ContentFilterGuardrail, ) From e7cc53f217d4d2d4412af6c013aab178a556ef2a Mon Sep 17 00:00:00 2001 From: Harshit Jain <48647625+Harshit28j@users.noreply.github.com> Date: Wed, 14 Jan 2026 22:12:04 +0530 Subject: [PATCH 02/73] fix(dynamic_rate_limiter_v3): fix TPM 25% limiting by ensuring priority logic only runs when configured (#19092) --- .../proxy/hooks/dynamic_rate_limiter_v3.py | 35 +++++++++++-------- 1 file changed, 21 insertions(+), 14 deletions(-) diff --git a/litellm/proxy/hooks/dynamic_rate_limiter_v3.py b/litellm/proxy/hooks/dynamic_rate_limiter_v3.py index 755f5fdc201..a659d62e3eb 100644 --- a/litellm/proxy/hooks/dynamic_rate_limiter_v3.py +++ b/litellm/proxy/hooks/dynamic_rate_limiter_v3.py @@ -114,25 +114,25 @@ class _PROXY_DynamicRateLimitHandlerV3(CustomLogger): ) -> Optional[str]: """ Get priority from user_api_key_dict. - + Checks team metadata first (takes precedence), then falls back to key metadata. - + Args: user_api_key_dict: User authentication info - + Returns: Priority string if found, None otherwise """ priority: Optional[str] = None - + # Check team metadata first (takes precedence) if user_api_key_dict.team_metadata is not None: priority = user_api_key_dict.team_metadata.get("priority", None) - + # Fall back to key metadata if priority is None: priority = user_api_key_dict.metadata.get("priority", None) - + return priority def _normalize_priority_weights( @@ -299,10 +299,13 @@ class _PROXY_DynamicRateLimitHandlerV3(CustomLogger): """ descriptors: List[RateLimitDescriptor] = [] + if litellm.priority_reservation is None: + return descriptors + # Get model group info - model_group_info: Optional[ModelGroupInfo] = ( - self.llm_router.get_model_group_info(model_group=model) - ) + model_group_info: Optional[ + ModelGroupInfo + ] = self.llm_router.get_model_group_info(model_group=model) if model_group_info is None: return descriptors @@ -577,9 +580,9 @@ class _PROXY_DynamicRateLimitHandlerV3(CustomLogger): ) # Get model configuration - model_group_info: Optional[ModelGroupInfo] = ( - self.llm_router.get_model_group_info(model_group=model) - ) + model_group_info: Optional[ + ModelGroupInfo + ] = self.llm_router.get_model_group_info(model_group=model) if model_group_info is None: verbose_proxy_logger.debug( f"No model group info for {model}, allowing request" @@ -703,7 +706,9 @@ class _PROXY_DynamicRateLimitHandlerV3(CustomLogger): # Get priority from user_api_key_auth_metadata in standard_logging_metadata # This is where user_api_key_dict.metadata is stored during pre-call - user_api_key_auth_metadata = standard_logging_metadata.get("user_api_key_auth_metadata") or {} + user_api_key_auth_metadata = ( + standard_logging_metadata.get("user_api_key_auth_metadata") or {} + ) key_priority: Optional[str] = user_api_key_auth_metadata.get("priority") # Get total tokens from response @@ -775,7 +780,9 @@ class _PROXY_DynamicRateLimitHandlerV3(CustomLogger): # Only log 'priority' if it's known safe; otherwise, redact. SAFE_PRIORITIES = {"low", "medium", "high", "default"} - logged_priority = key_priority if key_priority in SAFE_PRIORITIES else "REDACTED" + logged_priority = ( + key_priority if key_priority in SAFE_PRIORITIES else "REDACTED" + ) verbose_proxy_logger.debug( f"[Dynamic Rate Limiter] Incremented tokens by {total_tokens} for " f"model={model_group}, priority={logged_priority}" From e8c4cad8851088103177aee1b68a7f1dbb3301ab Mon Sep 17 00:00:00 2001 From: Harshit Jain <48647625+Harshit28j@users.noreply.github.com> Date: Wed, 14 Jan 2026 22:14:48 +0530 Subject: [PATCH 03/73] feat(proxy): cleanup spend logs cron verification, fix, and docs (#19085) --- .../docs/proxy/spend_logs_deletion.md | 12 ++ litellm/proxy/proxy_server.py | 115 ++++++++++++------ .../proxy/test_spend_log_cleanup.py | 108 ++++++++++++++++ 3 files changed, 199 insertions(+), 36 deletions(-) diff --git a/docs/my-website/docs/proxy/spend_logs_deletion.md b/docs/my-website/docs/proxy/spend_logs_deletion.md index 05627c07741..b021457173f 100644 --- a/docs/my-website/docs/proxy/spend_logs_deletion.md +++ b/docs/my-website/docs/proxy/spend_logs_deletion.md @@ -30,6 +30,9 @@ general_settings: # Optional: set how frequently cleanup should run - default is daily maximum_spend_logs_retention_interval: "1d" # Run cleanup daily + # Optional: set exact time for cleanup (Cron syntax) + maximum_spend_logs_cleanup_cron: "0 4 * * *" # Run at 04:00 AM daily + litellm_settings: cache: true cache_params: @@ -51,6 +54,15 @@ How long logs should be kept before deletion. Supported formats: How often the cleanup job should run. Uses the same format as above. If not set, cleanup will run every 24 hours if and only if `maximum_spend_logs_retention_period` is set. +#### `maximum_spend_logs_cleanup_cron` (optional) + +Schedule the cleanup using standard cron syntax. This takes precedence over `maximum_spend_logs_retention_interval`. + +Examples: +- `"0 4 * * *"` – Run at 04:00 AM daily +- `"0 0 * * 0"` – Run at midnight every Sunday +- `"*/30 * * * *"` – Run every 30 minutes + ## How it works ### Step 1. Lock Acquisition (Optional with Redis) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 1dfaf78cb5b..1b01ac3a304 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -3240,20 +3240,22 @@ class ProxyConfig: ) -> Optional[dict]: """ Get router_settings in priority order: Key > Team > Global - + Returns: dict: Combined router_settings, or None if no settings found """ if prisma_client is None: return None - + import json import yaml - + # 1. Try key-level router_settings if user_api_key_dict is not None: # Check if router_settings is available on the key object - key_router_settings_value = getattr(user_api_key_dict, "router_settings", None) + key_router_settings_value = getattr( + user_api_key_dict, "router_settings", None + ) if key_router_settings_value is not None: key_router_settings = None if isinstance(key_router_settings_value, str): @@ -3266,11 +3268,15 @@ class ProxyConfig: pass elif isinstance(key_router_settings_value, dict): key_router_settings = key_router_settings_value - + # If key has router_settings (non-empty dict), use it - if key_router_settings is not None and isinstance(key_router_settings, dict) and key_router_settings: + if ( + key_router_settings is not None + and isinstance(key_router_settings, dict) + and key_router_settings + ): return key_router_settings - + # 2. Try team-level router_settings if user_api_key_dict is not None and user_api_key_dict.team_id is not None: try: @@ -3278,37 +3284,51 @@ class ProxyConfig: where={"team_id": user_api_key_dict.team_id} ) if team_obj is not None: - team_router_settings_value = getattr(team_obj, "router_settings", None) + team_router_settings_value = getattr( + team_obj, "router_settings", None + ) if team_router_settings_value is not None: team_router_settings = None if isinstance(team_router_settings_value, str): try: - team_router_settings = yaml.safe_load(team_router_settings_value) + team_router_settings = yaml.safe_load( + team_router_settings_value + ) except (yaml.YAMLError, json.JSONDecodeError): try: - team_router_settings = json.loads(team_router_settings_value) + team_router_settings = json.loads( + team_router_settings_value + ) except json.JSONDecodeError: pass elif isinstance(team_router_settings_value, dict): team_router_settings = team_router_settings_value - + # If team has router_settings (non-empty dict), use it - if team_router_settings is not None and isinstance(team_router_settings, dict) and team_router_settings: + if ( + team_router_settings is not None + and isinstance(team_router_settings, dict) + and team_router_settings + ): return team_router_settings except Exception: # If team lookup fails, continue to global settings pass - + # 3. Try global router_settings try: db_router_settings = await prisma_client.db.litellm_config.find_first( where={"param_name": "router_settings"} ) - if db_router_settings is not None and isinstance(db_router_settings.param_value, dict) and db_router_settings.param_value: + if ( + db_router_settings is not None + and isinstance(db_router_settings.param_value, dict) + and db_router_settings.param_value + ): return db_router_settings.param_value except Exception: pass - + return None async def _add_router_settings_from_db_config( @@ -4675,27 +4695,48 @@ class ProxyStartupEvent: ### SPEND LOG CLEANUP ### if general_settings.get("maximum_spend_logs_retention_period") is not None: spend_log_cleanup = SpendLogCleanup() - # Get the interval from config or default to 1 day - retention_interval = general_settings.get( - "maximum_spend_logs_retention_interval", "1d" - ) - try: - interval_seconds = duration_in_seconds(retention_interval) - scheduler.add_job( - spend_log_cleanup.cleanup_old_spend_logs, - "interval", - seconds=interval_seconds - + random.randint(0, 60), # Add small random offset - # REMOVED jitter parameter - major cause of memory leak - args=[prisma_client], - id="spend_log_cleanup_job", - replace_existing=True, - misfire_grace_time=APSCHEDULER_MISFIRE_GRACE_TIME, - ) - except ValueError: - verbose_proxy_logger.error( - "Invalid maximum_spend_logs_retention_interval value" + cleanup_cron = general_settings.get("maximum_spend_logs_cleanup_cron") + + if cleanup_cron: + from apscheduler.triggers.cron import CronTrigger + + try: + cron_trigger = CronTrigger.from_crontab(cleanup_cron) + scheduler.add_job( + spend_log_cleanup.cleanup_old_spend_logs, + cron_trigger, + args=[prisma_client], + id="spend_log_cleanup_job", + replace_existing=True, + misfire_grace_time=APSCHEDULER_MISFIRE_GRACE_TIME, + ) + verbose_proxy_logger.info( + f"Spend log cleanup scheduled with cron: {cleanup_cron}" + ) + except ValueError: + verbose_proxy_logger.error( + f"Invalid maximum_spend_logs_cleanup_cron value: {cleanup_cron}" + ) + else: + # Interval-based scheduling (existing behavior) + retention_interval = general_settings.get( + "maximum_spend_logs_retention_interval", "1d" ) + try: + interval_seconds = duration_in_seconds(retention_interval) + scheduler.add_job( + spend_log_cleanup.cleanup_old_spend_logs, + "interval", + seconds=interval_seconds + random.randint(0, 60), + args=[prisma_client], + id="spend_log_cleanup_job", + replace_existing=True, + misfire_grace_time=APSCHEDULER_MISFIRE_GRACE_TIME, + ) + except ValueError: + verbose_proxy_logger.error( + "Invalid maximum_spend_logs_retention_interval value" + ) ### CHECK BATCH COST ### if llm_router is not None: try: @@ -9885,7 +9926,9 @@ async def get_config(): # noqa: PLR0915 _success_callbacks = normalize_callback(_success_callbacks) _failure_callbacks = normalize_callback(_failure_callbacks) - _success_and_failure_callbacks = normalize_callback(_success_and_failure_callbacks) + _success_and_failure_callbacks = normalize_callback( + _success_and_failure_callbacks + ) _data_to_return = [] """ diff --git a/tests/test_litellm/proxy/test_spend_log_cleanup.py b/tests/test_litellm/proxy/test_spend_log_cleanup.py index 6aa18c560c8..1ffbb83caef 100644 --- a/tests/test_litellm/proxy/test_spend_log_cleanup.py +++ b/tests/test_litellm/proxy/test_spend_log_cleanup.py @@ -10,6 +10,114 @@ import pytest from litellm.proxy.db.db_transaction_queue.spend_log_cleanup import SpendLogCleanup +def test_spend_log_cleanup_cron_scheduling(): + """Test that cron expressions are correctly parsed for spend log cleanup scheduling""" + from apscheduler.triggers.cron import CronTrigger + + # Valid cron expressions + cron_expr = "0 4 * * *" # 4:00 AM daily + trigger = CronTrigger.from_crontab(cron_expr) + assert trigger is not None + + # Every minute (useful for testing) + trigger_minute = CronTrigger.from_crontab("*/1 * * * *") + assert trigger_minute is not None + + # Specific day and hour + trigger_weekly = CronTrigger.from_crontab("0 3 * * 0") # 3 AM every Sunday + assert trigger_weekly is not None + + # Invalid cron expression should raise ValueError + with pytest.raises(ValueError): + CronTrigger.from_crontab("invalid cron") + + with pytest.raises(ValueError): + CronTrigger.from_crontab("60 25 * * *") # Invalid minute and hour + + +def test_spend_log_cleanup_cron_scheduler_integration(): + """ + Integration test: Verify the proxy_server scheduler logic correctly adds + cron-based cleanup job when maximum_spend_logs_cleanup_cron is configured. + + This tests the logic in proxy_server.py lines 4671-4717 without requiring + a real database connection. + """ + from unittest.mock import MagicMock + from apscheduler.triggers.cron import CronTrigger + + # Mock scheduler + mock_scheduler = MagicMock() + mock_prisma_client = MagicMock() + mock_cleanup_instance = MagicMock() + + # Test Case 1: Cron-based scheduling + general_settings_cron = { + "maximum_spend_logs_retention_period": "7d", + "maximum_spend_logs_cleanup_cron": "0 4 * * *", # 4 AM daily + } + + cleanup_cron = general_settings_cron.get("maximum_spend_logs_cleanup_cron") + assert cleanup_cron is not None + + # Simulate the scheduler logic from proxy_server.py + cron_trigger = CronTrigger.from_crontab(cleanup_cron) + mock_scheduler.add_job( + mock_cleanup_instance.cleanup_old_spend_logs, + cron_trigger, + args=[mock_prisma_client], + id="spend_log_cleanup_job", + replace_existing=True, + misfire_grace_time=3600, + ) + + # Verify scheduler was called correctly + mock_scheduler.add_job.assert_called_once() + call_args = mock_scheduler.add_job.call_args + + # Verify the trigger is a CronTrigger + assert isinstance(call_args[0][1], CronTrigger) + + # Verify job ID + assert call_args[1]["id"] == "spend_log_cleanup_job" + assert call_args[1]["replace_existing"] is True + + # Test Case 2: Interval-based scheduling (fallback) + mock_scheduler.reset_mock() + general_settings_interval = { + "maximum_spend_logs_retention_period": "7d", + # No cron, so it should fall back to interval + } + + cleanup_cron_fallback = general_settings_interval.get( + "maximum_spend_logs_cleanup_cron" + ) + assert cleanup_cron_fallback is None # No cron configured + + # Simulate interval-based scheduling fallback + retention_interval = general_settings_interval.get( + "maximum_spend_logs_retention_interval", "1d" + ) + from litellm.litellm_core_utils.duration_parser import duration_in_seconds + + interval_seconds = duration_in_seconds(retention_interval) + + mock_scheduler.add_job( + mock_cleanup_instance.cleanup_old_spend_logs, + "interval", + seconds=interval_seconds, + args=[mock_prisma_client], + id="spend_log_cleanup_job", + replace_existing=True, + ) + + # Verify interval scheduling was called + mock_scheduler.add_job.assert_called_once() + interval_call_args = mock_scheduler.add_job.call_args + assert interval_call_args[0][1] == "interval" + assert interval_call_args[1]["seconds"] == 86400 # 1 day in seconds + + @pytest.mark.asyncio async def test_should_delete_spend_logs(): # Test case 1: No retention set From 1391e419166b7ebf07e880fee53e1ed6aae323d9 Mon Sep 17 00:00:00 2001 From: Kris Xia Date: Thu, 15 Jan 2026 00:47:43 +0800 Subject: [PATCH 04/73] fix(vertex_ai): improve passthrough endpoint url parsing and construction (#17402) (#17526) * fix(vertex_ai): improve passthrough endpoint url parsing and construction (#17402) * test(proxy): add test for vertex passthrough load balancing Add a test that verifies _base_vertex_proxy_route uses get_available_deployment for proper load balancing instead of get_model_list. This ensures the correct deployment is selected from the router and vertex credentials are properly fetched. Also refactor the implementation to: - Use get_available_deployment instead of get_model_list - Add error handling for deployment retrieval - Improve code structure with try-except block * feat(proxy): add pass-through deployment filtering methods Add dedicated methods to filter and select deployments for pass-through endpoints: - Implement get_available_deployment_for_pass_through() to ensure only deployments with use_in_pass_through=True are considered - Implement async_get_available_deployment_for_pass_through() for async operations - Add _filter_pass_through_deployments() helper method to filter by use_in_pass_through flag - Update vertex pass-through route to use the new dedicated method This ensures pass-through endpoints respect the use_in_pass_through configuration and apply proper load balancing strategy only to configured deployments. Add comprehensive tests to verify filtering and load balancing behavior. --- litellm/llms/vertex_ai/common_utils.py | 19 + .../llm_passthrough_endpoints.py | 20 ++ litellm/router.py | 339 ++++++++++++++++++ .../vertex_ai/test_vertex_ai_common_utils.py | 57 ++- .../test_vertex_passthrough_load_balancing.py | 222 ++++++++++++ 5 files changed, 650 insertions(+), 7 deletions(-) create mode 100644 tests/test_litellm/proxy/pass_through_endpoints/test_vertex_passthrough_load_balancing.py diff --git a/litellm/llms/vertex_ai/common_utils.py b/litellm/llms/vertex_ai/common_utils.py index 2aa6a00c72b..5ccbb8cd088 100644 --- a/litellm/llms/vertex_ai/common_utils.py +++ b/litellm/llms/vertex_ai/common_utils.py @@ -771,6 +771,16 @@ def get_vertex_location_from_url(url: str) -> Optional[str]: return match.group(1) if match else None +def get_vertex_model_id_from_url(url: str) -> Optional[str]: + """ + Get the vertex model id from the url + + `https://${LOCATION}-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/${LOCATION}/publishers/google/models/${MODEL_ID}:streamGenerateContent` + """ + match = re.search(r"/models/([^/:]+)", url) + return match.group(1) if match else None + + def replace_project_and_location_in_route( requested_route: str, vertex_project: str, vertex_location: str ) -> str: @@ -820,6 +830,15 @@ def construct_target_url( if "cachedContent" in requested_route: vertex_version = "v1beta1" + # Check if the requested route starts with a version + # e.g. /v1beta1/publishers/google/models/gemini-3-pro-preview:streamGenerateContent + if requested_route.startswith("/v1/"): + vertex_version = "v1" + requested_route = requested_route.replace("/v1/", "/", 1) + elif requested_route.startswith("/v1beta1/"): + vertex_version = "v1beta1" + requested_route = requested_route.replace("/v1beta1/", "/", 1) + base_requested_route = "{}/projects/{}/locations/{}".format( vertex_version, vertex_project, vertex_location ) diff --git a/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py b/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py index 5299b30b52f..e48fd22bc8d 100644 --- a/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py +++ b/litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py @@ -1555,6 +1555,7 @@ async def _base_vertex_proxy_route( from litellm.llms.vertex_ai.common_utils import ( construct_target_url, get_vertex_location_from_url, + get_vertex_model_id_from_url, get_vertex_project_id_from_url, ) @@ -1584,6 +1585,25 @@ async def _base_vertex_proxy_route( vertex_location=vertex_location, ) + if vertex_project is None or vertex_location is None: + # Check if model is in router config + model_id = get_vertex_model_id_from_url(endpoint) + if model_id: + from litellm.proxy.proxy_server import llm_router + + if llm_router: + try: + # Use the dedicated pass-through deployment selection method to automatically filter use_in_pass_through=True + deployment = llm_router.get_available_deployment_for_pass_through(model=model_id) + if deployment: + litellm_params = deployment.get("litellm_params", {}) + vertex_project = litellm_params.get("vertex_project") + vertex_location = litellm_params.get("vertex_location") + except Exception as e: + verbose_proxy_logger.debug( + f"Error getting available deployment for model {model_id}: {e}" + ) + vertex_credentials = passthrough_endpoint_router.get_vertex_credentials( project_id=vertex_project, location=vertex_location, diff --git a/litellm/router.py b/litellm/router.py index 638df49ac05..6523b5513af 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -7994,6 +7994,154 @@ class Router: ) raise e + async def async_get_available_deployment_for_pass_through( + self, + model: str, + request_kwargs: Dict, + messages: Optional[List[Dict[str, str]]] = None, + input: Optional[Union[str, List]] = None, + specific_deployment: Optional[bool] = False, + ): + """ + Async version of get_available_deployment_for_pass_through + + Only returns deployments configured with use_in_pass_through=True + """ + try: + parent_otel_span = _get_parent_otel_span_from_kwargs(request_kwargs) + + # 1. Execute pre-routing hook + pre_routing_hook_response = await self.async_pre_routing_hook( + model=model, + request_kwargs=request_kwargs, + messages=messages, + input=input, + specific_deployment=specific_deployment, + ) + if pre_routing_hook_response is not None: + model = pre_routing_hook_response.model + messages = pre_routing_hook_response.messages + + # 2. Get healthy deployments + healthy_deployments = await self.async_get_healthy_deployments( + model=model, + request_kwargs=request_kwargs, + messages=messages, + input=input, + specific_deployment=specific_deployment, + parent_otel_span=parent_otel_span, + ) + + # 3. If specific deployment returned, verify if it supports pass-through + if isinstance(healthy_deployments, dict): + litellm_params = healthy_deployments.get("litellm_params", {}) + if litellm_params.get("use_in_pass_through"): + return healthy_deployments + else: + raise litellm.BadRequestError( + message=f"Deployment {healthy_deployments.get('model_info', {}).get('id')} does not support pass-through endpoint (use_in_pass_through=False)", + model=model, + llm_provider="", + ) + + # 4. Filter deployments that support pass-through + pass_through_deployments = self._filter_pass_through_deployments( + healthy_deployments=healthy_deployments + ) + + if len(pass_through_deployments) == 0: + raise litellm.BadRequestError( + message=f"Model {model} has no deployments configured with use_in_pass_through=True. Please add use_in_pass_through: true to the deployment configuration", + model=model, + llm_provider="", + ) + + # 5. Apply load balancing strategy + start_time = time.perf_counter() + if ( + self.routing_strategy == "usage-based-routing-v2" + and self.lowesttpm_logger_v2 is not None + ): + deployment = ( + await self.lowesttpm_logger_v2.async_get_available_deployments( + model_group=model, + healthy_deployments=pass_through_deployments, # type: ignore + messages=messages, + input=input, + ) + ) + elif ( + self.routing_strategy == "latency-based-routing" + and self.lowestlatency_logger is not None + ): + deployment = ( + await self.lowestlatency_logger.async_get_available_deployments( + model_group=model, + healthy_deployments=pass_through_deployments, # type: ignore + messages=messages, + input=input, + request_kwargs=request_kwargs, + ) + ) + elif self.routing_strategy == "simple-shuffle": + return simple_shuffle( + llm_router_instance=self, + healthy_deployments=pass_through_deployments, + model=model, + ) + elif ( + self.routing_strategy == "least-busy" + and self.leastbusy_logger is not None + ): + deployment = ( + await self.leastbusy_logger.async_get_available_deployments( + model_group=model, + healthy_deployments=pass_through_deployments, # type: ignore + ) + ) + else: + deployment = None + + if deployment is None: + exception = await async_raise_no_deployment_exception( + litellm_router_instance=self, + model=model, + parent_otel_span=parent_otel_span, + ) + raise exception + + verbose_router_logger.info( + f"async_get_available_deployment_for_pass_through model: {model}, selected deployment: {self.print_deployment(deployment)}" + ) + + end_time = time.perf_counter() + _duration = end_time - start_time + asyncio.create_task( + self.service_logger_obj.async_service_success_hook( + service=ServiceTypes.ROUTER, + duration=_duration, + call_type=".async_get_available_deployments", + parent_otel_span=parent_otel_span, + start_time=start_time, + end_time=end_time, + ) + ) + + return deployment + except Exception as e: + traceback_exception = traceback.format_exc() + if request_kwargs is not None: + logging_obj = request_kwargs.get("litellm_logging_obj", None) + if logging_obj is not None: + threading.Thread( + target=logging_obj.failure_handler, + args=(e, traceback_exception), + ).start() + asyncio.create_task( + logging_obj.async_failure_handler(e, traceback_exception) # type: ignore + ) + raise e + async def async_pre_routing_hook( self, model: str, @@ -8146,6 +8294,169 @@ class Router: ) return deployment + def get_available_deployment_for_pass_through( + self, + model: str, + messages: Optional[List[Dict[str, str]]] = None, + input: Optional[Union[str, List]] = None, + specific_deployment: Optional[bool] = False, + request_kwargs: Optional[Dict] = None, + ): + """ + Returns deployments available for pass-through endpoints (based on load balancing strategy) + + Similar to get_available_deployment, but only returns deployments with use_in_pass_through=True + + Args: + model: Model name + messages: Optional list of messages + input: Optional input data + specific_deployment: Whether to find a specific deployment + request_kwargs: Optional request parameters + + Returns: + Dict: Selected deployment configuration + + Raises: + BadRequestError: If no deployment is configured with use_in_pass_through=True + RouterRateLimitError: If no pass-through deployments are available + """ + # 1. Perform common checks to get healthy deployments list + model, healthy_deployments = self._common_checks_available_deployment( + model=model, + messages=messages, + input=input, + specific_deployment=specific_deployment, + ) + + # 2. If the returned is a specific deployment (Dict), verify and return directly + if isinstance(healthy_deployments, dict): + litellm_params = healthy_deployments.get("litellm_params", {}) + if litellm_params.get("use_in_pass_through"): + return healthy_deployments + else: + # Specific deployment does not support pass-through + raise litellm.BadRequestError( + message=f"Deployment {healthy_deployments.get('model_info', {}).get('id')} does not support pass-through endpoint (use_in_pass_through=False)", + model=model, + llm_provider="", + ) + + # 3. Filter deployments that support pass-through + pass_through_deployments = self._filter_pass_through_deployments( + healthy_deployments=healthy_deployments + ) + + if len(pass_through_deployments) == 0: + # No deployments support pass-through + raise litellm.BadRequestError( + message=f"Model {model} has no deployment configured with use_in_pass_through=True. Please add use_in_pass_through: true in the deployment configuration", + model=model, + llm_provider="", + ) + + # 4. Apply cooldown filtering + parent_otel_span: Optional[Span] = _get_parent_otel_span_from_kwargs( + request_kwargs + ) + cooldown_deployments = _get_cooldown_deployments( + litellm_router_instance=self, parent_otel_span=parent_otel_span + ) + pass_through_deployments = self._filter_cooldown_deployments( + healthy_deployments=pass_through_deployments, + cooldown_deployments=cooldown_deployments, + ) + + # 5. Apply pre-call checks (if enabled) + if self.enable_pre_call_checks and messages is not None: + pass_through_deployments = self._pre_call_checks( + model=model, + healthy_deployments=pass_through_deployments, + messages=messages, + request_kwargs=request_kwargs, + ) + + if len(pass_through_deployments) == 0: + model_ids = self.get_model_ids(model_name=model) + _cooldown_time = self.cooldown_cache.get_min_cooldown( + model_ids=model_ids, parent_otel_span=parent_otel_span + ) + _cooldown_list = _get_cooldown_deployments( + litellm_router_instance=self, parent_otel_span=parent_otel_span + ) + raise RouterRateLimitError( + model=model, + cooldown_time=_cooldown_time, + enable_pre_call_checks=self.enable_pre_call_checks, + cooldown_list=_cooldown_list, + ) + + # 6. Apply load balancing strategy + if self.routing_strategy == "least-busy" and self.leastbusy_logger is not None: + deployment = self.leastbusy_logger.get_available_deployments( + model_group=model, healthy_deployments=pass_through_deployments # type: ignore + ) + elif self.routing_strategy == "simple-shuffle": + return simple_shuffle( + llm_router_instance=self, + healthy_deployments=pass_through_deployments, + model=model, + ) + elif ( + self.routing_strategy == "latency-based-routing" + and self.lowestlatency_logger is not None + ): + deployment = self.lowestlatency_logger.get_available_deployments( + model_group=model, + healthy_deployments=pass_through_deployments, # type: ignore + request_kwargs=request_kwargs, + ) + elif ( + self.routing_strategy == "usage-based-routing" + and self.lowesttpm_logger is not None + ): + deployment = self.lowesttpm_logger.get_available_deployments( + model_group=model, + healthy_deployments=pass_through_deployments, # type: ignore + messages=messages, + input=input, + ) + elif ( + self.routing_strategy == "usage-based-routing-v2" + and self.lowesttpm_logger_v2 is not None + ): + deployment = self.lowesttpm_logger_v2.get_available_deployments( + model_group=model, + healthy_deployments=pass_through_deployments, # type: ignore + messages=messages, + input=input, + ) + else: + deployment = None + + if deployment is None: + verbose_router_logger.info( + f"get_available_deployment_for_pass_through model: {model}, no available deployments" + ) + model_ids = self.get_model_ids(model_name=model) + _cooldown_time = self.cooldown_cache.get_min_cooldown( + model_ids=model_ids, parent_otel_span=parent_otel_span + ) + _cooldown_list = _get_cooldown_deployments( + litellm_router_instance=self, parent_otel_span=parent_otel_span + ) + raise RouterRateLimitError( + model=model, + cooldown_time=_cooldown_time, + enable_pre_call_checks=self.enable_pre_call_checks, + cooldown_list=_cooldown_list, + ) + + verbose_router_logger.info( + f"get_available_deployment_for_pass_through model: {model}, selected deployment: {self.print_deployment(deployment)}" + ) + return deployment + def _filter_cooldown_deployments( self, healthy_deployments: List[Dict], cooldown_deployments: List[str] ) -> List[Dict]: @@ -8168,6 +8479,34 @@ class Router: if deployment["model_info"]["id"] not in cooldown_set ] + def _filter_pass_through_deployments( + self, healthy_deployments: List[Dict] + ) -> List[Dict]: + """ + Filter out deployments configured with use_in_pass_through=True + + Args: + healthy_deployments: List of healthy deployments + + Returns: + List[Dict]: Only includes a list of deployments that support pass-through + """ + verbose_router_logger.debug( + f"Filter pass-through deployments from {len(healthy_deployments)} healthy deployments" + ) + + pass_through_deployments = [ + deployment + for deployment in healthy_deployments + if deployment.get("litellm_params", {}).get("use_in_pass_through", False) + ] + + verbose_router_logger.debug( + f"Found {len(pass_through_deployments)} deployments with pass-through enabled" + ) + + return pass_through_deployments + def _track_deployment_metrics( self, deployment, parent_otel_span: Optional[Span], response=None ): diff --git a/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py b/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py index b5637db3e52..12e35a47280 100644 --- a/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py +++ b/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py @@ -1,7 +1,6 @@ import os import sys -from typing import Any, Dict -from unittest.mock import MagicMock, call, patch +from unittest.mock import patch import pytest @@ -11,7 +10,6 @@ sys.path.insert( 0, os.path.abspath("../../..") ) # Adds the parent directory to the system path -import litellm from litellm.llms.vertex_ai.common_utils import ( _get_vertex_url, convert_anyof_null_to_nullable, @@ -798,9 +796,54 @@ def test_fix_enum_empty_strings(): assert "mobile" in enum_values assert "tablet" in enum_values - # 3. Other properties preserved - assert input_schema["properties"]["user_agent_type"]["type"] == "string" - assert input_schema["properties"]["user_agent_type"]["description"] == "Device type for user agent" + +def test_get_vertex_model_id_from_url(): + """Test get_vertex_model_id_from_url with various URLs""" + from litellm.llms.vertex_ai.common_utils import get_vertex_model_id_from_url + + # Test with valid URL + url = "https://us-central1-aiplatform.googleapis.com/v1/projects/test-project/locations/us-central1/publishers/google/models/gemini-pro:streamGenerateContent" + model_id = get_vertex_model_id_from_url(url) + assert model_id == "gemini-pro" + + # Test with invalid URL + url = "https://invalid-url.com" + model_id = get_vertex_model_id_from_url(url) + assert model_id is None + + +def test_construct_target_url_with_version_prefix(): + """Test construct_target_url with version prefixes""" + from litellm.llms.vertex_ai.common_utils import construct_target_url + + # Test with /v1/ prefix + url = "/v1/publishers/google/models/gemini-pro:streamGenerateContent" + vertex_project = "test-project" + vertex_location = "us-central1" + base_url = "https://us-central1-aiplatform.googleapis.com" + + target_url = construct_target_url( + base_url=base_url, + requested_route=url, + vertex_project=vertex_project, + vertex_location=vertex_location, + ) + + expected_url = "https://us-central1-aiplatform.googleapis.com/v1/projects/test-project/locations/us-central1/publishers/google/models/gemini-pro:streamGenerateContent" + assert str(target_url) == expected_url + + # Test with /v1beta1/ prefix + url = "/v1beta1/publishers/google/models/gemini-pro:streamGenerateContent" + + target_url = construct_target_url( + base_url=base_url, + requested_route=url, + vertex_project=vertex_project, + vertex_location=vertex_location, + ) + + expected_url = "https://us-central1-aiplatform.googleapis.com/v1beta1/projects/test-project/locations/us-central1/publishers/google/models/gemini-pro:streamGenerateContent" + assert str(target_url) == expected_url def test_fix_enum_types(): @@ -862,7 +905,7 @@ def test_fix_enum_types(): "truncateMode": { "enum": ["auto", "none", "start", "end"], # Kept - string type "type": "string", - "description": "How to truncate content" + "description": "How to truncate content", }, "maxLength": { # enum removed "type": "integer", diff --git a/tests/test_litellm/proxy/pass_through_endpoints/test_vertex_passthrough_load_balancing.py b/tests/test_litellm/proxy/pass_through_endpoints/test_vertex_passthrough_load_balancing.py new file mode 100644 index 00000000000..ceb231eb4cb --- /dev/null +++ b/tests/test_litellm/proxy/pass_through_endpoints/test_vertex_passthrough_load_balancing.py @@ -0,0 +1,222 @@ + +import pytest +from unittest.mock import MagicMock, AsyncMock, patch +from litellm.proxy.pass_through_endpoints.llm_passthrough_endpoints import _base_vertex_proxy_route +from litellm.types.router import DeploymentTypedDict + +@pytest.mark.asyncio +async def test_vertex_passthrough_load_balancing(): + """ + Test that _base_vertex_proxy_route uses llm_router.get_available_deployment_for_pass_through + instead of get_model_list to ensure load balancing works with pass-through filtering. + """ + # Setup mocks + mock_request = MagicMock() + mock_response = MagicMock() + mock_handler = MagicMock() + + # Mock the router + mock_router = MagicMock() + mock_deployment = { + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "test-project-lb", + "vertex_location": "us-central1-lb", + "use_in_pass_through": True + } + } + mock_router.get_available_deployment_for_pass_through.return_value = mock_deployment + + # Mock get_vertex_model_id_from_url to return a model ID + with patch("litellm.llms.vertex_ai.common_utils.get_vertex_model_id_from_url", return_value="gemini-pro"), \ + patch("litellm.proxy.proxy_server.llm_router", mock_router), \ + patch("litellm.llms.vertex_ai.common_utils.get_vertex_project_id_from_url", return_value=None), \ + patch("litellm.llms.vertex_ai.common_utils.get_vertex_location_from_url", return_value=None), \ + patch("litellm.proxy.pass_through_endpoints.llm_passthrough_endpoints.passthrough_endpoint_router") as mock_pt_router, \ + patch("litellm.proxy.pass_through_endpoints.llm_passthrough_endpoints._prepare_vertex_auth_headers", new_callable=AsyncMock) as mock_prep_headers, \ + patch("litellm.proxy.pass_through_endpoints.llm_passthrough_endpoints.create_pass_through_route") as mock_create_route, \ + patch("litellm.proxy.pass_through_endpoints.llm_passthrough_endpoints.user_api_key_auth", new_callable=AsyncMock) as mock_auth: + + # Setup additional mocks to avoid side effects + mock_pt_router.get_vertex_credentials.return_value = MagicMock() + mock_prep_headers.return_value = ({}, "https://test.url", False, "test-project-lb", "us-central1-lb") + + mock_endpoint_func = AsyncMock() + mock_create_route.return_value = mock_endpoint_func + mock_auth.return_value = {} + + # Execute + await _base_vertex_proxy_route( + endpoint="https://us-central1-aiplatform.googleapis.com/v1/projects/my-project/locations/us-central1/publishers/google/models/gemini-pro:streamGenerateContent", + request=mock_request, + fastapi_response=mock_response, + get_vertex_pass_through_handler=mock_handler + ) + + # Verify + # 1. Check that get_available_deployment_for_pass_through was called with the correct model ID + mock_router.get_available_deployment_for_pass_through.assert_called_once_with(model="gemini-pro") + + # 2. Check that get_model_list was NOT called (this ensures we aren't doing the old logic) + mock_router.get_model_list.assert_not_called() + + # 3. Verify that the project and location from the deployment were used (passed to _prepare_vertex_auth_headers) + # The args are: request, vertex_credentials, router_credentials, vertex_project, vertex_location, ... + # We check the 4th and 5th args (index 3 and 4) + call_args = mock_prep_headers.call_args + assert call_args[1]['vertex_project'] == "test-project-lb" + assert call_args[1]['vertex_location'] == "us-central1-lb" + + +def test_get_available_deployment_for_pass_through_filters_correctly(): + """ + Test that get_available_deployment_for_pass_through filters deployments correctly + """ + from litellm.router import Router + + # Configure router with both pass-through and non-pass-through deployments + model_list = [ + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-1", + "vertex_location": "us-central1", + "use_in_pass_through": True, # Supports pass-through + } + }, + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-2", + "vertex_location": "us-west1", + "use_in_pass_through": False, # Does not support pass-through + } + }, + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-3", + "vertex_location": "us-east1", + # use_in_pass_through not set (defaults to False) + } + }, + ] + + router = Router(model_list=model_list, routing_strategy="simple-shuffle") + + # Test: Should only return project-1 (use_in_pass_through=True) + deployment = router.get_available_deployment_for_pass_through(model="gemini-pro") + + assert deployment is not None + assert deployment["litellm_params"]["vertex_project"] == "project-1" + assert deployment["litellm_params"]["use_in_pass_through"] is True + + +def test_get_available_deployment_for_pass_through_no_deployments(): + """ + Test that correct error is thrown when there are no pass-through deployments + """ + import litellm + from litellm.router import Router + + model_list = [ + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-1", + "vertex_location": "us-central1", + "use_in_pass_through": False, # Does not support pass-through + } + } + ] + + router = Router(model_list=model_list) + + # Should throw BadRequestError + with pytest.raises(litellm.BadRequestError) as exc_info: + router.get_available_deployment_for_pass_through(model="gemini-pro") + + assert "use_in_pass_through=True" in str(exc_info.value) + + +def test_get_available_deployment_for_pass_through_load_balancing(): + """ + Test load balancing for pass-through deployments + """ + from litellm.router import Router + + model_list = [ + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-1", + "vertex_location": "us-central1", + "use_in_pass_through": True, + "rpm": 100, + } + }, + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-2", + "vertex_location": "us-west1", + "use_in_pass_through": True, + "rpm": 200, # Higher RPM should be selected more frequently + } + }, + ] + + router = Router( + model_list=model_list, + routing_strategy="simple-shuffle" + ) + + # Call multiple times and track selected deployments + selections = {"project-1": 0, "project-2": 0} + for _ in range(100): + deployment = router.get_available_deployment_for_pass_through(model="gemini-pro") + project = deployment["litellm_params"]["vertex_project"] + selections[project] += 1 + + # Due to rpm weight, project-2 should be selected more times + assert selections["project-2"] > selections["project-1"] + + +@pytest.mark.asyncio +async def test_async_get_available_deployment_for_pass_through(): + """ + Test the async version of get_available_deployment_for_pass_through + """ + from litellm.router import Router + + model_list = [ + { + "model_name": "gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "project-1", + "vertex_location": "us-central1", + "use_in_pass_through": True, + } + } + ] + + router = Router( + model_list=model_list, + routing_strategy="simple-shuffle" + ) + + deployment = await router.async_get_available_deployment_for_pass_through( + model="gemini-pro", + request_kwargs={} + ) + + assert deployment is not None + assert deployment["litellm_params"]["use_in_pass_through"] is True + From d92a0168cc419d8fbbdcbb81e03f9b2e500fe6ed Mon Sep 17 00:00:00 2001 From: Rayan Pal <90289028+theonlypal@users.noreply.github.com> Date: Wed, 14 Jan 2026 09:28:05 -0800 Subject: [PATCH 05/73] fix: keep type field in Gemini schema when properties is empty (#18979) --- litellm/llms/vertex_ai/common_utils.py | 4 ++-- .../vertex_ai/test_gemini_empty_properties.py | 16 ++++++++++++++++ .../vertex_ai/test_vertex_ai_common_utils.py | 4 ++-- 3 files changed, 20 insertions(+), 4 deletions(-) create mode 100644 tests/test_litellm/llms/vertex_ai/test_gemini_empty_properties.py diff --git a/litellm/llms/vertex_ai/common_utils.py b/litellm/llms/vertex_ai/common_utils.py index 5ccbb8cd088..704e45e301d 100644 --- a/litellm/llms/vertex_ai/common_utils.py +++ b/litellm/llms/vertex_ai/common_utils.py @@ -660,11 +660,11 @@ def add_object_type(schema): if "required" in schema and schema["required"] is None: schema.pop("required", None) # Gemini doesn't accept empty properties for object types - # If properties is empty, remove it and the type field + # If properties is empty, remove it but keep type as object if not properties: schema.pop("properties", None) - schema.pop("type", None) schema.pop("required", None) + schema["type"] = "object" else: schema["type"] = "object" for name, value in properties.items(): diff --git a/tests/test_litellm/llms/vertex_ai/test_gemini_empty_properties.py b/tests/test_litellm/llms/vertex_ai/test_gemini_empty_properties.py new file mode 100644 index 00000000000..1a4e4d35ca9 --- /dev/null +++ b/tests/test_litellm/llms/vertex_ai/test_gemini_empty_properties.py @@ -0,0 +1,16 @@ +"""Test for Gemini schema handling with empty properties.""" + +import os +import sys + +sys.path.insert(0, os.path.abspath("../../../..")) + +from litellm.llms.vertex_ai.common_utils import add_object_type + + +def test_add_object_type_empty_properties_keeps_type(): + """Gemini requires type: object even when properties is empty.""" + schema = {"properties": {}, "type": "object"} + add_object_type(schema) + assert schema.get("type") == "object" + assert "properties" not in schema diff --git a/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py b/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py index 12e35a47280..19d8f174930 100644 --- a/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py +++ b/tests/test_litellm/llms/vertex_ai/test_vertex_ai_common_utils.py @@ -1297,8 +1297,8 @@ def test_build_vertex_schema_empty_properties(): # Verify empty properties was removed assert "properties" not in go_back_schema, "Empty properties should be removed" - # Verify type was also removed (since object without properties is invalid in Gemini) - assert "type" not in go_back_schema, "Type should be removed when properties is empty" + # Verify type is kept as object (Gemini requires type: object even without properties) + assert go_back_schema.get("type") == "object", "Type should be kept as object when properties is empty" # Verify required was also removed assert "required" not in go_back_schema, "Required should be removed when properties is empty" From eb49adb20180ac60d2333262c1eba5f7efff7c24 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Thu, 15 Jan 2026 18:36:06 +0530 Subject: [PATCH 06/73] Add user auth in standard logging object for bedrock passthrough --- litellm/litellm_core_utils/litellm_logging.py | 39 +++- .../test_standard_logging_payload.py | 185 ++++++++++++++++++ 2 files changed, 219 insertions(+), 5 deletions(-) diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py index 619c5d1cf00..15d578a7f99 100644 --- a/litellm/litellm_core_utils/litellm_logging.py +++ b/litellm/litellm_core_utils/litellm_logging.py @@ -4338,6 +4338,38 @@ class StandardLoggingPayloadSetup: return messages + @staticmethod + def merge_litellm_metadata(litellm_params: dict) -> dict: + """ + Merge both litellm_metadata and metadata from litellm_params. + + litellm_metadata contains model-related fields, metadata contains user API key fields. + We need both for complete standard logging payload. + + Args: + litellm_params: Dictionary containing metadata and litellm_metadata + + Returns: + dict: Merged metadata with user API key fields taking precedence + """ + merged_metadata: dict = {} + + # Start with metadata (user API key fields) - but skip non-serializable objects + if litellm_params.get("metadata") and isinstance(litellm_params.get("metadata"), dict): + for key, value in litellm_params["metadata"].items(): + # Skip non-serializable objects like UserAPIKeyAuth + if key == "user_api_key_auth": + continue + merged_metadata[key] = value + + # Then merge litellm_metadata (model-related fields) - this will NOT overwrite existing keys + if litellm_params.get("litellm_metadata") and isinstance(litellm_params.get("litellm_metadata"), dict): + for key, value in litellm_params["litellm_metadata"].items(): + if key not in merged_metadata: # Don't overwrite existing keys from metadata + merged_metadata[key] = value + + return merged_metadata + @staticmethod def get_standard_logging_metadata( metadata: Optional[Dict[str, Any]], @@ -5059,11 +5091,8 @@ def get_standard_logging_object_payload( litellm_params = kwargs.get("litellm_params", {}) or {} proxy_server_request = litellm_params.get("proxy_server_request") or {} - metadata: dict = ( - litellm_params.get("litellm_metadata") - or litellm_params.get("metadata", None) - or {} - ) + # Merge both litellm_metadata and metadata to get complete metadata + metadata: dict = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) completion_start_time = kwargs.get("completion_start_time", end_time) call_type = kwargs.get("call_type") diff --git a/tests/logging_callback_tests/test_standard_logging_payload.py b/tests/logging_callback_tests/test_standard_logging_payload.py index 4ead642c462..3d8ffbf1f7f 100644 --- a/tests/logging_callback_tests/test_standard_logging_payload.py +++ b/tests/logging_callback_tests/test_standard_logging_payload.py @@ -703,3 +703,188 @@ def test_cost_breakdown_missing_in_standard_logging_payload(): assert payload["response_cost"] == 0.0001 print("✅ Cost breakdown missing test passed!") + + +def test_merge_litellm_metadata_basic(): + """ + Test that merge_litellm_metadata correctly merges metadata and litellm_metadata. + User API key fields (from metadata) should take precedence over model-related fields (from litellm_metadata). + """ + litellm_params = { + "metadata": { + "user_api_key": "test-key-123", + "user_api_key_user_id": "user-456", + "user_api_key_team_id": "team-789", + }, + "litellm_metadata": { + "model_group": "gpt-4-group", + "model_info": {"id": "model-123"}, + "tags": ["tag1", "tag2"], + }, + } + + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + + # Check that user API key fields are present + assert result["user_api_key"] == "test-key-123" + assert result["user_api_key_user_id"] == "user-456" + assert result["user_api_key_team_id"] == "team-789" + + # Check that model-related fields are present + assert result["model_group"] == "gpt-4-group" + assert result["model_info"] == {"id": "model-123"} + assert result["tags"] == ["tag1", "tag2"] + + +def test_merge_litellm_metadata_precedence(): + """ + Test that metadata fields take precedence over litellm_metadata when there are conflicts. + """ + litellm_params = { + "metadata": { + "tags": ["user-tag1", "user-tag2"], + "custom_field": "from_metadata", + }, + "litellm_metadata": { + "tags": ["model-tag1", "model-tag2"], # This should NOT overwrite + "custom_field": "from_litellm_metadata", # This should NOT overwrite + "model_group": "gpt-4-group", # This should be included + }, + } + + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + + # metadata values should take precedence + assert result["tags"] == ["user-tag1", "user-tag2"] + assert result["custom_field"] == "from_metadata" + + # litellm_metadata values should only be included if not in metadata + assert result["model_group"] == "gpt-4-group" + + +def test_merge_litellm_metadata_skip_non_serializable(): + """ + Test that non-serializable objects like UserAPIKeyAuth are skipped. + """ + from litellm.proxy._types import UserAPIKeyAuth + + user_api_key_auth = UserAPIKeyAuth( + api_key="test-key", + user_id="test-user", + team_id="test-team", + ) + + litellm_params = { + "metadata": { + "user_api_key": "test-key-123", + "user_api_key_auth": user_api_key_auth, # This should be skipped + "safe_field": "safe_value", + }, + "litellm_metadata": { + "model_group": "gpt-4-group", + }, + } + + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + + # user_api_key_auth should be skipped + assert "user_api_key_auth" not in result + + # Other fields should be present + assert result["user_api_key"] == "test-key-123" + assert result["safe_field"] == "safe_value" + assert result["model_group"] == "gpt-4-group" + + +def test_merge_litellm_metadata_empty_params(): + """ + Test that merge_litellm_metadata handles empty or missing metadata gracefully. + """ + # Test with empty litellm_params + result = StandardLoggingPayloadSetup.merge_litellm_metadata({}) + assert result == {} + + # Test with only metadata + litellm_params = { + "metadata": { + "user_api_key": "test-key", + } + } + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + assert result == {"user_api_key": "test-key"} + + # Test with only litellm_metadata + litellm_params = { + "litellm_metadata": { + "model_group": "gpt-4-group", + } + } + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + assert result == {"model_group": "gpt-4-group"} + + # Test with None values + litellm_params = { + "metadata": None, + "litellm_metadata": None, + } + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + assert result == {} + + +def test_merge_litellm_metadata_bedrock_passthrough_scenario(): + """ + Test merge_litellm_metadata in a Bedrock passthrough scenario where both + user API key metadata and model metadata need to be merged. + + This is the specific scenario that was fixed - bedrock passthrough requests + should include complete user authentication metadata in logging. + """ + litellm_params = { + "metadata": { + # User API key fields from authentication + "user_api_key": "sk-bedrock-test-key-123", + "user_api_key_hash": "hashed-key-123", + "user_api_key_user_id": "bedrock-user-456", + "user_api_key_team_id": "bedrock-team-789", + "user_api_key_org_id": "bedrock-org-101", + "user_api_key_alias": "bedrock-key-alias", + "user_api_key_team_alias": "bedrock-team-alias", + "user_api_key_end_user_id": "end-user-123", + "user_api_key_request_route": "/bedrock/model/invoke", + }, + "litellm_metadata": { + # Model-related fields from Bedrock configuration + "model_group": "bedrock-claude-group", + "model_info": { + "id": "anthropic.claude-3-sonnet", + "mode": "chat", + }, + "aws_region_name": "us-east-1", + "tags": ["production", "bedrock"], + }, + } + + result = StandardLoggingPayloadSetup.merge_litellm_metadata(litellm_params) + + # Verify all user API key fields are present + assert result["user_api_key"] == "sk-bedrock-test-key-123" + assert result["user_api_key_hash"] == "hashed-key-123" + assert result["user_api_key_user_id"] == "bedrock-user-456" + assert result["user_api_key_team_id"] == "bedrock-team-789" + assert result["user_api_key_org_id"] == "bedrock-org-101" + assert result["user_api_key_alias"] == "bedrock-key-alias" + assert result["user_api_key_team_alias"] == "bedrock-team-alias" + assert result["user_api_key_end_user_id"] == "end-user-123" + assert result["user_api_key_request_route"] == "/bedrock/model/invoke" + + # Verify all model-related fields are present + assert result["model_group"] == "bedrock-claude-group" + assert result["model_info"] == { + "id": "anthropic.claude-3-sonnet", + "mode": "chat", + } + assert result["aws_region_name"] == "us-east-1" + assert result["tags"] == ["production", "bedrock"] + + # Verify total number of fields (9 user fields + 4 model fields = 13) + assert len(result) == 13 From 5676c6c1356eafd2c64634f090c00408f0eb777b Mon Sep 17 00:00:00 2001 From: burnerburnerburnerman Date: Thu, 15 Jan 2026 21:09:30 +0000 Subject: [PATCH 07/73] Chore: bump boto3 version (#19090) --- .circleci/config.yml | 46 ++++++++++---------- poetry.lock | 44 +++++++++---------- pyproject.toml | 2 +- requirements.txt | 4 +- tests/code_coverage_tests/license_cache.json | 4 +- 5 files changed, 50 insertions(+), 50 deletions(-) diff --git a/.circleci/config.yml b/.circleci/config.yml index 133a7184f9b..dc3e6d64e98 100644 --- a/.circleci/config.yml +++ b/.circleci/config.yml @@ -144,8 +144,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 pip install "azure-identity==1.16.1" @@ -260,8 +260,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 pip install "azure-identity==1.16.1" @@ -367,8 +367,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 pip install "azure-identity==1.16.1" @@ -637,8 +637,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install "langfuse>=2.0.0" pip install "logfire==0.29.0" @@ -759,8 +759,8 @@ jobs: pip install "google-cloud-aiplatform==1.43.0" pip install "google-genai==1.22.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 pip install "azure-identity==1.16.1" @@ -865,8 +865,8 @@ jobs: pip install "google-cloud-aiplatform==1.43.0" pip install "google-genai==1.22.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 pip install "azure-identity==1.16.1" @@ -972,8 +972,8 @@ jobs: pip install "google-cloud-aiplatform==1.43.0" pip install "google-genai==1.22.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 pip install "azure-identity==1.16.1" @@ -1198,7 +1198,7 @@ jobs: pip install "pytest-asyncio==0.21.1" pip install "respx==0.22.0" pip install "pydantic==2.10.2" - pip install "boto3==1.36.0" + pip install "boto3==1.40.15" # Run pytest and generate JUnit XML report - run: name: Run tests @@ -1879,7 +1879,7 @@ jobs: pip install aiohttp pip install openai pip install click - pip install "boto3==1.36.0" + pip install "boto3==1.40.15" pip install jinja2 pip install "tokenizers==0.20.0" pip install "uvloop==0.21.0" @@ -2176,8 +2176,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install "langfuse>=2.0.0" pip install "logfire==0.29.0" @@ -2316,8 +2316,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install "langchain_mcp_adapters==0.0.5" pip install "langfuse>=2.0.0" @@ -2462,8 +2462,8 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.36.0" - pip install "aioboto3==13.4.0" + pip install "boto3==1.40.15" + pip install "aioboto3==15.5.0" pip install langchain pip install "langfuse>=2.0.0" pip install "logfire==0.29.0" @@ -3118,7 +3118,7 @@ jobs: pip install "pytest==7.3.1" pip install "pytest-mock==3.12.0" pip install "pytest-asyncio==0.21.1" - pip install "boto3==1.36.0" + pip install "boto3==1.40.15" pip install "mypy==1.18.2" pip install pyarrow pip install numpydoc diff --git a/poetry.lock b/poetry.lock index 3bafdb157ca..249933b2ae1 100644 --- a/poetry.lock +++ b/poetry.lock @@ -1,4 +1,4 @@ -# This file is automatically @generated by Poetry 2.2.0 and should not be changed by hand. +# This file is automatically @generated by Poetry 2.2.1 and should not be changed by hand. [[package]] name = "aiofiles" @@ -525,36 +525,36 @@ files = [ [[package]] name = "boto3" -version = "1.36.0" +version = "1.40.15" description = "The AWS SDK for Python" optional = true -python-versions = ">=3.8" +python-versions = ">=3.9" groups = ["main"] markers = "extra == \"proxy\"" files = [ - {file = "boto3-1.36.0-py3-none-any.whl", hash = "sha256:d0ca7a58ce25701a52232cc8df9d87854824f1f2964b929305722ebc7959d5a9"}, - {file = "boto3-1.36.0.tar.gz", hash = "sha256:159898f51c2997a12541c0e02d6e5a8fe2993ddb307b9478fd9a339f98b57e00"}, + {file = "boto3-1.40.15-py3-none-any.whl", hash = "sha256:52b8aa78c9906c4e49dcec6817c041df33c9825073bf66e7df8fc00afbe47b4b"}, + {file = "boto3-1.40.15.tar.gz", hash = "sha256:271b379ce5ad35ca82f1009e917528a182eed0e2de197ccffb0c51acadec5c79"}, ] [package.dependencies] -botocore = ">=1.36.0,<1.37.0" +botocore = ">=1.40.15,<1.41.0" jmespath = ">=0.7.1,<2.0.0" -s3transfer = ">=0.11.0,<0.12.0" +s3transfer = ">=0.13.0,<0.14.0" [package.extras] crt = ["botocore[crt] (>=1.21.0,<2.0a0)"] [[package]] name = "botocore" -version = "1.36.26" +version = "1.40.76" description = "Low-level, data-driven core of boto 3." optional = true -python-versions = ">=3.8" +python-versions = ">=3.9" groups = ["main"] markers = "extra == \"proxy\"" files = [ - {file = "botocore-1.36.26-py3-none-any.whl", hash = "sha256:4e3f19913887a58502e71ef8d696fe7eaa54de7813ff73390cd5883f837dfa6e"}, - {file = "botocore-1.36.26.tar.gz", hash = "sha256:4a63bcef7ecf6146fd3a61dc4f9b33b7473b49bdaf1770e9aaca6eee0c9eab62"}, + {file = "botocore-1.40.76-py3-none-any.whl", hash = "sha256:fe425d386e48ac64c81cbb4a7181688d813df2e2b4c78b95ebe833c9e868c6f4"}, + {file = "botocore-1.40.76.tar.gz", hash = "sha256:2b16024d68b29b973005adfb5039adfe9099ebe772d40a90ca89f2e165c495dc"}, ] [package.dependencies] @@ -566,7 +566,7 @@ urllib3 = [ ] [package.extras] -crt = ["awscrt (==0.23.8)"] +crt = ["awscrt (==0.28.4)"] [[package]] name = "cachetools" @@ -2375,7 +2375,7 @@ description = "WSGI HTTP Server for UNIX" optional = true python-versions = ">=3.7" groups = ["main"] -markers = "extra == \"proxy\" or (extra == \"mlflow\" or extra == \"proxy\") and platform_system != \"Windows\" and python_version >= \"3.10\"" +markers = "python_version >= \"3.10\" and platform_system != \"Windows\" and (extra == \"proxy\" or extra == \"mlflow\") or extra == \"proxy\"" files = [ {file = "gunicorn-23.0.0-py3-none-any.whl", hash = "sha256:ec400d38950de4dfd418cff8328b2c8faed0edb0d517d3394e457c317908ca4d"}, {file = "gunicorn-23.0.0.tar.gz", hash = "sha256:f014447a0101dc57e294f6c18ca6b40227a4c90e9bdb586042628030cba004ec"}, @@ -3433,8 +3433,8 @@ files = [ [package.dependencies] numpy = [ {version = ">=1.23.3", markers = "python_version >= \"3.11\""}, - {version = ">1.20"}, {version = ">=1.21.2", markers = "python_version >= \"3.10\""}, + {version = ">1.20", markers = "python_version < \"3.10\""}, {version = ">=1.26.0", markers = "python_version >= \"3.12\""}, ] @@ -6255,22 +6255,22 @@ files = [ [[package]] name = "s3transfer" -version = "0.11.3" +version = "0.13.1" description = "An Amazon S3 Transfer Manager" optional = true -python-versions = ">=3.8" +python-versions = ">=3.9" groups = ["main"] markers = "extra == \"proxy\"" files = [ - {file = "s3transfer-0.11.3-py3-none-any.whl", hash = "sha256:ca855bdeb885174b5ffa95b9913622459d4ad8e331fc98eb01e6d5eb6a30655d"}, - {file = "s3transfer-0.11.3.tar.gz", hash = "sha256:edae4977e3a122445660c7c114bba949f9d191bae3b34a096f18a1c8c354527a"}, + {file = "s3transfer-0.13.1-py3-none-any.whl", hash = "sha256:a981aa7429be23fe6dfc13e80e4020057cbab622b08c0315288758d67cabc724"}, + {file = "s3transfer-0.13.1.tar.gz", hash = "sha256:c3fdba22ba1bd367922f27ec8032d6a1cf5f10c934fb5d68cf60fd5a23d936cf"}, ] [package.dependencies] -botocore = ">=1.36.0,<2.0a.0" +botocore = ">=1.37.4,<2.0a.0" [package.extras] -crt = ["botocore[crt] (>=1.36.0,<2.0a.0)"] +crt = ["botocore[crt] (>=1.37.4,<2.0a.0)"] [[package]] name = "scikit-learn" @@ -7201,7 +7201,7 @@ files = [ {file = "tomli-2.3.0-py3-none-any.whl", hash = "sha256:e95b1af3c5b07d9e643909b5abbec77cd9f1217e6d0bca72b0234736b9fb1f1b"}, {file = "tomli-2.3.0.tar.gz", hash = "sha256:64be704a875d2a59753d80ee8a533c3fe183e3f06807ff7dc2232938ccb01549"}, ] -markers = {main = "extra == \"utils\" and python_version == \"3.9\" or python_version == \"3.10\" and (extra == \"utils\" or extra == \"mlflow\")", dev = "python_version < \"3.11\"", proxy-dev = "python_version < \"3.11\""} +markers = {main = "python_version == \"3.10\" and (extra == \"utils\" or extra == \"mlflow\") or extra == \"utils\" and python_version == \"3.9\"", dev = "python_version < \"3.11\"", proxy-dev = "python_version < \"3.11\""} [[package]] name = "tomlkit" @@ -7981,4 +7981,4 @@ utils = ["numpydoc"] [metadata] lock-version = "2.1" python-versions = ">=3.9,<4.0" -content-hash = "ea62b77c662ab9fc486e421c576f0868bcde16d62a24703ee1f4916a0465ffb2" +content-hash = "a0d4bdda2742911291e79bab30faaaede14463f738c239425afcfe0f6b886d55" diff --git a/pyproject.toml b/pyproject.toml index aa8e6fd97be..f9d27f5317d 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -56,7 +56,7 @@ google-cloud-iam = {version = "^2.19.1", optional = true} resend = {version = ">=0.8.0", optional = true} pynacl = {version = "^1.5.0", optional = true} websockets = {version = "^15.0.1", optional = true} -boto3 = {version = "1.36.0", optional = true} +boto3 = {version = "1.40.15", optional = true} redisvl = {version = "^0.4.1", optional = true, markers = "python_version >= '3.9' and python_version < '3.14'"} mcp = {version = "^1.21.2", optional = true, python = ">=3.10"} litellm-proxy-extras = {version = "0.4.21", optional = true} diff --git a/requirements.txt b/requirements.txt index 5f00a269a7c..c95c78dfa8e 100644 --- a/requirements.txt +++ b/requirements.txt @@ -10,7 +10,7 @@ uvicorn==0.31.1 # server dep gunicorn==23.0.0 # server dep fastuuid==0.13.5 # for uuid4 uvloop==0.21.0 # uvicorn dep, gives us much better performance under load -boto3==1.36.0 # aws bedrock/sagemaker calls +boto3==1.40.15 # aws bedrock/sagemaker calls redis==5.2.1 # redis caching prisma==0.11.0 # for db nodejs-wheel-binaries==24.12.0 ## required by prisma for migrations, prevents runtime download (updated from nodejs-bin for security fixes) @@ -58,7 +58,7 @@ click==8.1.7 # for proxy cli rich==13.7.1 # for litellm proxy cli jinja2==3.1.6 # for prompt templates aiohttp==3.13.3 # for network calls -aioboto3==13.4.0 # for async sagemaker calls +aioboto3==15.5.0 # for async sagemaker calls tenacity==8.5.0 # for retrying requests, when litellm.num_retries set pydantic>=2.11,<3 # proxy + openai req. + mcp jsonschema>=4.23.0,<5.0.0 # validating json schema - aligned with openapi-core + mcp diff --git a/tests/code_coverage_tests/license_cache.json b/tests/code_coverage_tests/license_cache.json index 910ec931c86..21f74e26520 100644 --- a/tests/code_coverage_tests/license_cache.json +++ b/tests/code_coverage_tests/license_cache.json @@ -4,7 +4,7 @@ "pyyaml:6.0.2": "MIT", "gunicorn:22.0.0": "MIT", "uvloop:0.21.0": "MIT License", - "boto3:1.36.0": "Apache License 2.0", + "boto3:1.40.15": "Apache License 2.0", "redis:5.0.0": "MIT", "numpy:2.1.1": "Copyright (c) 2005-2024, NumPy Developers. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of the NumPy Developers nor the names of any contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. ---- The NumPy repository and source distributions bundle several libraries that are compatibly licensed. We list these here. Name: lapack-lite Files: numpy/linalg/lapack_lite/* License: BSD-3-Clause For details, see numpy/linalg/lapack_lite/LICENSE.txt Name: dragon4 Files: numpy/_core/src/multiarray/dragon4.c License: MIT For license text, see numpy/_core/src/multiarray/dragon4.c Name: libdivide Files: numpy/_core/include/numpy/libdivide/* License: Zlib For license text, see numpy/_core/include/numpy/libdivide/LICENSE.txt Note that the following files are vendored in the repository and sdist but not installed in built numpy packages: Name: Meson Files: vendored-meson/meson/* License: Apache 2.0 For license text, see vendored-meson/meson/COPYING Name: spin Files: .spin/cmds.py License: BSD-3 For license text, see .spin/LICENSE ---- This binary distribution of NumPy also bundles the following software: Name: OpenBLAS Files: numpy/.dylibs/libscipy_openblas*.so Description: bundled as a dynamically linked library Availability: https://github.com/OpenMathLib/OpenBLAS/ License: BSD-3-Clause Copyright (c) 2011-2014, The OpenBLAS Project All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the OpenBLAS project nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Name: LAPACK Files: numpy/.dylibs/libscipy_openblas*.so Description: bundled in OpenBLAS Availability: https://github.com/OpenMathLib/OpenBLAS/ License: BSD-3-Clause-Attribution Copyright (c) 1992-2013 The University of Tennessee and The University of Tennessee Research Foundation. All rights reserved. Copyright (c) 2000-2013 The University of California Berkeley. All rights reserved. Copyright (c) 2006-2013 The University of Colorado Denver. All rights reserved. $COPYRIGHT$ Additional copyrights may follow $HEADER$ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: - Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer listed in this license in the documentation and/or other materials provided with the distribution. - Neither the name of the copyright holders nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. The copyright holders provide no reassurances that the source code provided does not infringe any patent, copyright, or any other intellectual property rights of third parties. The copyright holders disclaim any liability to any recipient for claims brought against recipient by any third party for infringement of that parties intellectual property rights. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Name: GCC runtime library Files: numpy/.dylibs/libgfortran*, numpy/.dylibs/libgcc* Description: dynamically linked to files compiled with gcc Availability: https://gcc.gnu.org/git/?p=gcc.git;a=tree;f=libgfortran License: GPL-3.0-with-GCC-exception Copyright (C) 2002-2017 Free Software Foundation, Inc. Libgfortran is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version. Libgfortran is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. Under Section 7 of GPL version 3, you are granted additional permissions described in the GCC Runtime Library Exception, version 3.1, as published by the Free Software Foundation. You should have received a copy of the GNU General Public License and a copy of the GCC Runtime Library Exception along with this program; see the files COPYING3 and COPYING.RUNTIME respectively. If not, see . ---- Full text of license texts referred to above follows (that they are listed below does not necessarily imply the conditions apply to the present binary release): ---- GCC RUNTIME LIBRARY EXCEPTION Version 3.1, 31 March 2009 Copyright (C) 2009 Free Software Foundation, Inc. Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. This GCC Runtime Library Exception (\"Exception\") is an additional permission under section 7 of the GNU General Public License, version 3 (\"GPLv3\"). It applies to a given file (the \"Runtime Library\") that bears a notice placed by the copyright holder of the file stating that the file is governed by GPLv3 along with this Exception. When you use GCC to compile a program, GCC may combine portions of certain GCC header files and runtime libraries with the compiled program. The purpose of this Exception is to allow compilation of non-GPL (including proprietary) programs to use, in this way, the header files and runtime libraries covered by this Exception. 0. Definitions. A file is an \"Independent Module\" if it either requires the Runtime Library for execution after a Compilation Process, or makes use of an interface provided by the Runtime Library, but is not otherwise based on the Runtime Library. \"GCC\" means a version of the GNU Compiler Collection, with or without modifications, governed by version 3 (or a specified later version) of the GNU General Public License (GPL) with the option of using any subsequent versions published by the FSF. \"GPL-compatible Software\" is software whose conditions of propagation, modification and use would permit combination with GCC in accord with the license of GCC. \"Target Code\" refers to output from any compiler for a real or virtual target processor architecture, in executable form or suitable for input to an assembler, loader, linker and/or execution phase. Notwithstanding that, Target Code does not include data in any format that is used as a compiler intermediate representation, or used for producing a compiler intermediate representation. The \"Compilation Process\" transforms code entirely represented in non-intermediate languages designed for human-written code, and/or in Java Virtual Machine byte code, into Target Code. Thus, for example, use of source code generators and preprocessors need not be considered part of the Compilation Process, since the Compilation Process can be understood as starting with the output of the generators or preprocessors. A Compilation Process is \"Eligible\" if it is done using GCC, alone or with other GPL-compatible software, or if it is done without using any work based on GCC. For example, using non-GPL-compatible Software to optimize any GCC intermediate representations would not qualify as an Eligible Compilation Process. 1. Grant of Additional Permission. You have permission to propagate a work of Target Code formed by combining the Runtime Library with Independent Modules, even if such propagation would otherwise violate the terms of GPLv3, provided that all Target Code was generated by Eligible Compilation Processes. You may then convey such a combination under terms of your choice, consistent with the licensing of the Independent Modules. 2. No Weakening of GCC Copyleft. The availability of this Exception does not imply any general presumption that third-party software is unaffected by the copyleft requirements of the license of GCC. ---- GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007 Copyright (C) 2007 Free Software Foundation, Inc. Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. Preamble The GNU General Public License is a free, copyleft license for software and other kinds of works. The licenses for most software and other practical works are designed to take away your freedom to share and change the works. By contrast, the GNU General Public License is intended to guarantee your freedom to share and change all versions of a program--to make sure it remains free software for all its users. We, the Free Software Foundation, use the GNU General Public License for most of our software; it applies also to any other work released this way by its authors. You can apply it to your programs, too. When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for them if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs, and that you know you can do these things. To protect your rights, we need to prevent others from denying you these rights or asking you to surrender the rights. Therefore, you have certain responsibilities if you distribute copies of the software, or if you modify it: responsibilities to respect the freedom of others. For example, if you distribute copies of such a program, whether gratis or for a fee, you must pass on to the recipients the same freedoms that you received. You must make sure that they, too, receive or can get the source code. And you must show them these terms so they know their rights. Developers that use the GNU GPL protect your rights with two steps: (1) assert copyright on the software, and (2) offer you this License giving you legal permission to copy, distribute and/or modify it. For the developers' and authors' protection, the GPL clearly explains that there is no warranty for this free software. For both users' and authors' sake, the GPL requires that modified versions be marked as changed, so that their problems will not be attributed erroneously to authors of previous versions. Some devices are designed to deny users access to install or run modified versions of the software inside them, although the manufacturer can do so. This is fundamentally incompatible with the aim of protecting users' freedom to change the software. The systematic pattern of such abuse occurs in the area of products for individuals to use, which is precisely where it is most unacceptable. Therefore, we have designed this version of the GPL to prohibit the practice for those products. If such problems arise substantially in other domains, we stand ready to extend this provision to those domains in future versions of the GPL, as needed to protect the freedom of users. Finally, every program is threatened constantly by software patents. States should not allow patents to restrict development and use of software on general-purpose computers, but in those that do, we wish to avoid the special danger that patents applied to a free program could make it effectively proprietary. To prevent this, the GPL assures that patents cannot be used to render the program non-free. The precise terms and conditions for copying, distribution and modification follow. TERMS AND CONDITIONS 0. Definitions. \"This License\" refers to version 3 of the GNU General Public License. \"Copyright\" also means copyright-like laws that apply to other kinds of works, such as semiconductor masks. \"The Program\" refers to any copyrightable work licensed under this License. Each licensee is addressed as \"you\". \"Licensees\" and \"recipients\" may be individuals or organizations. To \"modify\" a work means to copy from or adapt all or part of the work in a fashion requiring copyright permission, other than the making of an exact copy. The resulting work is called a \"modified version\" of the earlier work or a work \"based on\" the earlier work. A \"covered work\" means either the unmodified Program or a work based on the Program. To \"propagate\" a work means to do anything with it that, without permission, would make you directly or secondarily liable for infringement under applicable copyright law, except executing it on a computer or modifying a private copy. Propagation includes copying, distribution (with or without modification), making available to the public, and in some countries other activities as well. To \"convey\" a work means any kind of propagation that enables other parties to make or receive copies. Mere interaction with a user through a computer network, with no transfer of a copy, is not conveying. An interactive user interface displays \"Appropriate Legal Notices\" to the extent that it includes a convenient and prominently visible feature that (1) displays an appropriate copyright notice, and (2) tells the user that there is no warranty for the work (except to the extent that warranties are provided), that licensees may convey the work under this License, and how to view a copy of this License. If the interface presents a list of user commands or options, such as a menu, a prominent item in the list meets this criterion. 1. Source Code. The \"source code\" for a work means the preferred form of the work for making modifications to it. \"Object code\" means any non-source form of a work. A \"Standard Interface\" means an interface that either is an official standard defined by a recognized standards body, or, in the case of interfaces specified for a particular programming language, one that is widely used among developers working in that language. The \"System Libraries\" of an executable work include anything, other than the work as a whole, that (a) is included in the normal form of packaging a Major Component, but which is not part of that Major Component, and (b) serves only to enable use of the work with that Major Component, or to implement a Standard Interface for which an implementation is available to the public in source code form. A \"Major Component\", in this context, means a major essential component (kernel, window system, and so on) of the specific operating system (if any) on which the executable work runs, or a compiler used to produce the work, or an object code interpreter used to run it. The \"Corresponding Source\" for a work in object code form means all the source code needed to generate, install, and (for an executable work) run the object code and to modify the work, including scripts to control those activities. However, it does not include the work's System Libraries, or general-purpose tools or generally available free programs which are used unmodified in performing those activities but which are not part of the work. For example, Corresponding Source includes interface definition files associated with source files for the work, and the source code for shared libraries and dynamically linked subprograms that the work is specifically designed to require, such as by intimate data communication or control flow between those subprograms and other parts of the work. The Corresponding Source need not include anything that users can regenerate automatically from other parts of the Corresponding Source. The Corresponding Source for a work in source code form is that same work. 2. Basic Permissions. All rights granted under this License are granted for the term of copyright on the Program, and are irrevocable provided the stated conditions are met. This License explicitly affirms your unlimited permission to run the unmodified Program. The output from running a covered work is covered by this License only if the output, given its content, constitutes a covered work. This License acknowledges your rights of fair use or other equivalent, as provided by copyright law. You may make, run and propagate covered works that you do not convey, without conditions so long as your license otherwise remains in force. You may convey covered works to others for the sole purpose of having them make modifications exclusively for you, or provide you with facilities for running those works, provided that you comply with the terms of this License in conveying all material for which you do not control copyright. Those thus making or running the covered works for you must do so exclusively on your behalf, under your direction and control, on terms that prohibit them from making any copies of your copyrighted material outside their relationship with you. Conveying under any other circumstances is permitted solely under the conditions stated below. Sublicensing is not allowed; section 10 makes it unnecessary. 3. Protecting Users' Legal Rights From Anti-Circumvention Law. No covered work shall be deemed part of an effective technological measure under any applicable law fulfilling obligations under article 11 of the WIPO copyright treaty adopted on 20 December 1996, or similar laws prohibiting or restricting circumvention of such measures. When you convey a covered work, you waive any legal power to forbid circumvention of technological measures to the extent such circumvention is effected by exercising rights under this License with respect to the covered work, and you disclaim any intention to limit operation or modification of the work as a means of enforcing, against the work's users, your or third parties' legal rights to forbid circumvention of technological measures. 4. Conveying Verbatim Copies. You may convey verbatim copies of the Program's source code as you receive it, in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate copyright notice; keep intact all notices stating that this License and any non-permissive terms added in accord with section 7 apply to the code; keep intact all notices of the absence of any warranty; and give all recipients a copy of this License along with the Program. You may charge any price or no price for each copy that you convey, and you may offer support or warranty protection for a fee. 5. Conveying Modified Source Versions. You may convey a work based on the Program, or the modifications to produce it from the Program, in the form of source code under the terms of section 4, provided that you also meet all of these conditions: a) The work must carry prominent notices stating that you modified it, and giving a relevant date. b) The work must carry prominent notices stating that it is released under this License and any conditions added under section 7. This requirement modifies the requirement in section 4 to \"keep intact all notices\". c) You must license the entire work, as a whole, under this License to anyone who comes into possession of a copy. This License will therefore apply, along with any applicable section 7 additional terms, to the whole of the work, and all its parts, regardless of how they are packaged. This License gives no permission to license the work in any other way, but it does not invalidate such permission if you have separately received it. d) If the work has interactive user interfaces, each must display Appropriate Legal Notices; however, if the Program has interactive interfaces that do not display Appropriate Legal Notices, your work need not make them do so. A compilation of a covered work with other separate and independent works, which are not by their nature extensions of the covered work, and which are not combined with it such as to form a larger program, in or on a volume of a storage or distribution medium, is called an \"aggregate\" if the compilation and its resulting copyright are not used to limit the access or legal rights of the compilation's users beyond what the individual works permit. Inclusion of a covered work in an aggregate does not cause this License to apply to the other parts of the aggregate. 6. Conveying Non-Source Forms. You may convey a covered work in object code form under the terms of sections 4 and 5, provided that you also convey the machine-readable Corresponding Source under the terms of this License, in one of these ways: a) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by the Corresponding Source fixed on a durable physical medium customarily used for software interchange. b) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by a written offer, valid for at least three years and valid for as long as you offer spare parts or customer support for that product model, to give anyone who possesses the object code either (1) a copy of the Corresponding Source for all the software in the product that is covered by this License, on a durable physical medium customarily used for software interchange, for a price no more than your reasonable cost of physically performing this conveying of source, or (2) access to copy the Corresponding Source from a network server at no charge. c) Convey individual copies of the object code with a copy of the written offer to provide the Corresponding Source. This alternative is allowed only occasionally and noncommercially, and only if you received the object code with such an offer, in accord with subsection 6b. d) Convey the object code by offering access from a designated place (gratis or for a charge), and offer equivalent access to the Corresponding Source in the same way through the same place at no further charge. You need not require recipients to copy the Corresponding Source along with the object code. If the place to copy the object code is a network server, the Corresponding Source may be on a different server (operated by you or a third party) that supports equivalent copying facilities, provided you maintain clear directions next to the object code saying where to find the Corresponding Source. Regardless of what server hosts the Corresponding Source, you remain obligated to ensure that it is available for as long as needed to satisfy these requirements. e) Convey the object code using peer-to-peer transmission, provided you inform other peers where the object code and Corresponding Source of the work are being offered to the general public at no charge under subsection 6d. A separable portion of the object code, whose source code is excluded from the Corresponding Source as a System Library, need not be included in conveying the object code work. A \"User Product\" is either (1) a \"consumer product\", which means any tangible personal property which is normally used for personal, family, or household purposes, or (2) anything designed or sold for incorporation into a dwelling. In determining whether a product is a consumer product, doubtful cases shall be resolved in favor of coverage. For a particular product received by a particular user, \"normally used\" refers to a typical or common use of that class of product, regardless of the status of the particular user or of the way in which the particular user actually uses, or expects or is expected to use, the product. A product is a consumer product regardless of whether the product has substantial commercial, industrial or non-consumer uses, unless such uses represent the only significant mode of use of the product. \"Installation Information\" for a User Product means any methods, procedures, authorization keys, or other information required to install and execute modified versions of a covered work in that User Product from a modified version of its Corresponding Source. The information must suffice to ensure that the continued functioning of the modified object code is in no case prevented or interfered with solely because modification has been made. If you convey an object code work under this section in, or with, or specifically for use in, a User Product, and the conveying occurs as part of a transaction in which the right of possession and use of the User Product is transferred to the recipient in perpetuity or for a fixed term (regardless of how the transaction is characterized), the Corresponding Source conveyed under this section must be accompanied by the Installation Information. But this requirement does not apply if neither you nor any third party retains the ability to install modified object code on the User Product (for example, the work has been installed in ROM). The requirement to provide Installation Information does not include a requirement to continue to provide support service, warranty, or updates for a work that has been modified or installed by the recipient, or for the User Product in which it has been modified or installed. Access to a network may be denied when the modification itself materially and adversely affects the operation of the network or violates the rules and protocols for communication across the network. Corresponding Source conveyed, and Installation Information provided, in accord with this section must be in a format that is publicly documented (and with an implementation available to the public in source code form), and must require no special password or key for unpacking, reading or copying. 7. Additional Terms. \"Additional permissions\" are terms that supplement the terms of this License by making exceptions from one or more of its conditions. Additional permissions that are applicable to the entire Program shall be treated as though they were included in this License, to the extent that they are valid under applicable law. If additional permissions apply only to part of the Program, that part may be used separately under those permissions, but the entire Program remains governed by this License without regard to the additional permissions. When you convey a copy of a covered work, you may at your option remove any additional permissions from that copy, or from any part of it. (Additional permissions may be written to require their own removal in certain cases when you modify the work.) You may place additional permissions on material, added by you to a covered work, for which you have or can give appropriate copyright permission. Notwithstanding any other provision of this License, for material you add to a covered work, you may (if authorized by the copyright holders of that material) supplement the terms of this License with terms: a) Disclaiming warranty or limiting liability differently from the terms of sections 15 and 16 of this License; or b) Requiring preservation of specified reasonable legal notices or author attributions in that material or in the Appropriate Legal Notices displayed by works containing it; or c) Prohibiting misrepresentation of the origin of that material, or requiring that modified versions of such material be marked in reasonable ways as different from the original version; or d) Limiting the use for publicity purposes of names of licensors or authors of the material; or e) Declining to grant rights under trademark law for use of some trade names, trademarks, or service marks; or f) Requiring indemnification of licensors and authors of that material by anyone who conveys the material (or modified versions of it) with contractual assumptions of liability to the recipient, for any liability that these contractual assumptions directly impose on those licensors and authors. All other non-permissive additional terms are considered \"further restrictions\" within the meaning of section 10. If the Program as you received it, or any part of it, contains a notice stating that it is governed by this License along with a term that is a further restriction, you may remove that term. If a license document contains a further restriction but permits relicensing or conveying under this License, you may add to a covered work material governed by the terms of that license document, provided that the further restriction does not survive such relicensing or conveying. If you add terms to a covered work in accord with this section, you must place, in the relevant source files, a statement of the additional terms that apply to those files, or a notice indicating where to find the applicable terms. Additional terms, permissive or non-permissive, may be stated in the form of a separately written license, or stated as exceptions; the above requirements apply either way. 8. Termination. You may not propagate or modify a covered work except as expressly provided under this License. Any attempt otherwise to propagate or modify it is void, and will automatically terminate your rights under this License (including any patent licenses granted under the third paragraph of section 11). However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation. Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice. Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, you do not qualify to receive new licenses for the same material under section 10. 9. Acceptance Not Required for Having Copies. You are not required to accept this License in order to receive or run a copy of the Program. Ancillary propagation of a covered work occurring solely as a consequence of using peer-to-peer transmission to receive a copy likewise does not require acceptance. However, nothing other than this License grants you permission to propagate or modify any covered work. These actions infringe copyright if you do not accept this License. Therefore, by modifying or propagating a covered work, you indicate your acceptance of this License to do so. 10. Automatic Licensing of Downstream Recipients. Each time you convey a covered work, the recipient automatically receives a license from the original licensors, to run, modify and propagate that work, subject to this License. You are not responsible for enforcing compliance by third parties with this License. An \"entity transaction\" is a transaction transferring control of an organization, or substantially all assets of one, or subdividing an organization, or merging organizations. If propagation of a covered work results from an entity transaction, each party to that transaction who receives a copy of the work also receives whatever licenses to the work the party's predecessor in interest had or could give under the previous paragraph, plus a right to possession of the Corresponding Source of the work from the predecessor in interest, if the predecessor has it or can get it with reasonable efforts. You may not impose any further restrictions on the exercise of the rights granted or affirmed under this License. For example, you may not impose a license fee, royalty, or other charge for exercise of rights granted under this License, and you may not initiate litigation (including a cross-claim or counterclaim in a lawsuit) alleging that any patent claim is infringed by making, using, selling, offering for sale, or importing the Program or any portion of it. 11. Patents. A \"contributor\" is a copyright holder who authorizes use under this License of the Program or a work on which the Program is based. The work thus licensed is called the contributor's \"contributor version\". A contributor's \"essential patent claims\" are all patent claims owned or controlled by the contributor, whether already acquired or hereafter acquired, that would be infringed by some manner, permitted by this License, of making, using, or selling its contributor version, but do not include claims that would be infringed only as a consequence of further modification of the contributor version. For purposes of this definition, \"control\" includes the right to grant patent sublicenses in a manner consistent with the requirements of this License. Each contributor grants you a non-exclusive, worldwide, royalty-free patent license under the contributor's essential patent claims, to make, use, sell, offer for sale, import and otherwise run, modify and propagate the contents of its contributor version. In the following three paragraphs, a \"patent license\" is any express agreement or commitment, however denominated, not to enforce a patent (such as an express permission to practice a patent or covenant not to sue for patent infringement). To \"grant\" such a patent license to a party means to make such an agreement or commitment not to enforce a patent against the party. If you convey a covered work, knowingly relying on a patent license, and the Corresponding Source of the work is not available for anyone to copy, free of charge and under the terms of this License, through a publicly available network server or other readily accessible means, then you must either (1) cause the Corresponding Source to be so available, or (2) arrange to deprive yourself of the benefit of the patent license for this particular work, or (3) arrange, in a manner consistent with the requirements of this License, to extend the patent license to downstream recipients. \"Knowingly relying\" means you have actual knowledge that, but for the patent license, your conveying the covered work in a country, or your recipient's use of the covered work in a country, would infringe one or more identifiable patents in that country that you have reason to believe are valid. If, pursuant to or in connection with a single transaction or arrangement, you convey, or propagate by procuring conveyance of, a covered work, and grant a patent license to some of the parties receiving the covered work authorizing them to use, propagate, modify or convey a specific copy of the covered work, then the patent license you grant is automatically extended to all recipients of the covered work and works based on it. A patent license is \"discriminatory\" if it does not include within the scope of its coverage, prohibits the exercise of, or is conditioned on the non-exercise of one or more of the rights that are specifically granted under this License. You may not convey a covered work if you are a party to an arrangement with a third party that is in the business of distributing software, under which you make payment to the third party based on the extent of your activity of conveying the work, and under which the third party grants, to any of the parties who would receive the covered work from you, a discriminatory patent license (a) in connection with copies of the covered work conveyed by you (or copies made from those copies), or (b) primarily for and in connection with specific products or compilations that contain the covered work, unless you entered into that arrangement, or that patent license was granted, prior to 28 March 2007. Nothing in this License shall be construed as excluding or limiting any implied license or other defenses to infringement that may otherwise be available to you under applicable patent law. 12. No Surrender of Others' Freedom. If conditions are imposed on you (whether by court order, agreement or otherwise) that contradict the conditions of this License, they do not excuse you from the conditions of this License. If you cannot convey a covered work so as to satisfy simultaneously your obligations under this License and any other pertinent obligations, then as a consequence you may not convey it at all. For example, if you agree to terms that obligate you to collect a royalty for further conveying from those to whom you convey the Program, the only way you could satisfy both those terms and this License would be to refrain entirely from conveying the Program. 13. Use with the GNU Affero General Public License. Notwithstanding any other provision of this License, you have permission to link or combine any covered work with a work licensed under version 3 of the GNU Affero General Public License into a single combined work, and to convey the resulting work. The terms of this License will continue to apply to the part which is the covered work, but the special requirements of the GNU Affero General Public License, section 13, concerning interaction through a network will apply to the combination as such. 14. Revised Versions of this License. The Free Software Foundation may publish revised and/or new versions of the GNU General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. Each version is given a distinguishing version number. If the Program specifies that a certain numbered version of the GNU General Public License \"or any later version\" applies to it, you have the option of following the terms and conditions either of that numbered version or of any later version published by the Free Software Foundation. If the Program does not specify a version number of the GNU General Public License, you may choose any version ever published by the Free Software Foundation. If the Program specifies that a proxy can decide which future versions of the GNU General Public License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Program. Later license versions may give you additional or different permissions. However, no additional obligations are imposed on any author or copyright holder as a result of your choosing to follow a later version. 15. Disclaimer of Warranty. THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM \"AS IS\" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 16. Limitation of Liability. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. 17. Interpretation of Sections 15 and 16. If the disclaimer of warranty and limitation of liability provided above cannot be given local legal effect according to their terms, reviewing courts shall apply local law that most closely approximates an absolute waiver of all civil liability in connection with the Program, unless a warranty or assumption of liability accompanies a copy of the Program in return for a fee. END OF TERMS AND CONDITIONS How to Apply These Terms to Your New Programs If you develop a new program, and you want it to be of the greatest possible use to the public, the best way to achieve this is to make it free software which everyone can redistribute and change under these terms. To do so, attach the following notices to the program. It is safest to attach them to the start of each source file to most effectively state the exclusion of warranty; and each file should have at least the \"copyright\" line and a pointer to where the full notice is found. Copyright (C) This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . Also add information on how to contact you by electronic and paper mail. If the program does terminal interaction, make it output a short notice like this when it starts in an interactive mode: Copyright (C) This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. This is free software, and you are welcome to redistribute it under certain conditions; type `show c' for details. The hypothetical commands `show w' and `show c' should show the appropriate parts of the General Public License. Of course, your program's commands might be different; for a GUI interface, you would use an \"about box\". You should also get your employer (if you work as a programmer) or school, if any, to sign a \"copyright disclaimer\" for the program, if necessary. For more information on this, and how to apply and follow the GNU GPL, see . The GNU General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Lesser General Public License instead of this License. But first, please read . Name: libquadmath Files: numpy/.dylibs/libquadmath*.so Description: dynamically linked to files compiled with gcc Availability: https://gcc.gnu.org/git/?p=gcc.git;a=tree;f=libquadmath License: LGPL-2.1-or-later GCC Quad-Precision Math Library Copyright (C) 2010-2019 Free Software Foundation, Inc. Written by Francois-Xavier Coudert This file is part of the libquadmath library. Libquadmath is free software; you can redistribute it and/or modify it under the terms of the GNU Library General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. Libquadmath is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. https://www.gnu.org/licenses/old-licenses/lgpl-2.1.html", "prisma:0.11.0": "APACHE", @@ -35,7 +35,7 @@ "click:8.1.7": "BSD-3-Clause", "certifi:2024.12.14": "MPL-2.0", "aiohttp:3.10.2": "Apache 2", - "aioboto3:13.4.0": "Apache-2.0", + "aioboto3:15.5.0": "Apache-2.0", "tenacity:8.2.3": "Apache 2.0", "pydantic:2.10.0": "MIT", "jsonschema:4.22.0": "MIT", From 92827ead659e149e982490981d5674136a6ae328 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E6=8B=90=E7=88=B7=26=26=E8=80=81=E6=8B=90=E7=98=A6?= Date: Fri, 16 Jan 2026 06:00:34 +0800 Subject: [PATCH 08/73] Add pricing for volcengine models (deepseek-v3-2, glm-4-7, kimi-k2-thinking) (#19076) Co-authored-by: Claude Opus 4.5 --- model_prices_and_context_window.json | 42 ++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index a130aefa5de..91708fa13f3 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -10170,6 +10170,48 @@ "mode": "completion", "output_cost_per_token": 5e-07 }, + "deepseek-v3-2-251201": { + "input_cost_per_token": 0.0, + "litellm_provider": "volcengine", + "max_input_tokens": 98304, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 0.0, + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "glm-4-7-251222": { + "input_cost_per_token": 0.0, + "litellm_provider": "volcengine", + "max_input_tokens": 204800, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 0.0, + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "kimi-k2-thinking-251104": { + "input_cost_per_token": 0.0, + "litellm_provider": "volcengine", + "max_input_tokens": 229376, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 0.0, + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, "doubao-embedding": { "input_cost_per_token": 0.0, "litellm_provider": "volcengine", From ccc0e342f27f98ff4baa8193e876722161744027 Mon Sep 17 00:00:00 2001 From: Kris Xia Date: Fri, 16 Jan 2026 06:02:59 +0800 Subject: [PATCH 09/73] Make keepalive_timeout parameter work for Gunicorn (#19087) * [Fix] Containers API - Allow routing to regional endpoints (#19118) * fix get_complete_url * fix url resolution containers API * TestContainerRegionalApiBase * feat(proxy): add keepalive_timeout support for Gunicorn server Add configurable keepalive timeout parameter for Gunicorn workers to match existing Uvicorn functionality. This allows users to tune the keep-alive connection timeout based on their deployment requirements. Changes: - Add keepalive_timeout parameter to _run_gunicorn_server method - Configure Gunicorn's keepalive setting (defaults to 90s if not specified) - Update --keepalive_timeout CLI help text to document both Uvicorn and Gunicorn behavior - Pass keepalive_timeout from run_server to _run_gunicorn_server Tests: - Add test to verify keepalive_timeout flag is properly passed to Gunicorn - Add test to verify default 90s timeout when flag is not specified Co-Authored-By: lizhen921 <294474470@qq.com> Signed-off-by: Kris Xia --------- Signed-off-by: Kris Xia Co-authored-by: Ishaan Jaff Co-authored-by: lizhen921 <294474470@qq.com> --- litellm/containers/main.py | 48 +++++- .../llms/openai/containers/transformation.py | 9 +- litellm/proxy/proxy_cli.py | 8 +- .../test_container_regional_api_base.py | 163 ++++++++++++++++++ tests/test_litellm/proxy/test_proxy_cli.py | 69 ++++++++ 5 files changed, 288 insertions(+), 9 deletions(-) create mode 100644 tests/test_litellm/containers/test_container_regional_api_base.py diff --git a/litellm/containers/main.py b/litellm/containers/main.py index 625a291fb55..105e999ffe8 100644 --- a/litellm/containers/main.py +++ b/litellm/containers/main.py @@ -199,7 +199,13 @@ def create_container( return response # get llm provider logic - litellm_params = GenericLiteLLMParams(**kwargs) + # Pass credential params explicitly since they're named args, not in kwargs + litellm_params = GenericLiteLLMParams( + api_key=api_key, + api_base=api_base, + api_version=api_version, + **kwargs, + ) # get provider config container_provider_config: Optional[BaseContainerConfig] = ( ProviderConfigManager.get_provider_container_config( @@ -406,7 +412,13 @@ def list_containers( return response # get llm provider logic - litellm_params = GenericLiteLLMParams(**kwargs) + # Pass credential params explicitly since they're named args, not in kwargs + litellm_params = GenericLiteLLMParams( + api_key=api_key, + api_base=api_base, + api_version=api_version, + **kwargs, + ) # get provider config container_provider_config: Optional[BaseContainerConfig] = ( ProviderConfigManager.get_provider_container_config( @@ -594,7 +606,13 @@ def retrieve_container( return response # get llm provider logic - litellm_params = GenericLiteLLMParams(**kwargs) + # Pass credential params explicitly since they're named args, not in kwargs + litellm_params = GenericLiteLLMParams( + api_key=api_key, + api_base=api_base, + api_version=api_version, + **kwargs, + ) # get provider config container_provider_config: Optional[BaseContainerConfig] = ( ProviderConfigManager.get_provider_container_config( @@ -774,7 +792,13 @@ def delete_container( return response # get llm provider logic - litellm_params = GenericLiteLLMParams(**kwargs) + # Pass credential params explicitly since they're named args, not in kwargs + litellm_params = GenericLiteLLMParams( + api_key=api_key, + api_base=api_base, + api_version=api_version, + **kwargs, + ) # get provider config container_provider_config: Optional[BaseContainerConfig] = ( ProviderConfigManager.get_provider_container_config( @@ -968,7 +992,13 @@ def list_container_files( return response # get llm provider logic - litellm_params = GenericLiteLLMParams(**kwargs) + # Pass credential params explicitly since they're named args, not in kwargs + litellm_params = GenericLiteLLMParams( + api_key=api_key, + api_base=api_base, + api_version=api_version, + **kwargs, + ) # get provider config container_provider_config: Optional[BaseContainerConfig] = ( ProviderConfigManager.get_provider_container_config( @@ -1203,7 +1233,13 @@ def upload_container_file( return response # get llm provider logic - litellm_params = GenericLiteLLMParams(**kwargs) + # Pass credential params explicitly since they're named args, not in kwargs + litellm_params = GenericLiteLLMParams( + api_key=api_key, + api_base=api_base, + api_version=api_version, + **kwargs, + ) # get provider config container_provider_config: Optional[BaseContainerConfig] = ( ProviderConfigManager.get_provider_container_config( diff --git a/litellm/llms/openai/containers/transformation.py b/litellm/llms/openai/containers/transformation.py index 46718816f37..e67bfbe0c62 100644 --- a/litellm/llms/openai/containers/transformation.py +++ b/litellm/llms/openai/containers/transformation.py @@ -83,8 +83,13 @@ class OpenAIContainerConfig(BaseContainerConfig): ) -> str: """Get the complete URL for OpenAI container API. """ - if api_base is None: - api_base = "https://api.openai.com/v1" + api_base = ( + api_base + or litellm.api_base + or get_secret_str("OPENAI_BASE_URL") + or get_secret_str("OPENAI_API_BASE") + or "https://api.openai.com/v1" + ) return f"{api_base.rstrip('/')}/containers" diff --git a/litellm/proxy/proxy_cli.py b/litellm/proxy/proxy_cli.py index 2059246674b..ddc79a2865d 100644 --- a/litellm/proxy/proxy_cli.py +++ b/litellm/proxy/proxy_cli.py @@ -187,6 +187,7 @@ class ProxyInitializationHelpers: ssl_certfile_path: str, ssl_keyfile_path: str, max_requests_before_restart: Optional[int] = None, + keepalive_timeout: Optional[int] = None, ): """ Run litellm with `gunicorn` @@ -267,6 +268,10 @@ class ProxyInitializationHelpers: "access_log_format": '%(h)s %(l)s %(u)s %(t)s "%(r)s" %(s)s %(b)s', } + # Optional: set keepalive timeout if specified by user + if keepalive_timeout is not None: + gunicorn_options["keepalive"] = keepalive_timeout + # Optional: recycle workers after N requests to mitigate memory growth if max_requests_before_restart is not None: gunicorn_options["max_requests"] = max_requests_before_restart @@ -489,7 +494,7 @@ class ProxyInitializationHelpers: "--keepalive_timeout", default=None, type=int, - help="Set the uvicorn keepalive timeout in seconds (uvicorn timeout_keep_alive parameter)", + help="Set the keepalive timeout in seconds. For Uvicorn: timeout_keep_alive parameter. For Gunicorn: keepalive parameter. Default: Uvicorn uses ~75s, Gunicorn uses 90s", envvar="KEEPALIVE_TIMEOUT", ) @click.option( @@ -859,6 +864,7 @@ def run_server( # noqa: PLR0915 ssl_certfile_path=ssl_certfile_path, ssl_keyfile_path=ssl_keyfile_path, max_requests_before_restart=max_requests_before_restart, + keepalive_timeout=keepalive_timeout, ) elif run_hypercorn is True: ProxyInitializationHelpers._init_hypercorn_server( diff --git a/tests/test_litellm/containers/test_container_regional_api_base.py b/tests/test_litellm/containers/test_container_regional_api_base.py new file mode 100644 index 00000000000..7c6154867f0 --- /dev/null +++ b/tests/test_litellm/containers/test_container_regional_api_base.py @@ -0,0 +1,163 @@ +""" +Tests for OpenAI Containers API regional api_base support. + +Validates that litellm.create_container and litellm.upload_container_file +correctly use regional endpoints like https://us.api.openai.com/v1 for +US Data Residency instead of defaulting to https://api.openai.com/v1. +""" + +import os +import sys +from unittest.mock import MagicMock, patch + +import httpx +import pytest + +sys.path.insert(0, os.path.abspath("../../..")) + +import litellm + + +class TestContainerRegionalApiBase: + """Test suite for container API regional api_base support.""" + + def setup_method(self): + """Set up test fixtures.""" + os.environ["OPENAI_API_KEY"] = "sk-test123" + + def teardown_method(self): + """Clean up after tests.""" + if "OPENAI_API_KEY" in os.environ: + del os.environ["OPENAI_API_KEY"] + if "OPENAI_BASE_URL" in os.environ: + del os.environ["OPENAI_BASE_URL"] + if "OPENAI_API_BASE" in os.environ: + del os.environ["OPENAI_API_BASE"] + litellm.api_base = None + + @patch("litellm.llms.custom_httpx.http_handler.HTTPHandler.post") + def test_create_container_uses_regional_api_base(self, mock_post): + """ + Test that litellm.create_container uses the regional api_base when provided. + + This validates the fix for US Data Residency support where requests should + go to https://us.api.openai.com/v1 instead of https://api.openai.com/v1. + """ + mock_response = MagicMock(spec=httpx.Response) + mock_response.status_code = 200 + mock_response.json.return_value = { + "id": "cntr_123456", + "object": "container", + "created_at": 1747857508, + "status": "running", + "expires_after": {"anchor": "last_active_at", "minutes": 20}, + "last_active_at": 1747857508, + "name": "Test Container" + } + mock_post.return_value = mock_response + + litellm.create_container( + name="Test Container", + custom_llm_provider="openai", + api_base="https://us.api.openai.com/v1", + ) + + mock_post.assert_called_once() + call_args = mock_post.call_args + called_url = call_args[1]["url"] + + assert "us.api.openai.com" in called_url, f"Expected US regional URL, got: {called_url}" + assert called_url == "https://us.api.openai.com/v1/containers" + + @patch("litellm.llms.custom_httpx.http_handler.HTTPHandler.post") + def test_create_container_uses_env_var_openai_base_url(self, mock_post): + """ + Test that litellm.create_container uses OPENAI_BASE_URL env var. + """ + os.environ["OPENAI_BASE_URL"] = "https://us.api.openai.com/v1" + + mock_response = MagicMock(spec=httpx.Response) + mock_response.status_code = 200 + mock_response.json.return_value = { + "id": "cntr_123456", + "object": "container", + "created_at": 1747857508, + "status": "running", + "expires_after": {"anchor": "last_active_at", "minutes": 20}, + "last_active_at": 1747857508, + "name": "Test Container" + } + mock_post.return_value = mock_response + + litellm.create_container( + name="Test Container", + custom_llm_provider="openai", + ) + + mock_post.assert_called_once() + call_args = mock_post.call_args + called_url = call_args[1]["url"] + + assert "us.api.openai.com" in called_url, f"Expected US regional URL, got: {called_url}" + + @patch("litellm.llms.custom_httpx.http_handler.HTTPHandler.post") + def test_create_container_defaults_to_standard_openai(self, mock_post): + """ + Test that litellm.create_container defaults to standard OpenAI URL + when no regional api_base is configured. + """ + mock_response = MagicMock(spec=httpx.Response) + mock_response.status_code = 200 + mock_response.json.return_value = { + "id": "cntr_123456", + "object": "container", + "created_at": 1747857508, + "status": "running", + "expires_after": {"anchor": "last_active_at", "minutes": 20}, + "last_active_at": 1747857508, + "name": "Test Container" + } + mock_post.return_value = mock_response + + litellm.create_container( + name="Test Container", + custom_llm_provider="openai", + ) + + mock_post.assert_called_once() + call_args = mock_post.call_args + called_url = call_args[1]["url"] + + assert called_url == "https://api.openai.com/v1/containers" + + @patch("litellm.llms.custom_httpx.http_handler.HTTPHandler.post") + def test_upload_container_file_uses_regional_api_base(self, mock_post): + """ + Test that litellm.upload_container_file uses the regional api_base when provided. + """ + mock_response = MagicMock(spec=httpx.Response) + mock_response.status_code = 200 + mock_response.json.return_value = { + "id": "file_123456", + "object": "container.file", + "created_at": 1747857508, + "container_id": "cntr_123456", + "path": "/mnt/user/data.csv", + "source": "user", + } + mock_post.return_value = mock_response + + litellm.upload_container_file( + container_id="cntr_123456", + file=("data.csv", b"col1,col2\n1,2", "text/csv"), + custom_llm_provider="openai", + api_base="https://us.api.openai.com/v1", + ) + + mock_post.assert_called_once() + call_args = mock_post.call_args + called_url = call_args[1]["url"] + + assert "us.api.openai.com" in called_url, f"Expected US regional URL, got: {called_url}" + assert "cntr_123456/files" in called_url + diff --git a/tests/test_litellm/proxy/test_proxy_cli.py b/tests/test_litellm/proxy/test_proxy_cli.py index 5f03ef18171..99b4ebba064 100644 --- a/tests/test_litellm/proxy/test_proxy_cli.py +++ b/tests/test_litellm/proxy/test_proxy_cli.py @@ -483,6 +483,75 @@ class TestProxyInitializationHelpers: # Verify that uvicorn.run was called again mock_uvicorn_run.assert_called_once() + @patch("litellm.proxy.proxy_cli.ProxyInitializationHelpers._run_gunicorn_server") + @patch("builtins.print") + def test_gunicorn_keepalive_timeout_flag(self, mock_print, mock_gunicorn): + """Test that the keepalive_timeout flag is properly passed to Gunicorn""" + from click.testing import CliRunner + + from litellm.proxy.proxy_cli import run_server + + runner = CliRunner() + + mock_app = MagicMock() + mock_proxy_config = MagicMock() + mock_key_mgmt = MagicMock() + mock_save_worker_config = MagicMock() + + with patch.dict( + "sys.modules", + { + "proxy_server": MagicMock( + app=mock_app, + ProxyConfig=mock_proxy_config, + KeyManagementSettings=mock_key_mgmt, + save_worker_config=mock_save_worker_config, + ) + }, + ): + result = runner.invoke( + run_server, ["--local", "--run_gunicorn", "--keepalive_timeout", "120"] + ) + assert result.exit_code == 0 + + # Verify _run_gunicorn_server was called with keepalive_timeout + mock_gunicorn.assert_called_once() + call_kwargs = mock_gunicorn.call_args.kwargs + assert call_kwargs["keepalive_timeout"] == 120 + + @patch("litellm.proxy.proxy_cli.ProxyInitializationHelpers._run_gunicorn_server") + @patch("builtins.print") + def test_gunicorn_keepalive_default(self, mock_print, mock_gunicorn): + """Test that Gunicorn uses default 90s when keepalive_timeout not specified""" + from click.testing import CliRunner + + from litellm.proxy.proxy_cli import run_server + + runner = CliRunner() + + mock_app = MagicMock() + mock_proxy_config = MagicMock() + mock_key_mgmt = MagicMock() + mock_save_worker_config = MagicMock() + + with patch.dict( + "sys.modules", + { + "proxy_server": MagicMock( + app=mock_app, + ProxyConfig=mock_proxy_config, + KeyManagementSettings=mock_key_mgmt, + save_worker_config=mock_save_worker_config, + ) + }, + ): + result = runner.invoke(run_server, ["--local", "--run_gunicorn"]) + assert result.exit_code == 0 + + # Verify default behavior (keepalive_timeout is None, Gunicorn will use 90) + call_kwargs = mock_gunicorn.call_args.kwargs + assert call_kwargs.get("keepalive_timeout") is None + class TestHealthAppFactory: """Test cases for the health app factory module""" From ae7b70b9178b8972643e5c011679702d82808ab2 Mon Sep 17 00:00:00 2001 From: danielnyari-seon Date: Thu, 15 Jan 2026 23:04:40 +0100 Subject: [PATCH 10/73] Update prisma_migration.py (#19083) --- litellm/proxy/prisma_migration.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/litellm/proxy/prisma_migration.py b/litellm/proxy/prisma_migration.py index 251d1e56287..62909b8b2c7 100644 --- a/litellm/proxy/prisma_migration.py +++ b/litellm/proxy/prisma_migration.py @@ -26,3 +26,5 @@ if exit_code != 0: verbose_proxy_logger.error( f"'prisma generate' stderr: {result.stderr}" ) # Log stderr + +sys.exit(exit_code) \ No newline at end of file From 1c1b6faa8234c8f52bd1be6505b7524bd944ecd7 Mon Sep 17 00:00:00 2001 From: Harshit Jain <48647625+Harshit28j@users.noreply.github.com> Date: Fri, 16 Jan 2026 03:37:20 +0530 Subject: [PATCH 11/73] fix: model-level guardrails not taking effect (#18363) (#18895) * fix: model-level guardrails not taking effect (#18363) * fix(proxy): add support event-based deployment hooks * fix(proxy): add type safety check for guardrails --- litellm/proxy/common_request_processing.py | 4 +++- litellm/proxy/litellm_pre_call_utils.py | 23 ++++++++++++++++++---- 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/litellm/proxy/common_request_processing.py b/litellm/proxy/common_request_processing.py index 52f7f227b52..5b669bd048f 100644 --- a/litellm/proxy/common_request_processing.py +++ b/litellm/proxy/common_request_processing.py @@ -49,7 +49,9 @@ if TYPE_CHECKING: ProxyConfig = _ProxyConfig else: ProxyConfig = Any -from litellm.proxy.litellm_pre_call_utils import add_litellm_data_to_request +from litellm.proxy.litellm_pre_call_utils import ( + add_litellm_data_to_request, +) from litellm.types.utils import ModelResponse, ModelResponseStream, Usage diff --git a/litellm/proxy/litellm_pre_call_utils.py b/litellm/proxy/litellm_pre_call_utils.py index 3f844f21eb0..ad0ab6b7a38 100644 --- a/litellm/proxy/litellm_pre_call_utils.py +++ b/litellm/proxy/litellm_pre_call_utils.py @@ -846,7 +846,9 @@ async def add_litellm_data_to_request( # noqa: PLR0915 # Add headers to metadata for guardrails to access (fixes #17477) # Guardrails use metadata["headers"] to access request headers (e.g., User-Agent) - if _metadata_variable_name in data and isinstance(data[_metadata_variable_name], dict): + if _metadata_variable_name in data and isinstance( + data[_metadata_variable_name], dict + ): data[_metadata_variable_name]["headers"] = _headers # check for forwardable headers @@ -1307,6 +1309,9 @@ def move_guardrails_to_metadata( - If guardrails set on API Key metadata then sets guardrails on request metadata - If guardrails not set on API key, then checks request metadata + + Note: We copy (not pop) guardrails from data to metadata to ensure deployment-level + guardrails merged by the router remain in kwargs for async_pre_call_deployment_hook. """ # Check key-level guardrails _add_guardrails_from_key_or_team_metadata( @@ -1319,15 +1324,25 @@ def move_guardrails_to_metadata( ######################################################################################### # User's might send "guardrails" in the request body, we need to add them to the request metadata. # Since downstream logic requires "guardrails" to be in the request metadata + # + # IMPORTANT: We copy instead of pop to preserve guardrails in kwargs for + # async_pre_call_deployment_hook (custom_guardrail.py:290) which checks kwargs.get("guardrails"). + # This is the event-based approach for deployment-level guardrails. ######################################################################################### if "guardrails" in data: - request_body_guardrails = data.pop("guardrails") + request_body_guardrails = data.get("guardrails") + if request_body_guardrails is None: + return if "guardrails" in data[_metadata_variable_name] and isinstance( data[_metadata_variable_name]["guardrails"], list ): - data[_metadata_variable_name]["guardrails"].extend(request_body_guardrails) + # Merge unique guardrails + existing = data[_metadata_variable_name]["guardrails"] + for g in request_body_guardrails: + if g not in existing: + existing.append(g) else: - data[_metadata_variable_name]["guardrails"] = request_body_guardrails + data[_metadata_variable_name]["guardrails"] = list(request_body_guardrails) ######################################################################################### if "guardrail_config" in data: From 41d8f799294bf2d5fe9122710c0091bb1cab7561 Mon Sep 17 00:00:00 2001 From: Harshit Jain <48647625+Harshit28j@users.noreply.github.com> Date: Fri, 16 Jan 2026 03:41:21 +0530 Subject: [PATCH 12/73] fix: models loadbalancing billing issue by filter (#18891) * fix: models loadbalancing billing issue by filter * fix: separate key and team access groups in metadata --- litellm/proxy/auth/model_checks.py | 25 +- litellm/proxy/litellm_pre_call_utils.py | 59 +++-- litellm/router.py | 12 +- litellm/router_utils/common_utils.py | 79 +++++- ...est_filter_deployments_by_access_groups.py | 227 ++++++++++++++++++ 5 files changed, 382 insertions(+), 20 deletions(-) create mode 100644 tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py diff --git a/litellm/proxy/auth/model_checks.py b/litellm/proxy/auth/model_checks.py index 71ae1348f39..af2574d88ee 100644 --- a/litellm/proxy/auth/model_checks.py +++ b/litellm/proxy/auth/model_checks.py @@ -64,6 +64,27 @@ def _get_models_from_access_groups( return all_models +def get_access_groups_from_models( + model_access_groups: Dict[str, List[str]], + models: List[str], +) -> List[str]: + """ + Extract access group names from a models list. + + Given a models list like ["gpt-4", "beta-models", "claude-v1"] + and access groups like {"beta-models": ["gpt-5", "gpt-6"]}, + returns ["beta-models"]. + + This is used to pass allowed access groups to the router for filtering + deployments during load balancing (GitHub issue #18333). + """ + access_groups = [] + for model in models: + if model in model_access_groups: + access_groups.append(model) + return access_groups + + async def get_mcp_server_ids( user_api_key_dict: UserAPIKeyAuth, ) -> List[str]: @@ -80,7 +101,6 @@ async def get_mcp_server_ids( # Make a direct SQL query to get just the mcp_servers try: - result = await prisma_client.db.litellm_objectpermissiontable.find_unique( where={"object_permission_id": user_api_key_dict.object_permission_id}, ) @@ -176,6 +196,7 @@ def get_complete_model_list( """ unique_models = [] + def append_unique(models): for model in models: if model not in unique_models: @@ -188,7 +209,7 @@ def get_complete_model_list( else: append_unique(proxy_model_list) if include_model_access_groups: - append_unique(list(model_access_groups.keys())) # TODO: keys order + append_unique(list(model_access_groups.keys())) # TODO: keys order if user_model: append_unique([user_model]) diff --git a/litellm/proxy/litellm_pre_call_utils.py b/litellm/proxy/litellm_pre_call_utils.py index ad0ab6b7a38..7a49c1f6520 100644 --- a/litellm/proxy/litellm_pre_call_utils.py +++ b/litellm/proxy/litellm_pre_call_utils.py @@ -173,12 +173,12 @@ def _get_dynamic_logging_metadata( user_api_key_dict: UserAPIKeyAuth, proxy_config: ProxyConfig ) -> Optional[TeamCallbackMetadata]: callback_settings_obj: Optional[TeamCallbackMetadata] = None - key_dynamic_logging_settings: Optional[ - dict - ] = KeyAndTeamLoggingSettings.get_key_dynamic_logging_settings(user_api_key_dict) - team_dynamic_logging_settings: Optional[ - dict - ] = KeyAndTeamLoggingSettings.get_team_dynamic_logging_settings(user_api_key_dict) + key_dynamic_logging_settings: Optional[dict] = ( + KeyAndTeamLoggingSettings.get_key_dynamic_logging_settings(user_api_key_dict) + ) + team_dynamic_logging_settings: Optional[dict] = ( + KeyAndTeamLoggingSettings.get_team_dynamic_logging_settings(user_api_key_dict) + ) ######################################################################################### # Key-based callbacks ######################################################################################### @@ -661,11 +661,11 @@ class LiteLLMProxyRequestSetup: ## KEY-LEVEL SPEND LOGS / TAGS if "tags" in key_metadata and key_metadata["tags"] is not None: - data[_metadata_variable_name][ - "tags" - ] = LiteLLMProxyRequestSetup._merge_tags( - request_tags=data[_metadata_variable_name].get("tags"), - tags_to_add=key_metadata["tags"], + data[_metadata_variable_name]["tags"] = ( + LiteLLMProxyRequestSetup._merge_tags( + request_tags=data[_metadata_variable_name].get("tags"), + tags_to_add=key_metadata["tags"], + ) ) if "disable_global_guardrails" in key_metadata and isinstance( key_metadata["disable_global_guardrails"], bool @@ -933,9 +933,9 @@ async def add_litellm_data_to_request( # noqa: PLR0915 data[_metadata_variable_name]["litellm_api_version"] = version if general_settings is not None: - data[_metadata_variable_name][ - "global_max_parallel_requests" - ] = general_settings.get("global_max_parallel_requests", None) + data[_metadata_variable_name]["global_max_parallel_requests"] = ( + general_settings.get("global_max_parallel_requests", None) + ) ### KEY-LEVEL Controls key_metadata = user_api_key_dict.metadata @@ -1002,6 +1002,37 @@ async def add_litellm_data_to_request( # noqa: PLR0915 "user_api_key_model_max_budget" ] = user_api_key_dict.model_max_budget + # Extract allowed access groups for router filtering (GitHub issue #18333) + # This allows the router to filter deployments based on key's and team's access groups + # NOTE: We keep key and team access groups SEPARATE because a key doesn't always + # inherit all team access groups (per maintainer feedback). + if llm_router is not None: + from litellm.proxy.auth.model_checks import get_access_groups_from_models + + model_access_groups = llm_router.get_model_access_groups() + + # Key-level access groups (from user_api_key_dict.models) + key_models = list(user_api_key_dict.models) if user_api_key_dict.models else [] + key_allowed_access_groups = get_access_groups_from_models( + model_access_groups=model_access_groups, models=key_models + ) + if key_allowed_access_groups: + data[_metadata_variable_name][ + "user_api_key_allowed_access_groups" + ] = key_allowed_access_groups + + # Team-level access groups (from user_api_key_dict.team_models) + team_models = ( + list(user_api_key_dict.team_models) if user_api_key_dict.team_models else [] + ) + team_allowed_access_groups = get_access_groups_from_models( + model_access_groups=model_access_groups, models=team_models + ) + if team_allowed_access_groups: + data[_metadata_variable_name][ + "user_api_key_team_allowed_access_groups" + ] = team_allowed_access_groups + data[_metadata_variable_name]["user_api_key_metadata"] = user_api_key_dict.metadata _headers = dict(request.headers) _headers.pop( diff --git a/litellm/router.py b/litellm/router.py index b77e3c9c299..bd02e8e019c 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -86,6 +86,7 @@ from litellm.router_utils.clientside_credential_handler import ( is_clientside_credential, ) from litellm.router_utils.common_utils import ( + filter_deployments_by_access_groups, filter_team_based_models, filter_web_search_deployments, ) @@ -7819,10 +7820,17 @@ class Router: request_kwargs=request_kwargs, ) - verbose_router_logger.debug( - f"healthy_deployments after web search filter: {healthy_deployments}" + verbose_router_logger.debug(f"healthy_deployments after web search filter: {healthy_deployments}") + + # Filter by allowed access groups (GitHub issue #18333) + # This prevents cross-team load balancing when teams have models with same name in different access groups + healthy_deployments = filter_deployments_by_access_groups( + healthy_deployments=healthy_deployments, + request_kwargs=request_kwargs, ) + verbose_router_logger.debug(f"healthy_deployments after access group filter: {healthy_deployments}") + if isinstance(healthy_deployments, dict): return healthy_deployments diff --git a/litellm/router_utils/common_utils.py b/litellm/router_utils/common_utils.py index 10acc343abd..2c0ea5976d6 100644 --- a/litellm/router_utils/common_utils.py +++ b/litellm/router_utils/common_utils.py @@ -75,6 +75,7 @@ def filter_team_based_models( if deployment.get("model_info", {}).get("id") not in ids_to_remove ] + def _deployment_supports_web_search(deployment: Dict) -> bool: """ Check if a deployment supports web search. @@ -112,7 +113,7 @@ def filter_web_search_deployments( is_web_search_request = False tools = request_kwargs.get("tools") or [] for tool in tools: - # These are the two websearch tools for OpenAI / Azure. + # These are the two websearch tools for OpenAI / Azure. if tool.get("type") == "web_search" or tool.get("type") == "web_search_preview": is_web_search_request = True break @@ -121,8 +122,82 @@ def filter_web_search_deployments( return healthy_deployments # Filter out deployments that don't support web search - final_deployments = [d for d in healthy_deployments if _deployment_supports_web_search(d)] + final_deployments = [ + d for d in healthy_deployments if _deployment_supports_web_search(d) + ] if len(healthy_deployments) > 0 and len(final_deployments) == 0: verbose_logger.warning("No deployments support web search for request") return final_deployments + +def filter_deployments_by_access_groups( + healthy_deployments: Union[List[Dict], Dict], + request_kwargs: Optional[Dict] = None, +) -> Union[List[Dict], Dict]: + """ + Filter deployments to only include those matching the user's allowed access groups. + + Reads from TWO separate metadata fields (per maintainer feedback): + - `user_api_key_allowed_access_groups`: Access groups from the API Key's models. + - `user_api_key_team_allowed_access_groups`: Access groups from the Team's models. + + A deployment is included if its access_groups overlap with EITHER the key's + or the team's allowed access groups. Deployments with no access_groups are + always included (not restricted). + + This prevents cross-team load balancing when multiple teams have models with + the same name but in different access groups (GitHub issue #18333). + """ + if request_kwargs is None: + return healthy_deployments + + if isinstance(healthy_deployments, dict): + return healthy_deployments + + metadata = request_kwargs.get("metadata") or {} + litellm_metadata = request_kwargs.get("litellm_metadata") or {} + + # Gather key-level allowed access groups + key_allowed_access_groups = ( + metadata.get("user_api_key_allowed_access_groups") + or litellm_metadata.get("user_api_key_allowed_access_groups") + or [] + ) + + # Gather team-level allowed access groups + team_allowed_access_groups = ( + metadata.get("user_api_key_team_allowed_access_groups") + or litellm_metadata.get("user_api_key_team_allowed_access_groups") + or [] + ) + + # Combine both for the final allowed set + combined_allowed_access_groups = list(key_allowed_access_groups) + list( + team_allowed_access_groups + ) + + # If no access groups specified from either source, return all deployments (backwards compatible) + if not combined_allowed_access_groups: + return healthy_deployments + + allowed_set = set(combined_allowed_access_groups) + filtered = [] + for deployment in healthy_deployments: + model_info = deployment.get("model_info") or {} + deployment_access_groups = model_info.get("access_groups") or [] + + # If deployment has no access groups, include it (not restricted) + if not deployment_access_groups: + filtered.append(deployment) + continue + + # Include if any of deployment's groups overlap with allowed groups + if set(deployment_access_groups) & allowed_set: + filtered.append(deployment) + + if len(healthy_deployments) > 0 and len(filtered) == 0: + verbose_logger.warning( + f"No deployments match allowed access groups {combined_allowed_access_groups}" + ) + + return filtered diff --git a/tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py b/tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py new file mode 100644 index 00000000000..9ac5072c5d8 --- /dev/null +++ b/tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py @@ -0,0 +1,227 @@ +""" +Unit tests for filter_deployments_by_access_groups function. + +Tests the fix for GitHub issue #18333: Models loadbalanced outside of Model Access Group. +""" + +import pytest + +from litellm.router_utils.common_utils import filter_deployments_by_access_groups + + +class TestFilterDeploymentsByAccessGroups: + """Tests for the filter_deployments_by_access_groups function.""" + + def test_no_filter_when_no_access_groups_in_metadata(self): + """When no allowed_access_groups in metadata, return all deployments.""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1"]}}, + {"model_info": {"id": "2", "access_groups": ["AG2"]}}, + ] + request_kwargs = {"metadata": {"user_api_key_team_id": "team-1"}} + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + assert len(result) == 2 # All deployments returned + + def test_filter_to_single_access_group(self): + """Filter to only deployments matching allowed access group.""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1"]}}, + {"model_info": {"id": "2", "access_groups": ["AG2"]}}, + ] + request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + assert len(result) == 1 + assert result[0]["model_info"]["id"] == "2" + + def test_filter_with_multiple_allowed_groups(self): + """Filter with multiple allowed access groups.""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1"]}}, + {"model_info": {"id": "2", "access_groups": ["AG2"]}}, + {"model_info": {"id": "3", "access_groups": ["AG3"]}}, + ] + request_kwargs = { + "metadata": {"user_api_key_allowed_access_groups": ["AG1", "AG2"]} + } + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + assert len(result) == 2 + ids = [d["model_info"]["id"] for d in result] + assert "1" in ids + assert "2" in ids + assert "3" not in ids + + def test_deployment_with_multiple_access_groups(self): + """Deployment with multiple access groups should match if any overlap.""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1", "AG2"]}}, + {"model_info": {"id": "2", "access_groups": ["AG3"]}}, + ] + request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + assert len(result) == 1 + assert result[0]["model_info"]["id"] == "1" + + def test_deployment_without_access_groups_included(self): + """Deployments without access groups should be included (not restricted).""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1"]}}, + {"model_info": {"id": "2"}}, # No access_groups + {"model_info": {"id": "3", "access_groups": []}}, # Empty access_groups + ] + request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + # Should include deployments 2 and 3 (no restrictions) + assert len(result) == 2 + ids = [d["model_info"]["id"] for d in result] + assert "2" in ids + assert "3" in ids + + def test_dict_deployment_passes_through(self): + """When deployment is a dict (specific deployment), pass through.""" + deployment = {"model_info": {"id": "1", "access_groups": ["AG1"]}} + request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} + + result = filter_deployments_by_access_groups( + healthy_deployments=deployment, + request_kwargs=request_kwargs, + ) + + assert result == deployment # Unchanged + + def test_none_request_kwargs_passes_through(self): + """When request_kwargs is None, return deployments unchanged.""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1"]}}, + ] + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=None, + ) + + assert result == deployments + + def test_litellm_metadata_fallback(self): + """Should also check litellm_metadata for allowed access groups.""" + deployments = [ + {"model_info": {"id": "1", "access_groups": ["AG1"]}}, + {"model_info": {"id": "2", "access_groups": ["AG2"]}}, + ] + request_kwargs = { + "litellm_metadata": {"user_api_key_allowed_access_groups": ["AG1"]} + } + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + assert len(result) == 1 + assert result[0]["model_info"]["id"] == "1" + + +def test_filter_deployments_by_access_groups_issue_18333(): + """ + Regression test for GitHub issue #18333. + + Scenario: Two models named 'gpt-5' in different access groups (AG1, AG2). + Team2 has access to AG2 only. When Team2 requests 'gpt-5', only the AG2 + deployment should be available for load balancing. + """ + deployments = [ + { + "model_name": "gpt-5", + "litellm_params": {"model": "gpt-4.1", "api_key": "key-1"}, + "model_info": {"id": "ag1-deployment", "access_groups": ["AG1"]}, + }, + { + "model_name": "gpt-5", + "litellm_params": {"model": "gpt-4o", "api_key": "key-2"}, + "model_info": {"id": "ag2-deployment", "access_groups": ["AG2"]}, + }, + ] + + # Team2's request with allowed access groups + request_kwargs = { + "metadata": { + "user_api_key_team_id": "team-2", + "user_api_key_allowed_access_groups": ["AG2"], + } + } + + result = filter_deployments_by_access_groups( + healthy_deployments=deployments, + request_kwargs=request_kwargs, + ) + + # Only AG2 deployment should be returned + assert len(result) == 1 + assert result[0]["model_info"]["id"] == "ag2-deployment" + assert result[0]["litellm_params"]["model"] == "gpt-4o" + + +def test_get_access_groups_from_models(): + """ + Test the helper function that extracts access group names from models list. + This is used by the proxy to populate user_api_key_allowed_access_groups. + """ + from litellm.proxy.auth.model_checks import get_access_groups_from_models + + # Setup: access groups definition + model_access_groups = { + "AG1": ["gpt-4", "gpt-5"], + "AG2": ["claude-v1", "claude-v2"], + "beta-models": ["gpt-5-turbo"], + } + + # Test 1: Extract access groups from models list + models = ["gpt-4", "AG1", "AG2", "some-other-model"] + result = get_access_groups_from_models( + model_access_groups=model_access_groups, models=models + ) + assert set(result) == {"AG1", "AG2"} + + # Test 2: No access groups in models list + models = ["gpt-4", "claude-v1", "some-model"] + result = get_access_groups_from_models( + model_access_groups=model_access_groups, models=models + ) + assert result == [] + + # Test 3: Empty models list + result = get_access_groups_from_models( + model_access_groups=model_access_groups, models=[] + ) + assert result == [] + + # Test 4: All access groups + models = ["AG1", "AG2", "beta-models"] + result = get_access_groups_from_models( + model_access_groups=model_access_groups, models=models + ) + assert set(result) == {"AG1", "AG2", "beta-models"} From d76f3acb8052a2ce3e26c7a7110a554badac514d Mon Sep 17 00:00:00 2001 From: choigawoon Date: Fri, 16 Jan 2026 07:15:25 +0900 Subject: [PATCH 13/73] fix: video status/content credential injection for wildcard models (#18854) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * fix: video status/content credential injection for wildcard models When using wildcard model patterns like `vertex_ai/*`, the video status and content endpoints failed to resolve the model_name correctly, causing credential injection to be skipped. Changes: - router.py: Added `custom_llm_provider` parameter to `resolve_model_name_from_model_id` method - router.py: Added Strategy 2 (provider prefix matching) and Strategy 4 (wildcard pattern matching) - endpoints.py: Pass `provider_from_id` to resolver in video_status, video_content, and video_remix endpoints This allows video_id like `vertex_ai:veo-3.0-generate-preview:...` to correctly match `vertex_ai/*` wildcard pattern and inject credentials from the model config. Fixes: Video status returns "Your default credentials were not found" when using Vertex AI video generation with wildcard model patterns. * pr18845-video기능버그픽스 (vibe-kanban e43e2d2d) pr코멘트 대응 litellm fork해서 branch만들고 작업후 pull request를 올렸는데 피드백을줬어. 이 내용 파악해서 내가 올린 pr 브랜치에 해당 작업 이어서 해야할거같아. https://github.com/BerriAI/litellm/pull/18854#discussion\_r2677026995 여기 내용 읽고 현황 파악해서 작업하자. 테스트코드 작성해달라는데 테스트코드작성후 로컬에서 테스트명령어 한번 돌리고 커밋 푸시하려고. litellm에서 pull request를 위한 문서가 있어. https://docs.litellm.ai/docs/extras/contributing\_code CRA서명은 했어. 그다음거부터 양식에 맞게 해야할듯. 지금 버그만 바로 고쳐서 pr했거든. * fix: resolve mypy type error in resolve_model_name_from_model_id Rename loop variable to avoid type conflict between DeploymentTypedDict and Dict[Any, Any] from pattern_router.route() return type. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 --------- Co-authored-by: Claude Opus 4.5 --- litellm/proxy/video_endpoints/endpoints.py | 12 +- litellm/router.py | 35 +++- tests/test_litellm/test_router.py | 187 +++++++++++++++++++++ 3 files changed, 227 insertions(+), 7 deletions(-) diff --git a/litellm/proxy/video_endpoints/endpoints.py b/litellm/proxy/video_endpoints/endpoints.py index 5e00eb58455..a3c4af9ae5d 100644 --- a/litellm/proxy/video_endpoints/endpoints.py +++ b/litellm/proxy/video_endpoints/endpoints.py @@ -256,7 +256,9 @@ async def video_status( # Resolve model_name from model_id if available # This allows the router to automatically inject litellm_params from the model config if model_id_from_decoded and llm_router: - resolved_model = llm_router.resolve_model_name_from_model_id(model_id_from_decoded) + resolved_model = llm_router.resolve_model_name_from_model_id( + model_id_from_decoded, custom_llm_provider=provider_from_id + ) if resolved_model: data["model"] = resolved_model @@ -354,7 +356,9 @@ async def video_content( # Resolve model_name from model_id if available # This allows the router to automatically inject litellm_params from the model config if model_id_from_decoded and llm_router: - resolved_model = llm_router.resolve_model_name_from_model_id(model_id_from_decoded) + resolved_model = llm_router.resolve_model_name_from_model_id( + model_id_from_decoded, custom_llm_provider=provider_from_id + ) if resolved_model: data["model"] = resolved_model # Process request using ProxyBaseLLMRequestProcessing @@ -466,7 +470,9 @@ async def video_remix( # Resolve model_name from model_id if available # This allows the router to automatically inject litellm_params from the model config if model_id_from_decoded and llm_router: - resolved_model = llm_router.resolve_model_name_from_model_id(model_id_from_decoded) + resolved_model = llm_router.resolve_model_name_from_model_id( + model_id_from_decoded, custom_llm_provider=provider_from_id + ) if resolved_model: data["model"] = resolved_model diff --git a/litellm/router.py b/litellm/router.py index bd02e8e019c..f73d907c8c7 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -6972,7 +6972,7 @@ class Router: return candidate_id in self.model_id_to_deployment_index_map def resolve_model_name_from_model_id( - self, model_id: Optional[str] + self, model_id: Optional[str], custom_llm_provider: Optional[str] = None ) -> Optional[str]: """ Resolve model_name from model_id. @@ -6982,12 +6982,15 @@ class Router: Strategy: 1. First, check if model_id directly matches a model_name or deployment ID - 2. If not, search through router's model_list to find a match by litellm_params.model - 3. Return the model_name if found, None otherwise + 2. If custom_llm_provider is provided, check with provider prefix + 3. Search through router's model_list to find a match by litellm_params.model + 4. If custom_llm_provider is provided, try to find a wildcard pattern match + 5. Return the model_name if found, None otherwise Args: model_id: The model_id extracted from decoded video_id (could be model_name or litellm_params.model value) + custom_llm_provider: The provider name (e.g., "vertex_ai") for wildcard matching Returns: model_name if found, None otherwise. If None, the request will fall through @@ -7000,15 +7003,26 @@ class Router: if model_id in self.model_names or self.has_model_id(model_id): return model_id - # Strategy 2: Search through router's model_list to find by litellm_params.model + # Strategy 2: Check with provider prefix (e.g., "vertex_ai/veo-3.0-generate-preview") + if custom_llm_provider: + full_model_name = f"{custom_llm_provider}/{model_id}" + if full_model_name in self.model_names or self.has_model_id(full_model_name): + return full_model_name + + # Strategy 3: Search through router's model_list to find by litellm_params.model all_models = self.get_model_list(model_name=None) if not all_models: return None + # First pass: exact matches (non-wildcard) for deployment in all_models: litellm_params = deployment.get("litellm_params", {}) actual_model = litellm_params.get("model") + # Skip wildcard patterns in first pass + if actual_model and actual_model.endswith("/*"): + continue + # Match by exact match or by checking if actual_model ends with /model_id or :model_id # e.g., model_id="veo-2.0-generate-001" matches actual_model="vertex_ai/veo-2.0-generate-001" matches = ( @@ -7022,6 +7036,19 @@ class Router: if model_name: return model_name + # Strategy 4: Wildcard patterns using PatternMatchRouter + # For video status/content, we need to match model_id like "veo-3.0-generate-preview" + # to wildcard patterns like "vertex_ai/*" + if custom_llm_provider: + full_model_name = f"{custom_llm_provider}/{model_id}" + pattern_deployments = self.pattern_router.route(full_model_name) + if pattern_deployments: + # Return the first matching wildcard model_name + for pattern_deployment in pattern_deployments: + matched_model_name = pattern_deployment.get("model_name") + if matched_model_name: + return matched_model_name + # No match found return None diff --git a/tests/test_litellm/test_router.py b/tests/test_litellm/test_router.py index 6279e96305f..7201b961588 100644 --- a/tests/test_litellm/test_router.py +++ b/tests/test_litellm/test_router.py @@ -2054,3 +2054,190 @@ async def test_aguardrail(): assert result["result"] == "success" assert result["selected_guardrail"]["id"] == "guardrail-1" + + +def test_resolve_model_name_from_model_id_wildcard_pattern(): + """ + Test that resolve_model_name_from_model_id correctly resolves model names + for wildcard patterns using PatternMatchRouter. + + This is critical for video status/content endpoints where model_id extracted + from video_id (e.g., "veo-3.0-generate-preview") needs to match wildcard + patterns like "vertex_ai/*" to inject credentials from the model config. + """ + # Set up router with wildcard pattern + router = litellm.Router( + model_list=[ + { + "model_name": "vertex_ai/*", + "litellm_params": { + "model": "vertex_ai/*", + "vertex_project": "test-project", + "vertex_location": "us-central1", + }, + }, + { + "model_name": "specific-model", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "specific-project", + "vertex_location": "us-east1", + }, + }, + ], + ) + + # Test Case 1: Wildcard pattern matching with custom_llm_provider + # This simulates video_id like "vertex_ai:veo-3.0-generate-preview:..." + result = router.resolve_model_name_from_model_id( + model_id="veo-3.0-generate-preview", + custom_llm_provider="vertex_ai", + ) + assert result == "vertex_ai/*", f"Expected 'vertex_ai/*', got '{result}'" + + # Test Case 2: Different model name should also match wildcard + result = router.resolve_model_name_from_model_id( + model_id="gemini-2.0-flash", + custom_llm_provider="vertex_ai", + ) + assert result == "vertex_ai/*", f"Expected 'vertex_ai/*', got '{result}'" + + # Test Case 3: Without custom_llm_provider, should not match wildcard + result = router.resolve_model_name_from_model_id( + model_id="veo-3.0-generate-preview", + custom_llm_provider=None, + ) + assert result is None, f"Expected None without provider, got '{result}'" + + # Test Case 4: Exact model_name match should take precedence + result = router.resolve_model_name_from_model_id( + model_id="specific-model", + custom_llm_provider="vertex_ai", + ) + assert result == "specific-model", f"Expected 'specific-model', got '{result}'" + + +def test_resolve_model_name_from_model_id_exact_match(): + """ + Test that resolve_model_name_from_model_id correctly resolves exact model names. + """ + router = litellm.Router( + model_list=[ + { + "model_name": "my-gpt-model", + "litellm_params": { + "model": "azure/gpt-4", + "api_key": "test-key", + }, + }, + { + "model_name": "veo-model", + "litellm_params": { + "model": "vertex_ai/veo-2.0-generate-001", + "vertex_project": "test-project", + }, + }, + ], + ) + + # Test Case 1: Direct model_name match + result = router.resolve_model_name_from_model_id(model_id="my-gpt-model") + assert result == "my-gpt-model", f"Expected 'my-gpt-model', got '{result}'" + + # Test Case 2: Match by litellm_params.model suffix + result = router.resolve_model_name_from_model_id(model_id="veo-2.0-generate-001") + assert result == "veo-model", f"Expected 'veo-model', got '{result}'" + + # Test Case 3: Non-existent model should return None + result = router.resolve_model_name_from_model_id(model_id="non-existent-model") + assert result is None, f"Expected None, got '{result}'" + + +def test_resolve_model_name_from_model_id_provider_prefix(): + """ + Test that resolve_model_name_from_model_id handles provider prefix correctly. + """ + router = litellm.Router( + model_list=[ + { + "model_name": "vertex_ai/gemini-pro", + "litellm_params": { + "model": "vertex_ai/gemini-pro", + "vertex_project": "test-project", + }, + }, + ], + ) + + # Test Case 1: Full model name with provider prefix as model_name + result = router.resolve_model_name_from_model_id( + model_id="vertex_ai/gemini-pro", + custom_llm_provider=None, + ) + assert result == "vertex_ai/gemini-pro", f"Expected 'vertex_ai/gemini-pro', got '{result}'" + + # Test Case 2: Model ID with provider prefix constructed from custom_llm_provider + result = router.resolve_model_name_from_model_id( + model_id="gemini-pro", + custom_llm_provider="vertex_ai", + ) + assert result == "vertex_ai/gemini-pro", f"Expected 'vertex_ai/gemini-pro', got '{result}'" + + +def test_resolve_model_name_from_model_id_multiple_wildcards(): + """ + Test that resolve_model_name_from_model_id works with multiple wildcard patterns. + """ + router = litellm.Router( + model_list=[ + { + "model_name": "vertex_ai/*", + "litellm_params": { + "model": "vertex_ai/*", + "vertex_project": "vertex-project", + }, + }, + { + "model_name": "openai/*", + "litellm_params": { + "model": "openai/*", + "api_key": "openai-key", + }, + }, + { + "model_name": "anthropic/*", + "litellm_params": { + "model": "anthropic/*", + "api_key": "anthropic-key", + }, + }, + ], + ) + + # Test Case 1: Match vertex_ai wildcard + result = router.resolve_model_name_from_model_id( + model_id="veo-3.0-generate-preview", + custom_llm_provider="vertex_ai", + ) + assert result == "vertex_ai/*", f"Expected 'vertex_ai/*', got '{result}'" + + # Test Case 2: Match openai wildcard + result = router.resolve_model_name_from_model_id( + model_id="gpt-4o", + custom_llm_provider="openai", + ) + assert result == "openai/*", f"Expected 'openai/*', got '{result}'" + + # Test Case 3: Match anthropic wildcard + result = router.resolve_model_name_from_model_id( + model_id="claude-3-opus", + custom_llm_provider="anthropic", + ) + assert result == "anthropic/*", f"Expected 'anthropic/*', got '{result}'" + + # Test Case 4: Non-matching provider should return None + result = router.resolve_model_name_from_model_id( + model_id="some-model", + custom_llm_provider="bedrock", + ) + assert result is None, f"Expected None for non-matching provider, got '{result}'" From 812ac7e838643d4d19a612e5167877f61681a0af Mon Sep 17 00:00:00 2001 From: yuneng-jiang Date: Thu, 15 Jan 2026 15:19:03 -0800 Subject: [PATCH 14/73] Reusable model select --- .../app/(dashboard)/hooks/models/useModels.ts | 34 +- .../hooks/organizations/useOrganizations.ts | 32 +- .../app/(dashboard)/hooks/teams/useTeams.ts | 30 +- .../ModelSelect/ModelSelect.test.tsx | 367 ++++++++++++++++++ .../components/ModelSelect/ModelSelect.tsx | 157 ++++++++ .../components/ModelSelect/modelUtils.test.ts | 67 ++++ .../src/components/ModelSelect/modelUtils.ts | 21 + .../src/components/organizations.tsx | 19 +- 8 files changed, 707 insertions(+), 20 deletions(-) create mode 100644 ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx create mode 100644 ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.tsx create mode 100644 ui/litellm-dashboard/src/components/ModelSelect/modelUtils.test.ts create mode 100644 ui/litellm-dashboard/src/components/ModelSelect/modelUtils.ts diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/models/useModels.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/models/useModels.ts index 9c7ddf18f54..fa7ab911ecd 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/hooks/models/useModels.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/models/useModels.ts @@ -1,9 +1,23 @@ import { useQuery } from "@tanstack/react-query"; import { createQueryKeys } from "../common/queryKeysFactory"; -import { modelInfoCall, modelHubCall } from "@/components/networking"; +import { modelInfoCall, modelHubCall, modelAvailableCall } from "@/components/networking"; import useAuthorized from "../useAuthorized"; + +export interface ProxyModel { + id: string; + object: string; + created: number; + owned_by: string; +} + +export interface AllProxyModelsResponse { + data: ProxyModel[]; +} + const modelKeys = createQueryKeys("models"); const modelHubKeys = createQueryKeys("modelHub"); +const allProxyModelsKeys = createQueryKeys("allProxyModels"); +const selectedTeamModelsKeys = createQueryKeys("selectedTeamModels"); export const useModelsInfo = () => { const { accessToken, userId, userRole } = useAuthorized(); @@ -27,3 +41,21 @@ export const useModelHub = () => { enabled: Boolean(accessToken), }); }; + +export const useAllProxyModels = () => { + const { accessToken, userId, userRole } = useAuthorized(); + return useQuery({ + queryKey: allProxyModelsKeys.list({}), + queryFn: async () => await modelAvailableCall(accessToken!, userId!, userRole!, true), + enabled: Boolean(accessToken && userId && userRole), + }); +}; + +export const useSelectedTeamModels = (teamID: string | null) => { + const { accessToken, userId, userRole } = useAuthorized(); + return useQuery({ + queryKey: selectedTeamModelsKeys.list({}), + queryFn: async () => await modelAvailableCall(accessToken!, userId!, userRole!, true, teamID!), + enabled: Boolean(accessToken && userId && userRole && teamID), + }); +}; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/organizations/useOrganizations.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/organizations/useOrganizations.ts index 27a946d112a..323270f4360 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/hooks/organizations/useOrganizations.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/organizations/useOrganizations.ts @@ -1,10 +1,9 @@ -import { useQuery, UseQueryResult } from "@tanstack/react-query"; -import { createQueryKeys } from "../common/queryKeysFactory"; -import { organizationListCall, Organization } from "@/components/networking"; import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; +import { Organization, organizationInfoCall, organizationListCall } from "@/components/networking"; +import { useQuery, useQueryClient, UseQueryResult } from "@tanstack/react-query"; +import { createQueryKeys } from "../common/queryKeysFactory"; const organizationKeys = createQueryKeys("organizations"); - export const useOrganizations = (): UseQueryResult => { const { accessToken, userId, userRole } = useAuthorized(); return useQuery({ @@ -13,3 +12,28 @@ export const useOrganizations = (): UseQueryResult => { enabled: Boolean(accessToken && userId && userRole), }); }; + +export const useOrganization = (organizationID?: string) => { + const queryClient = useQueryClient(); + const { accessToken } = useAuthorized(); + return useQuery({ + queryKey: organizationKeys.detail(organizationID!), + enabled: Boolean(accessToken && organizationID), + + queryFn: async () => { + if (!accessToken || !organizationID) { + throw new Error("Missing auth or teamId"); + } + + return organizationInfoCall(accessToken, organizationID); + }, + + initialData: () => { + if (!organizationID) return undefined; + + const organizations = queryClient.getQueryData(organizationKeys.list({})); + + return organizations?.find((organization: Organization) => organization.organization_id === organizationID); + }, + }); +}; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/hooks/teams/useTeams.ts b/ui/litellm-dashboard/src/app/(dashboard)/hooks/teams/useTeams.ts index 5d2008a4d29..2beebb18718 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/hooks/teams/useTeams.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/hooks/teams/useTeams.ts @@ -1,17 +1,41 @@ -import { useQuery, UseQueryResult } from "@tanstack/react-query"; +import { useQuery, useQueryClient, UseQueryResult } from "@tanstack/react-query"; import { Team } from "@/components/key_team_helpers/key_list"; import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; import { fetchTeams } from "@/app/(dashboard)/networking"; import { createQueryKeys } from "@/app/(dashboard)/hooks/common/queryKeysFactory"; +import { teamInfoCall } from "@/components/networking"; const teamKeys = createQueryKeys("teams"); - export const useTeams = (): UseQueryResult => { const { accessToken, userId, userRole } = useAuthorized(); - return useQuery({ queryKey: teamKeys.list({}), queryFn: async () => await fetchTeams(accessToken!, userId, userRole, null), enabled: Boolean(accessToken), }); }; + +export const useTeam = (teamId?: string) => { + const { accessToken } = useAuthorized(); + const queryClient = useQueryClient(); + return useQuery({ + queryKey: teamKeys.detail(teamId!), + enabled: Boolean(accessToken && teamId), + + queryFn: async () => { + if (!accessToken || !teamId) { + throw new Error("Missing auth or teamId"); + } + + return teamInfoCall(accessToken, teamId); + }, + + initialData: () => { + if (!teamId) return undefined; + + const teams = queryClient.getQueryData(teamKeys.list({})); + + return teams?.find((team) => team.team_id === teamId); + }, + }); +}; diff --git a/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx b/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx new file mode 100644 index 00000000000..4afc6ea7d27 --- /dev/null +++ b/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx @@ -0,0 +1,367 @@ +import { describe, it, expect, vi, beforeEach } from "vitest"; +import { screen, waitFor } from "@testing-library/react"; +import userEvent from "@testing-library/user-event"; +import { renderWithProviders } from "../../../tests/test-utils"; +import { ModelSelect } from "./ModelSelect"; +import type { ProxyModel } from "@/app/(dashboard)/hooks/models/useModels"; +import type { Organization } from "@/components/networking"; +import type { Team } from "@/components/key_team_helpers/key_list"; + +vi.mock("@/app/(dashboard)/hooks/models/useModels", () => ({ + useAllProxyModels: vi.fn(), +})); + +vi.mock("@/app/(dashboard)/hooks/teams/useTeams", () => ({ + useTeam: vi.fn(), +})); + +vi.mock("@/app/(dashboard)/hooks/organizations/useOrganizations", () => ({ + useOrganization: vi.fn(), +})); + +vi.mock("antd", async (importOriginal) => { + const actual = await importOriginal(); + return { + ...actual, + Select: ({ + value, + onChange, + options, + "data-testid": dataTestId, + allowClear, + maxTagCount, + maxTagPlaceholder, + mode, + ...props + }: any) => { + return ( +
+ +
+ ); + }, + Skeleton: { + Input: ({ active, block }: any) =>
, + }, + Tooltip: ({ children }: { children: React.ReactNode }) => <>{children}, + }; +}); + +import { useAllProxyModels } from "@/app/(dashboard)/hooks/models/useModels"; +import { useTeam } from "@/app/(dashboard)/hooks/teams/useTeams"; +import { useOrganization } from "@/app/(dashboard)/hooks/organizations/useOrganizations"; + +const mockUseAllProxyModels = vi.mocked(useAllProxyModels); +const mockUseTeam = vi.mocked(useTeam); +const mockUseOrganization = vi.mocked(useOrganization); + +describe("ModelSelect", () => { + const mockProxyModels: ProxyModel[] = [ + { id: "gpt-4", object: "model", created: 1234567890, owned_by: "openai" }, + { id: "claude-3", object: "model", created: 1234567890, owned_by: "anthropic" }, + { id: "openai/*", object: "model", created: 1234567890, owned_by: "openai" }, + { id: "anthropic/*", object: "model", created: 1234567890, owned_by: "anthropic" }, + ]; + + const mockOnChange = vi.fn(); + + beforeEach(() => { + vi.clearAllMocks(); + mockUseAllProxyModels.mockReturnValue({ + data: { data: mockProxyModels }, + isLoading: false, + } as any); + mockUseTeam.mockReturnValue({ + data: undefined, + isLoading: false, + } as any); + mockUseOrganization.mockReturnValue({ + data: undefined, + isLoading: false, + } as any); + }); + + it("should render", async () => { + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByTestId("model-select")).toBeInTheDocument(); + }); + }); + + it("should show skeleton loader when loading", () => { + mockUseAllProxyModels.mockReturnValue({ + data: undefined, + isLoading: true, + } as any); + + renderWithProviders(); + + expect(screen.getByTestId("skeleton-input")).toBeInTheDocument(); + expect(screen.queryByTestId("model-select")).not.toBeInTheDocument(); + }); + + it("should show skeleton loader when team is loading", () => { + mockUseTeam.mockReturnValue({ + data: undefined, + isLoading: true, + } as any); + + renderWithProviders(); + + expect(screen.getByTestId("skeleton-input")).toBeInTheDocument(); + }); + + it("should show skeleton loader when organization is loading", () => { + mockUseOrganization.mockReturnValue({ + data: undefined, + isLoading: true, + } as any); + + renderWithProviders(); + + expect(screen.getByTestId("skeleton-input")).toBeInTheDocument(); + }); + + it("should render special options group", async () => { + renderWithProviders(); + + await waitFor(() => { + const select = screen.getByTestId("model-select"); + expect(select).toBeInTheDocument(); + expect(screen.getByText("All Proxy Models")).toBeInTheDocument(); + expect(screen.getByText("No Default Models")).toBeInTheDocument(); + }); + }); + + it("should render wildcard options group", async () => { + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByText("All Openai models")).toBeInTheDocument(); + expect(screen.getByText("All Anthropic models")).toBeInTheDocument(); + }); + }); + + it("should render regular models group", async () => { + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByText("gpt-4")).toBeInTheDocument(); + expect(screen.getByText("claude-3")).toBeInTheDocument(); + }); + }); + + it("should call onChange when selecting a regular model", async () => { + const user = userEvent.setup(); + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByTestId("model-select")).toBeInTheDocument(); + }); + + const select = screen.getByRole("listbox"); + await user.selectOptions(select, "gpt-4"); + + expect(mockOnChange).toHaveBeenCalledWith(["gpt-4"]); + }); + + it("should call onChange with only last special option when multiple special options are selected", async () => { + const user = userEvent.setup(); + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByTestId("model-select")).toBeInTheDocument(); + }); + + const select = screen.getByRole("listbox"); + await user.selectOptions(select, ["all-proxy-models", "no-default-models"]); + + expect(mockOnChange).toHaveBeenCalledWith(["no-default-models"]); + }); + + it("should disable regular models when special option is selected", async () => { + renderWithProviders( + , + ); + + await waitFor(() => { + const gpt4Option = screen.getByRole("option", { name: "gpt-4" }); + expect(gpt4Option).toBeDisabled(); + }); + }); + + it("should disable wildcard models when special option is selected", async () => { + renderWithProviders( + , + ); + + await waitFor(() => { + const openaiWildcardOption = screen.getByRole("option", { name: "All Openai models" }); + expect(openaiWildcardOption).toBeDisabled(); + }); + }); + + it("should disable other special options when one special option is selected", async () => { + renderWithProviders( + , + ); + + await waitFor(() => { + const noDefaultOption = screen.getByRole("option", { name: "No Default Models" }); + expect(noDefaultOption).toBeDisabled(); + }); + }); + + it("should filter models when showAllProxyModelsOverride is true", async () => { + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByText("gpt-4")).toBeInTheDocument(); + expect(screen.getByText("claude-3")).toBeInTheDocument(); + }); + }); + + it("should filter models when organization has all-proxy-models in models array", async () => { + const mockOrganization: Organization = { + organization_id: "org-1", + organization_alias: "Test Org", + budget_id: "budget-1", + metadata: {}, + models: ["all-proxy-models"], + spend: 0, + model_spend: {}, + created_at: "2024-01-01", + created_by: "user-1", + updated_at: "2024-01-01", + updated_by: "user-1", + litellm_budget_table: null, + teams: null, + users: null, + members: null, + }; + + mockUseOrganization.mockReturnValue({ + data: mockOrganization, + isLoading: false, + } as any); + + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByText("gpt-4")).toBeInTheDocument(); + expect(screen.getByText("claude-3")).toBeInTheDocument(); + }); + }); + + it("should return empty models array when organization does not have all-proxy-models", async () => { + const mockOrganization: Organization = { + organization_id: "org-1", + organization_alias: "Test Org", + budget_id: "budget-1", + metadata: {}, + models: ["gpt-4"], + spend: 0, + model_spend: {}, + created_at: "2024-01-01", + created_by: "user-1", + updated_at: "2024-01-01", + updated_by: "user-1", + litellm_budget_table: null, + teams: null, + users: null, + members: null, + }; + + mockUseOrganization.mockReturnValue({ + data: mockOrganization, + isLoading: false, + } as any); + + renderWithProviders(); + + await waitFor(() => { + expect(screen.queryByText("gpt-4")).not.toBeInTheDocument(); + expect(screen.queryByText("claude-3")).not.toBeInTheDocument(); + }); + }); + + it("should use custom dataTestId when provided", async () => { + renderWithProviders( + , + ); + + await waitFor(() => { + expect(screen.getByTestId("custom-test-id")).toBeInTheDocument(); + }); + }); + + it("should handle multiple model selections", async () => { + const user = userEvent.setup(); + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByTestId("model-select")).toBeInTheDocument(); + }); + + const select = screen.getByRole("listbox"); + await user.selectOptions(select, "gpt-4"); + expect(mockOnChange).toHaveBeenCalledWith(["gpt-4"]); + + await user.selectOptions(select, "claude-3"); + expect(mockOnChange).toHaveBeenCalled(); + const allCalls = mockOnChange.mock.calls.map((call) => call[0]); + expect(allCalls.some((call) => Array.isArray(call) && call.includes("gpt-4"))).toBe(true); + expect(allCalls.some((call) => Array.isArray(call) && call.includes("claude-3"))).toBe(true); + }); + + it("should capitalize provider name in wildcard options", async () => { + renderWithProviders(); + + await waitFor(() => { + expect(screen.getByText("All Openai models")).toBeInTheDocument(); + expect(screen.getByText("All Anthropic models")).toBeInTheDocument(); + }); + }); + + it("should deduplicate models with same id", async () => { + const duplicateModels: ProxyModel[] = [ + { id: "gpt-4", object: "model", created: 1234567890, owned_by: "openai" }, + { id: "gpt-4", object: "model", created: 1234567890, owned_by: "openai" }, + ]; + + mockUseAllProxyModels.mockReturnValue({ + data: { data: duplicateModels }, + isLoading: false, + } as any); + + renderWithProviders(); + + await waitFor(() => { + const gpt4Options = screen.getAllByText("gpt-4"); + expect(gpt4Options.length).toBeGreaterThan(0); + }); + }); +}); diff --git a/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.tsx b/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.tsx new file mode 100644 index 00000000000..5aa1ba6a30a --- /dev/null +++ b/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.tsx @@ -0,0 +1,157 @@ +import { ProxyModel, useAllProxyModels } from "@/app/(dashboard)/hooks/models/useModels"; +import { useTeam } from "@/app/(dashboard)/hooks/teams/useTeams"; +import { Select, Skeleton, Tooltip, type SelectProps } from "antd"; +import { Organization, Team } from "../networking"; +import { useOrganization } from "@/app/(dashboard)/hooks/organizations/useOrganizations"; +import { splitWildcardModels } from "./modelUtils"; + +const MODEL_SELECT_SPECIAL_VALUES = { + ALL_PROXY_MODELS: { + label: "All Proxy Models", + value: "all-proxy-models", + }, + NO_DEFAULT_MODELS: { + label: "No Default Models", + value: "no-default-models", + }, +}; + +const MODEL_SELECT_SPECIAL_VALUES_ARRAY = Object.values(MODEL_SELECT_SPECIAL_VALUES); + +export interface ModelSelectContext { + teamID?: string; + organizationID?: string; + includeUserModels?: boolean; + showAllTeamModelsOption?: boolean; + showAllProxyModelsOverride?: boolean; + includeSpecialOptions?: boolean; + dataTestId?: string; + value?: string[]; + onChange: (values: string[]) => void; +} + +const filterModels = ( + allProxyModels: ProxyModel[], + ctx: ModelSelectContext, + { + selectedTeam, + selectedOrganization, + userModels, + }: { selectedTeam?: Team; selectedOrganization?: Organization; userModels?: ProxyModel[] }, +): ProxyModel[] => { + const deduplicatedProxyModels = Array.from(new Map(allProxyModels.map((model) => [model.id, model])).values()); + if (ctx.showAllProxyModelsOverride) { + return deduplicatedProxyModels; + } + + if (selectedOrganization) { + if (selectedOrganization.models.includes(MODEL_SELECT_SPECIAL_VALUES.ALL_PROXY_MODELS.value)) { + return deduplicatedProxyModels; + } + } + + return []; +}; + +export const ModelSelect = (ctx: ModelSelectContext) => { + const { + teamID, + organizationID, + includeUserModels, + showAllTeamModelsOption, + showAllProxyModelsOverride, + includeSpecialOptions, + dataTestId, + value = [], + onChange, + } = ctx; + const { data: allProxyModels, isLoading: isLoadingAllProxyModels } = useAllProxyModels(); + const { data: team, isLoading: isLoadingTeam } = useTeam(teamID); + const { data: organization, isLoading: isLoadingOrganization } = useOrganization(organizationID); + + const isSpecialOption = (value: string) => MODEL_SELECT_SPECIAL_VALUES_ARRAY.some((sv) => sv.value === value); + const hasSpecialOptionSelected = value.some(isSpecialOption); + const isLoading = isLoadingAllProxyModels || isLoadingTeam || isLoadingOrganization; + + if (isLoading) { + return ; + } + + const optionRender: NonNullable = (option) => { + return {option.label}; + }; + + const handleChange = (values: string[]) => { + const specialValues = values.filter(isSpecialOption); + + let finalValues: string[]; + if (specialValues.length > 0) { + const lastSelectedSpecial = specialValues[specialValues.length - 1]; + finalValues = [lastSelectedSpecial]; + } else { + finalValues = values; + } + + onChange(finalValues); + }; + + const filteredModels = filterModels(allProxyModels?.data ?? [], ctx, { + selectedTeam: team, + selectedOrganization: organization, + }); + + const { wildcard, regular } = splitWildcardModels(filteredModels); + return ( + - {(() => { - let shouldShowAllProxyModels = false; - - if (organization) { - // Team is in an organization - if (organization.models.length === 0 || organization.models.includes("all-proxy-models")) { - // Organization has empty array [] or "all-proxy-models" - shouldShowAllProxyModels = true; - } - // Otherwise (organization has specific models), don't show "all-proxy-models" - } else { - // Team is not in an organization - shouldShowAllProxyModels = is_proxy_admin || userModels.includes("all-proxy-models"); - } - - return shouldShowAllProxyModels ? ( - - All Proxy Models - - ) : null; - })()} - {(() => { - // Show "no-default-models" option if: - // 1. Team is not in an organization, OR - // 2. Team is in an organization and organization's models include "no-default-models" - const shouldShowNoDefaultModels = - !organization || organization.models.includes("no-default-models"); - - return shouldShowNoDefaultModels ? ( - - No Default Models - - ) : null; - })()} - {Array.from(new Set(modelsToPick)).map((model, idx) => ( - - {getModelDisplayName(model)} - - ))} - + form.setFieldValue("models", values)} + teamID={teamId} + organizationID={teamData?.team_info?.organization_id || undefined} + options={{ + includeSpecialOptions: true, + includeUserModels: !teamData?.team_info?.organization_id, + showAllProxyModelsOverride: isProxyAdminRole(userRole) && !teamData?.team_info?.organization_id, + }} + context="team" + dataTestId="models-select" + /> diff --git a/ui/litellm-dashboard/src/components/view_users/types.ts b/ui/litellm-dashboard/src/components/view_users/types.ts index d674db5c7db..744aa00a88b 100644 --- a/ui/litellm-dashboard/src/components/view_users/types.ts +++ b/ui/litellm-dashboard/src/components/view_users/types.ts @@ -5,6 +5,7 @@ export interface UserInfo { user_role: string; spend: number; max_budget: number | null; + models: string[]; key_count: number; created_at: string; updated_at: string; From fba61f8e2a7cd32d810e4d79f4baf2c5cd491de9 Mon Sep 17 00:00:00 2001 From: yuneng-jiang Date: Thu, 15 Jan 2026 21:42:58 -0800 Subject: [PATCH 50/73] adding mocks --- .../ModelSelect/ModelSelect.test.tsx | 23 ++++++++----------- 1 file changed, 10 insertions(+), 13 deletions(-) diff --git a/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx b/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx index 9a56d219bac..80acf5d75ff 100644 --- a/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx +++ b/ui/litellm-dashboard/src/components/ModelSelect/ModelSelect.test.tsx @@ -192,7 +192,12 @@ describe("ModelSelect", () => { } as any); renderWithProviders( - , + , ); await waitFor(() => { @@ -392,9 +397,7 @@ describe("ModelSelect", () => { isLoading: false, } as any); - renderWithProviders( - , - ); + renderWithProviders(); await waitFor(() => { expect(screen.getByText("gpt-4")).toBeInTheDocument(); @@ -426,9 +429,7 @@ describe("ModelSelect", () => { isLoading: false, } as any); - renderWithProviders( - , - ); + renderWithProviders(); await waitFor(() => { expect(screen.getByText("gpt-4")).toBeInTheDocument(); @@ -510,9 +511,7 @@ describe("ModelSelect", () => { isLoading: false, } as any); - renderWithProviders( - , - ); + renderWithProviders(); await waitFor(() => { expect(screen.getByText("gpt-4")).toBeInTheDocument(); @@ -555,9 +554,7 @@ describe("ModelSelect", () => { isLoading: false, } as any); - renderWithProviders( - , - ); + renderWithProviders(); await waitFor(() => { expect(screen.getByText("gpt-4")).toBeInTheDocument(); From f85840a34f6f4b016f9f4cff0ba8b2aaea3d7978 Mon Sep 17 00:00:00 2001 From: Yuta Saito Date: Fri, 16 Jan 2026 14:58:36 +0900 Subject: [PATCH 51/73] =?UTF-8?q?bump:=20version=201.80.16=20=E2=86=92=201?= =?UTF-8?q?.80.17?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- pyproject.toml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/pyproject.toml b/pyproject.toml index aa8e6fd97be..88222da1f0c 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -1,6 +1,6 @@ [tool.poetry] name = "litellm" -version = "1.80.16" +version = "1.80.17" description = "Library to easily interface with LLM API providers" authors = ["BerriAI"] license = "MIT" @@ -167,7 +167,7 @@ requires = ["poetry-core", "wheel"] build-backend = "poetry.core.masonry.api" [tool.commitizen] -version = "1.80.16" +version = "1.80.17" version_files = [ "pyproject.toml:^version" ] From cf32eb573746f39035aa15a8e7af9a373014afab Mon Sep 17 00:00:00 2001 From: yuneng-jiang Date: Thu, 15 Jan 2026 22:08:45 -0800 Subject: [PATCH 52/73] refactor team member icon buttons --- .../components/team/team_member_view.test.tsx | 28 +++++++++++++++ .../src/components/team/team_member_view.tsx | 35 +++++++++---------- 2 files changed, 44 insertions(+), 19 deletions(-) diff --git a/ui/litellm-dashboard/src/components/team/team_member_view.test.tsx b/ui/litellm-dashboard/src/components/team/team_member_view.test.tsx index ba0f3132f64..30a06179c2f 100644 --- a/ui/litellm-dashboard/src/components/team/team_member_view.test.tsx +++ b/ui/litellm-dashboard/src/components/team/team_member_view.test.tsx @@ -20,6 +20,7 @@ vi.mock("@/utils/roles", () => ({ import { useUISettings } from "@/app/(dashboard)/hooks/uiSettings/useUISettings"; import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; +import { isProxyAdminRole, isUserTeamAdminForSingleTeam } from "@/utils/roles"; describe("TeamMembersComponent", () => { const mockHandleMemberDelete = vi.fn(); @@ -162,4 +163,31 @@ describe("TeamMembersComponent", () => { expect(screen.getByText("Add Member")).toBeInTheDocument(); }); + + it("should show delete button for proxy admin when canEditTeam is true", () => { + vi.mocked(isProxyAdminRole).mockReturnValue(true); + vi.mocked(isUserTeamAdminForSingleTeam).mockReturnValue(false); + + const { container } = renderWithProviders( + , + ); + + // Verify that action buttons are rendered when canEditTeam is true + // For proxy admin, both edit and delete buttons should be visible + // Check for clickable icon elements (Tremor Icon components with cursor-pointer class) + const clickableIcons = container.querySelectorAll('[class*="cursor-pointer"]'); + // Should have at least 4 icons: 2 edit buttons + 2 delete buttons for 2 members + expect(clickableIcons.length).toBeGreaterThanOrEqual(4); + + // Verify members are rendered + expect(screen.getAllByText("user1@test.com").length).toBeGreaterThan(0); + expect(screen.getAllByText("user2@test.com").length).toBeGreaterThan(0); + }); }); diff --git a/ui/litellm-dashboard/src/components/team/team_member_view.tsx b/ui/litellm-dashboard/src/components/team/team_member_view.tsx index 534d4c67e64..10b3cbd83e6 100644 --- a/ui/litellm-dashboard/src/components/team/team_member_view.tsx +++ b/ui/litellm-dashboard/src/components/team/team_member_view.tsx @@ -1,25 +1,24 @@ -import React from "react"; +import { useUISettings } from "@/app/(dashboard)/hooks/uiSettings/useUISettings"; +import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; import { Member } from "@/components/networking"; +import { formatNumberWithCommas } from "@/utils/dataUtils"; +import { isProxyAdminRole, isUserTeamAdminForSingleTeam } from "@/utils/roles"; +import { InfoCircleOutlined } from "@ant-design/icons"; import { Card, Table, - TableHead, - TableRow, - TableHeaderCell, TableBody, TableCell, + TableHead, + TableHeaderCell, + TableRow, Text, - Icon, Button as TremorButton, } from "@tremor/react"; -import { InfoCircleOutlined } from "@ant-design/icons"; import { Tooltip } from "antd"; +import React from "react"; +import TableIconActionButton from "../common_components/IconActionButton/TableIconActionButtons/TableIconActionButton"; import { TeamData } from "./team_info"; -import { PencilAltIcon, TrashIcon } from "@heroicons/react/outline"; -import { formatNumberWithCommas } from "@/utils/dataUtils"; -import { useUISettings } from "@/app/(dashboard)/hooks/uiSettings/useUISettings"; -import { isUserTeamAdminForSingleTeam, isProxyAdminRole } from "@/utils/roles"; -import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; interface TeamMembersComponentProps { teamData: TeamData; @@ -154,9 +153,9 @@ const TeamMembersComponent: React.FC = ({ {canEditTeam && (
- { // Get budget and rate limit data from team membership const membership = teamData.team_memberships.find((tm) => tm.user_id === member.user_id); @@ -169,14 +168,12 @@ const TeamMembersComponent: React.FC = ({ setSelectedEditMember(enhancedMember); setIsEditMemberModalVisible(true); }} - className="cursor-pointer hover:text-blue-600" /> {(isProxyAdmin || (isUserTeamAdmin && !disableTeamAdminDeleteTeamUser)) && ( - handleMemberDelete(member)} - className="cursor-pointer hover:text-red-600" /> )}
From c0e5637eae71895e454eb198dc2e5b8eacbf84ea Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 11:40:49 +0530 Subject: [PATCH 53/73] =?UTF-8?q?Fix:=20[Bug]:=20stream=5Ftimeout=EF=BC=9A?= =?UTF-8?q?The=20function=20of=20this=20parameter=20has=20been=20changed?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../llms/custom_httpx/aiohttp_transport.py | 1 - .../custom_httpx/test_aiohttp_transport.py | 87 +++++++++++++++++-- 2 files changed, 81 insertions(+), 7 deletions(-) diff --git a/litellm/llms/custom_httpx/aiohttp_transport.py b/litellm/llms/custom_httpx/aiohttp_transport.py index f845bf7cb90..a7b83d8c802 100644 --- a/litellm/llms/custom_httpx/aiohttp_transport.py +++ b/litellm/llms/custom_httpx/aiohttp_transport.py @@ -245,7 +245,6 @@ class LiteLLMAiohttpTransport(AiohttpTransport): allow_redirects=False, auto_decompress=False, timeout=ClientTimeout( - total=timeout.get("read"), sock_connect=timeout.get("connect"), sock_read=timeout.get("read"), connect=timeout.get("pool"), diff --git a/tests/test_litellm/llms/custom_httpx/test_aiohttp_transport.py b/tests/test_litellm/llms/custom_httpx/test_aiohttp_transport.py index f0dac113645..002fa81b9b5 100644 --- a/tests/test_litellm/llms/custom_httpx/test_aiohttp_transport.py +++ b/tests/test_litellm/llms/custom_httpx/test_aiohttp_transport.py @@ -333,15 +333,18 @@ def _make_mock_response(should_fail=False, fail_count={"count": 0}): @pytest.mark.asyncio -async def test_handle_async_request_total_timeout_triggers(): +async def test_handle_async_request_sock_read_timeout_triggers(): """ Ensure that LiteLLMAiohttpTransport raises httpx.TimeoutException - when the total timeout duration elapses. + when the sock_read timeout duration elapses (individual read operation timeout). + This is the correct behavior for stream_timeout - it should timeout on slow reads, + not on the total duration of the stream. """ import asyncio from aiohttp import web async def slow_handler(request): + # Sleep longer than the sock_read timeout await asyncio.sleep(0.3) return web.Response(text="ok") @@ -361,11 +364,12 @@ async def test_handle_async_request_total_timeout_triggers(): request = httpx.Request("GET", f"http://127.0.0.1:{port}/") + # Set a short sock_read timeout - this should trigger + # Note: total timeout is NOT set, allowing long-running streams request.extensions["timeout"] = { - "connect": 0.1, - "read": 0.1, - "pool": 0.1, - "total": 0.1, + "connect": 5.0, + "read": 0.1, # Short timeout for individual reads + "pool": 5.0, } try: @@ -376,6 +380,77 @@ async def test_handle_async_request_total_timeout_triggers(): await runner.cleanup() +@pytest.mark.asyncio +async def test_handle_async_request_streaming_does_not_timeout_on_total_duration(): + """ + Ensure that LiteLLMAiohttpTransport does NOT timeout on long-running + streaming responses as long as individual chunks arrive within the sock_read timeout. + This is the fix for issue #19184 - stream_timeout should only control the timeout + for individual chunks, not the total stream duration. + """ + import asyncio + from aiohttp import web + + async def streaming_handler(request): + # Simulate a streaming response that takes longer than a single timeout + # but each chunk arrives quickly + response = web.StreamResponse() + await response.prepare(request) + + # Send 5 chunks over 0.5 seconds total (0.1s between chunks) + for i in range(5): + await asyncio.sleep(0.05) # Less than sock_read timeout + await response.write(f"chunk{i}\n".encode()) + + await response.write_eof() + return response + + app = web.Application() + app.router.add_get("/stream", streaming_handler) + runner = web.AppRunner(app) + await runner.setup() + site = web.TCPSite(runner, "127.0.0.1", 0) + await site.start() + + port = site._server.sockets[0].getsockname()[1] + + def factory(): + return aiohttp.ClientSession() + + transport = LiteLLMAiohttpTransport(client=factory) # type: ignore + + request = httpx.Request("GET", f"http://127.0.0.1:{port}/stream") + + # Set sock_read timeout that's longer than individual chunk delays + # but shorter than total stream duration + # Total duration: ~0.25s, sock_read timeout: 0.15s per chunk + # This should NOT timeout because each chunk arrives within 0.15s + request.extensions["timeout"] = { + "connect": 5.0, + "read": 0.15, # Timeout for individual reads + "pool": 5.0, + # Note: total is NOT set - this is the fix! + } + + try: + # This should succeed without timing out + response = await transport.handle_async_request(request) + assert response.status_code == 200 + + # Read the streaming response + chunks = [] + async for chunk in response.aiter_bytes(): + chunks.append(chunk) + + # Verify we got all chunks + full_response = b"".join(chunks).decode() + assert "chunk0" in full_response + assert "chunk4" in full_response + finally: + await transport.aclose() + await runner.cleanup() + + def _make_mock_session(closed=False): """Helper to create a mock aiohttp session""" From f1bde3c5494fd501acae3bedb26a933976ffce5f Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 12:47:56 +0530 Subject: [PATCH 54/73] Add sanititzation for anthropic messages --- .../prompt_templates/factory.py | 220 ++++++++++ .../anthropic/test_message_sanitization.py | 380 ++++++++++++++++++ 2 files changed, 600 insertions(+) create mode 100644 tests/test_litellm/llms/anthropic/test_message_sanitization.py diff --git a/litellm/litellm_core_utils/prompt_templates/factory.py b/litellm/litellm_core_utils/prompt_templates/factory.py index 4320f756454..2f57ad9e813 100644 --- a/litellm/litellm_core_utils/prompt_templates/factory.py +++ b/litellm/litellm_core_utils/prompt_templates/factory.py @@ -1989,6 +1989,223 @@ def anthropic_process_openai_file_message( ) +def _sanitize_empty_text_content( + message: AllMessageValues, +) -> AllMessageValues: + """ + Case C: Sanitize empty text content + - Replace empty or whitespace-only text content with a placeholder message. + + Returns: + The message with sanitized content if needed, otherwise the original message + """ + if message.get("role") in ["user", "assistant"]: + content = message.get("content") + if isinstance(content, str): + if not content or not content.strip(): + message = dict(message) # Make a copy + message["content"] = "[System: Empty message content sanitised to satisfy protocol]" + verbose_logger.debug( + f"_sanitize_empty_text_content: Replaced empty text content in {message.get('role')} message" + ) + return message + + +def _add_missing_tool_results( + current_message: AllMessageValues, + messages: List[AllMessageValues], + current_index: int, +) -> List[AllMessageValues]: + """ + Case A: Missing tool_result for tool_use (orphaned tool calls) + - If an assistant message has tool_calls but no corresponding tool result follows, + add a dummy tool result message indicating the user did not provide the result. + + Returns: + A list containing the assistant message followed by any dummy tool results needed + """ + result_messages: List[AllMessageValues] = [] + tool_calls = current_message.get("tool_calls") + + if not tool_calls or len(tool_calls) == 0: + return [current_message] + + # Collect all tool_call_ids from this assistant message + expected_tool_call_ids = set() + for tool_call in tool_calls: + tool_call_id = None + if isinstance(tool_call, dict): + tool_call_id = tool_call.get("id") + else: + tool_call_id = getattr(tool_call, "id", None) + if tool_call_id: + expected_tool_call_ids.add(tool_call_id) + + found_tool_call_ids = set() + j = current_index + 1 + + while j < len(messages): + next_msg = messages[j] + next_role = next_msg.get("role") + + if next_role == "assistant": + break + + if next_role in ["tool", "function"]: + tool_call_id = next_msg.get("tool_call_id") + if tool_call_id: + found_tool_call_ids.add(tool_call_id) + + j += 1 + + # Find missing tool results + missing_tool_call_ids = expected_tool_call_ids - found_tool_call_ids + + if missing_tool_call_ids: + verbose_logger.debug( + f"_add_missing_tool_results: Found {len(missing_tool_call_ids)} orphaned tool calls. Adding dummy tool results." + ) + + result_messages.append(current_message) + + for tool_call_id in missing_tool_call_ids: + tool_name = "unknown_tool" + for tool_call in tool_calls: + tc_id = None + if isinstance(tool_call, dict): + tc_id = tool_call.get("id") + else: + tc_id = getattr(tool_call, "id", None) + + if tc_id == tool_call_id: + if isinstance(tool_call, dict): + function = tool_call.get("function", {}) + if isinstance(function, dict): + tool_name = function.get("name", "unknown_tool") + else: + tool_name = getattr(function, "name", "unknown_tool") + else: + function = getattr(tool_call, "function", None) + if function: + tool_name = getattr(function, "name", "unknown_tool") + break + + dummy_tool_result: ChatCompletionToolMessage = { + "role": "tool", + "tool_call_id": tool_call_id, + "content": f"[System: Tool execution skipped/interrupted by user. No result provided for tool '{tool_name}'.]", + } + result_messages.append(dummy_tool_result) + + return result_messages + + return [current_message] + + +def _is_orphaned_tool_result( + current_message: AllMessageValues, + sanitized_messages: List[AllMessageValues], +) -> bool: + """ + Case B: Orphaned tool_result (unexpected result) + - Check if a tool message references a tool_call_id that doesn't exist in the previous + assistant message. + + Returns: + True if this is an orphaned tool result that should be removed, False otherwise + """ + if current_message.get("role") not in ["tool", "function"]: + return False + + tool_call_id = current_message.get("tool_call_id") + + if not tool_call_id: + return False + + # Look back to find the most recent assistant message with tool_calls + found_matching_tool_call = False + + for j in range(len(sanitized_messages) - 1, -1, -1): + prev_msg = sanitized_messages[j] + if prev_msg.get("role") == "assistant": + tool_calls = prev_msg.get("tool_calls") + if tool_calls: + for tool_call in tool_calls: + tc_id = None + if isinstance(tool_call, dict): + tc_id = tool_call.get("id") + else: + tc_id = getattr(tool_call, "id", None) + + if tc_id == tool_call_id: + found_matching_tool_call = True + break + + break + + if not found_matching_tool_call: + verbose_logger.debug( + f"_is_orphaned_tool_result: Found orphaned tool result with tool_call_id={tool_call_id}" + ) + return True + + return False + + +def sanitize_messages_for_tool_calling( + messages: List[AllMessageValues], +) -> List[AllMessageValues]: + """ + Sanitize messages for tool calling to handle common issues when modify_params=True: + + Case A: Missing tool_result for tool_use (orphaned tool calls) + - If an assistant message has tool_calls but no corresponding tool result follows, + add a dummy tool result message indicating the user did not provide the result. + + Case B: Orphaned tool_result (unexpected result) + - If a tool message references a tool_call_id that doesn't exist in the previous + assistant message, remove that tool message. + + Case C: Empty text content + - Replace empty or whitespace-only text content with a placeholder message. + + This function operates on OpenAI format messages before they are converted to + provider-specific formats. + """ + if not litellm.modify_params: + return messages + + sanitized_messages: List[AllMessageValues] = [] + i = 0 + + while i < len(messages): + current_message = messages[i] + + # Case C: Sanitize empty text content + current_message = _sanitize_empty_text_content(current_message) + + # Case A: Check if assistant message has tool_calls without following tool results + if current_message.get("role") == "assistant": + result_messages = _add_missing_tool_results(current_message, messages, i) + + # If dummy tool results were added, extend sanitized_messages and continue + if len(result_messages) > 1: + sanitized_messages.extend(result_messages) + i += 1 + continue + + # Case B: Check for orphaned tool results + if _is_orphaned_tool_result(current_message, sanitized_messages): + i += 1 + continue # Skip this orphaned tool result + + # Add the message to sanitized list + sanitized_messages.append(current_message) + i += 1 + + return sanitized_messages + + def anthropic_messages_pt( # noqa: PLR0915 messages: List[AllMessageValues], model: str, @@ -2008,6 +2225,9 @@ def anthropic_messages_pt( # noqa: PLR0915 5. System messages are a separate param to the Messages API 6. Ensure we only accept role, content. (message.name is not supported) """ + # Sanitize messages for tool calling issues when modify_params=True + messages = sanitize_messages_for_tool_calling(messages) + # add role=tool support to allow function call result/error submission user_message_types = {"user", "tool", "function"} # reformat messages to ensure user/assistant are alternating, if there's either 2 consecutive 'user' messages or 2 consecutive 'assistant' message, merge them. diff --git a/tests/test_litellm/llms/anthropic/test_message_sanitization.py b/tests/test_litellm/llms/anthropic/test_message_sanitization.py new file mode 100644 index 00000000000..489ef527b48 --- /dev/null +++ b/tests/test_litellm/llms/anthropic/test_message_sanitization.py @@ -0,0 +1,380 @@ +""" +Test message sanitization for Anthropic API when modify_params=True + +Tests three cases: +A. Missing tool_result for tool_use (orphaned tool calls) +B. Orphaned tool_result without matching tool_use +C. Empty text content +""" + +import pytest +import sys +import os + +# Add the parent directory to the path so we can import litellm +sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), "../../../.."))) + +import litellm +from litellm.litellm_core_utils.prompt_templates.factory import ( + sanitize_messages_for_tool_calling, + anthropic_messages_pt, +) + + +class TestMessageSanitization: + """Test message sanitization for tool calling scenarios""" + + def setup_method(self): + """Setup for each test""" + # Save original modify_params value + self.original_modify_params = litellm.modify_params + litellm.modify_params = True + + def teardown_method(self): + """Cleanup after each test""" + # Restore original modify_params value + litellm.modify_params = self.original_modify_params + + def test_case_a_orphaned_tool_call_single(self): + """ + Test Case A: Assistant message with tool_calls but no tool result + Should add a dummy tool result message + """ + messages = [ + { + "role": "user", + "content": "What is the weather in Nashik?" + }, + { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "toolu_01Kus2cC3ydjBW7UK4GJqBP4", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Nashik, India"}' + } + } + ] + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + # Should have 3 messages: user, assistant, and dummy tool result + assert len(sanitized) == 3 + assert sanitized[0]["role"] == "user" + assert sanitized[1]["role"] == "assistant" + assert sanitized[2]["role"] == "tool" + assert sanitized[2]["tool_call_id"] == "toolu_01Kus2cC3ydjBW7UK4GJqBP4" + assert "skipped" in sanitized[2]["content"].lower() or "interrupted" in sanitized[2]["content"].lower() + assert "get_weather" in sanitized[2]["content"] + + def test_case_a_orphaned_tool_call_multiple(self): + """ + Test Case A: Assistant message with multiple tool_calls, some missing results + """ + messages = [ + { + "role": "user", + "content": "Get weather for Nashik and Mumbai" + }, + { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "call_1", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Nashik"}' + } + }, + { + "id": "call_2", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Mumbai"}' + } + } + ] + }, + { + "role": "tool", + "tool_call_id": "call_1", + "content": "Weather in Nashik: 25°C" + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + # Should have 4 messages: user, assistant, tool result for call_1, dummy for call_2 + assert len(sanitized) == 4 + assert sanitized[0]["role"] == "user" + assert sanitized[1]["role"] == "assistant" + assert sanitized[2]["tool_call_id"] == "call_2" # Dummy added first + assert sanitized[3]["tool_call_id"] == "call_1" # Original tool result + + def test_case_b_orphaned_tool_result(self): + """ + Test Case B: Tool result without matching tool_call in previous assistant message + Should remove the orphaned tool result + """ + messages = [ + { + "role": "user", + "content": "Hello" + }, + { + "role": "assistant", + "content": "Hi there!" + }, + { + "role": "tool", + "tool_call_id": "nonexistent_id", + "content": "Some result" + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + # Should have only 2 messages, orphaned tool result removed + assert len(sanitized) == 2 + assert sanitized[0]["role"] == "user" + assert sanitized[1]["role"] == "assistant" + + def test_case_b_valid_tool_result_preserved(self): + """ + Test Case B: Valid tool result with matching tool_call should be preserved + """ + messages = [ + { + "role": "user", + "content": "What's the weather?" + }, + { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "call_123", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Boston"}' + } + } + ] + }, + { + "role": "tool", + "tool_call_id": "call_123", + "content": "Weather: 20°C" + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + # All messages should be preserved + assert len(sanitized) == 3 + assert sanitized[2]["role"] == "tool" + assert sanitized[2]["tool_call_id"] == "call_123" + + def test_case_c_empty_text_content_user(self): + """ + Test Case C: Empty text content in user message + Should replace with placeholder + """ + messages = [ + { + "role": "user", + "content": "" + }, + { + "role": "assistant", + "content": "Hello!" + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + assert len(sanitized) == 2 + assert sanitized[0]["role"] == "user" + assert sanitized[0]["content"] == "[System: Empty message content sanitised to satisfy protocol]" + + def test_case_c_whitespace_only_content(self): + """ + Test Case C: Whitespace-only content + Should replace with placeholder + """ + messages = [ + { + "role": "user", + "content": " \n \t " + }, + { + "role": "assistant", + "content": " " + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + assert len(sanitized) == 2 + assert sanitized[0]["content"] == "[System: Empty message content sanitised to satisfy protocol]" + assert sanitized[1]["content"] == "[System: Empty message content sanitised to satisfy protocol]" + + def test_case_c_valid_content_preserved(self): + """ + Test Case C: Valid non-empty content should be preserved + """ + messages = [ + { + "role": "user", + "content": "Hello" + }, + { + "role": "assistant", + "content": "Hi there!" + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + assert len(sanitized) == 2 + assert sanitized[0]["content"] == "Hello" + assert sanitized[1]["content"] == "Hi there!" + + def test_combined_cases(self): + """ + Test combination of multiple cases + """ + messages = [ + { + "role": "user", + "content": "Get weather" + }, + { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "call_1", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "NYC"}' + } + } + ] + }, + # Missing tool result for call_1 + { + "role": "user", + "content": "" # Empty content + }, + { + "role": "assistant", + "content": "Response" + }, + { + "role": "tool", + "tool_call_id": "orphaned_id", # Orphaned tool result + "content": "Some data" + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + # Should have: user, assistant, dummy tool result, user (sanitized), assistant + # Orphaned tool result should be removed + assert len(sanitized) == 5 + assert sanitized[0]["role"] == "user" + assert sanitized[1]["role"] == "assistant" + assert sanitized[2]["role"] == "tool" + assert sanitized[2]["tool_call_id"] == "call_1" # Dummy added + assert sanitized[3]["role"] == "user" + assert sanitized[3]["content"] == "[System: Empty message content sanitised to satisfy protocol]" + assert sanitized[4]["role"] == "assistant" + + def test_modify_params_false_no_sanitization(self): + """ + Test that sanitization is skipped when modify_params=False + """ + litellm.modify_params = False + + messages = [ + { + "role": "user", + "content": "" + }, + { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "call_1", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{}' + } + } + ] + } + ] + + sanitized = sanitize_messages_for_tool_calling(messages) + + # Messages should be unchanged + assert len(sanitized) == 2 + assert sanitized[0]["content"] == "" + assert len(sanitized[1].get("tool_calls", [])) == 1 + + def test_anthropic_messages_pt_integration(self): + """ + Test that sanitization is integrated into anthropic_messages_pt + """ + litellm.modify_params = True + + messages = [ + { + "role": "user", + "content": "What is the weather in Nashik?" + }, + { + "role": "assistant", + "content": None, + "tool_calls": [ + { + "id": "toolu_01Kus2cC3ydjBW7UK4GJqBP4", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Nashik, India"}' + } + } + ] + } + ] + + # This should not raise an error and should add dummy tool result + result = anthropic_messages_pt( + messages=messages, + model="claude-sonnet-4-5", + llm_provider="anthropic" + ) + + # Should have at least 2 messages (user and assistant) + # The tool result will be merged into user content + assert len(result) >= 2 + assert result[0]["role"] == "user" + assert result[1]["role"] == "assistant" + + +if __name__ == "__main__": + pytest.main([__file__, "-v"]) From 8be3712e829ca28f739ea68780fbeb8e64d2f52c Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 12:52:13 +0530 Subject: [PATCH 55/73] Add docs for message sanitisation --- .../docs/completion/message_sanitization.md | 468 ++++++++++++++++++ docs/my-website/sidebars.js | 1 + 2 files changed, 469 insertions(+) create mode 100644 docs/my-website/docs/completion/message_sanitization.md diff --git a/docs/my-website/docs/completion/message_sanitization.md b/docs/my-website/docs/completion/message_sanitization.md new file mode 100644 index 00000000000..0a1f766e2fd --- /dev/null +++ b/docs/my-website/docs/completion/message_sanitization.md @@ -0,0 +1,468 @@ +import Tabs from '@theme/Tabs'; +import TabItem from '@theme/TabItem'; + +# Message Sanitization for Tool Calling for anthropic models + +**Automatically fix common message formatting issues when using tool calling with `modify_params=True`** + +LiteLLM can automatically sanitize messages to handle common issues that occur during tool calling workflows, especially when using OpenAI-compatible clients with providers that have strict message format requirements (like Anthropic Claude). + +## Overview + +When `litellm.modify_params = True` is enabled, LiteLLM automatically sanitizes messages to fix three common issues: + +1. **Orphaned Tool Calls** - Assistant messages with tool_calls but missing tool results +2. **Orphaned Tool Results** - Tool messages that reference non-existent tool_call_ids +3. **Empty Message Content** - Messages with empty or whitespace-only text content + +This ensures your tool calling workflows work seamlessly across different LLM providers without manual message validation. + +## Why Message Sanitization? + +Different LLM providers have varying requirements for message formats, especially during tool calling: + +- **Anthropic Claude** requires every tool_call to have a corresponding tool result +- Some providers reject messages with empty content +- OpenAI-compatible clients may not always maintain perfect message consistency + +Without sanitization, these issues cause API errors that interrupt your workflows. With `modify_params=True`, LiteLLM handles these edge cases automatically. + +## Quick Start + + + + +```python +import litellm + +# Enable automatic message sanitization +litellm.modify_params = True + +# This will work even if messages have formatting issues +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=[ + {"role": "user", "content": "What's the weather in Boston?"}, + { + "role": "assistant", + "tool_calls": [ + { + "id": "call_123", + "type": "function", + "function": {"name": "get_weather", "arguments": '{"city": "Boston"}'} + } + ] + # Missing tool result - LiteLLM will add a dummy result automatically + }, + {"role": "user", "content": "Thanks!"} + ], + tools=[{ + "type": "function", + "function": { + "name": "get_weather", + "description": "Get weather for a city", + "parameters": { + "type": "object", + "properties": {"city": {"type": "string"}}, + "required": ["city"] + } + } + }] +) +``` + + + + +```yaml +litellm_settings: + modify_params: true # Enable automatic message sanitization + +model_list: + - model_name: claude-3-5-sonnet + litellm_params: + model: anthropic/claude-3-5-sonnet-20241022 +``` + + + + +## Sanitization Cases + +### Case A: Orphaned Tool Calls (Missing Tool Results) + +**Problem:** An assistant message contains `tool_calls`, but no corresponding tool result messages follow. + +**Solution:** LiteLLM automatically adds dummy tool result messages for any missing tool results. + +**Example:** + +```python +import litellm +litellm.modify_params = True + +# Messages with orphaned tool calls +messages = [ + {"role": "user", "content": "Search for Python tutorials"}, + { + "role": "assistant", + "tool_calls": [ + { + "id": "call_abc123", + "type": "function", + "function": {"name": "web_search", "arguments": '{"query": "Python tutorials"}'} + } + ] + }, + # Missing tool result here! + {"role": "user", "content": "What about JavaScript?"} +] + +# LiteLLM automatically adds: +# { +# "role": "tool", +# "tool_call_id": "call_abc123", +# "content": "[System: Tool execution skipped/interrupted by user. No result provided for tool 'web_search'.]" +# } + +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=messages, + tools=[...] +) +``` + +**When this happens:** +- User interrupts tool execution +- Client loses tool results due to network issues +- Conversation flow changes before tool completes +- Multi-turn conversations where tools are optional + +### Case B: Orphaned Tool Results (Invalid tool_call_id) + +**Problem:** A tool message references a `tool_call_id` that doesn't exist in any previous assistant message. + +**Solution:** LiteLLM automatically removes these orphaned tool result messages. + +**Example:** + +```python +import litellm +litellm.modify_params = True + +# Messages with orphaned tool result +messages = [ + {"role": "user", "content": "Hello"}, + {"role": "assistant", "content": "Hi! How can I help?"}, + { + "role": "tool", + "tool_call_id": "call_nonexistent", # This tool_call_id doesn't exist! + "content": "Some result" + } +] + +# LiteLLM automatically removes the orphaned tool message + +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=messages +) +``` + +**When this happens:** +- Message history is manually edited +- Tool results are duplicated or mismatched +- Conversation state is restored incorrectly +- Messages are merged from different conversations + +### Case C: Empty Message Content + +**Problem:** User or assistant messages have empty or whitespace-only content. + +**Solution:** LiteLLM replaces empty content with a system placeholder message. + +**Example:** + +```python +import litellm +litellm.modify_params = True + +# Messages with empty content +messages = [ + {"role": "user", "content": ""}, # Empty content + {"role": "assistant", "content": " "}, # Whitespace only +] + +# LiteLLM automatically replaces with: +# {"role": "user", "content": "[System: Empty message content sanitised to satisfy protocol]"} +# {"role": "assistant", "content": "[System: Empty message content sanitised to satisfy protocol]"} + +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=messages +) +``` + +**When this happens:** +- UI sends empty messages +- Content is stripped during preprocessing +- Placeholder messages in conversation history +- Edge cases in message construction + +## Configuration + +### Enable Globally + + + + +```python +import litellm + +# Enable for all completion calls +litellm.modify_params = True +``` + + + + +```yaml +litellm_settings: + modify_params: true +``` + + + + +```bash +export LITELLM_MODIFY_PARAMS=True +``` + + + + +### Enable Per-Request + +```python +import litellm + +# Enable only for specific requests +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=messages, + modify_params=True # Override global setting +) +``` + +## Supported Providers + +Message sanitization works with all LLM providers that support tool calling: + +- ✅ Anthropic (Claude) +- ✅ OpenAI (GPT-4, GPT-3.5) +- ✅ AWS Bedrock (Claude, Titan) +- ✅ Google Vertex AI (Claude, Gemini) +- ✅ Azure OpenAI +- ✅ And all other providers with tool calling support + +## Implementation Details + +### How It Works + +The message sanitization process runs **before** messages are converted to provider-specific formats: + +1. **Input:** OpenAI-format messages with potential issues +2. **Sanitization:** Three helper functions process the messages: + - `_sanitize_empty_text_content()` - Fixes empty content + - `_add_missing_tool_results()` - Adds dummy tool results + - `_is_orphaned_tool_result()` - Identifies orphaned results +3. **Output:** Clean, provider-compatible messages + +### Code Reference + +The sanitization logic is implemented in: +- `litellm/litellm_core_utils/prompt_templates/factory.py` +- Function: `sanitize_messages_for_tool_calling()` + +### Logging + +When sanitization occurs, LiteLLM logs debug messages: + +```python +import litellm +litellm.set_verbose = True # Enable debug logging + +# You'll see logs like: +# "_add_missing_tool_results: Found 1 orphaned tool calls. Adding dummy tool results." +# "_is_orphaned_tool_result: Found orphaned tool result with tool_call_id=call_123" +# "_sanitize_empty_text_content: Replaced empty text content in user message" +``` + +## Best Practices + +### 1. Enable for Production Workflows + +```python +# Recommended for production +litellm.modify_params = True + +# Ensures robust handling of edge cases +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=messages, + tools=tools +) +``` + +### 2. Preserve Tool Results When Possible + +While sanitization handles missing tool results, it's better to provide actual results: + +```python +# Good: Provide actual tool results +messages = [ + {"role": "user", "content": "Search for Python"}, + {"role": "assistant", "tool_calls": [...]}, + {"role": "tool", "tool_call_id": "call_123", "content": "Actual search results"} +] + +# Fallback: Sanitization adds dummy result if missing +messages = [ + {"role": "user", "content": "Search for Python"}, + {"role": "assistant", "tool_calls": [...]}, + # Missing tool result - sanitization adds dummy +] +``` + +### 3. Monitor Sanitization Events + +Use logging to track when sanitization occurs: + +```python +import litellm +import logging + +# Enable debug logging +litellm.set_verbose = True +logging.basicConfig(level=logging.DEBUG) + +# Track sanitization events in your application +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=messages +) +``` + +### 4. Test Edge Cases + +Ensure your application handles sanitized messages correctly: + +```python +import litellm +litellm.modify_params = True + +# Test orphaned tool calls +test_messages = [ + {"role": "user", "content": "Test"}, + {"role": "assistant", "tool_calls": [{"id": "call_1", "type": "function", "function": {"name": "test", "arguments": "{}"}}]}, + {"role": "user", "content": "Continue"} # No tool result +] + +response = litellm.completion( + model="anthropic/claude-3-5-sonnet-20241022", + messages=test_messages, + tools=[...] +) + +# Verify the response handles the dummy tool result appropriately +``` + +## Related Features + +- **[Drop Params](./drop_params.md)** - Drop unsupported parameters for specific providers +- **[Message Trimming](./message_trimming.md)** - Trim messages to fit token limits +- **[Function Calling](./function_call.md)** - Complete guide to tool/function calling +- **[Reasoning Content](../reasoning_content.md)** - Extended thinking with tool calling + +## Troubleshooting + +### Sanitization Not Working + +**Issue:** Messages still cause errors despite `modify_params=True` + +**Solution:** +1. Verify `modify_params` is enabled: + ```python + import litellm + print(litellm.modify_params) # Should be True + ``` + +2. Check if the issue is provider-specific: + ```python + litellm.set_verbose = True # Enable debug logging + ``` + +3. Ensure you're using a recent version of LiteLLM: + ```bash + pip install --upgrade litellm + ``` + +### Unexpected Dummy Tool Results + +**Issue:** Dummy tool results appear when you expect actual results + +**Cause:** Tool result messages are missing or have incorrect `tool_call_id` + +**Solution:** +1. Verify tool result messages have correct `tool_call_id`: + ```python + # Correct + {"role": "tool", "tool_call_id": "call_123", "content": "result"} + + # Incorrect - will be treated as orphaned + {"role": "tool", "tool_call_id": "wrong_id", "content": "result"} + ``` + +2. Ensure tool results immediately follow assistant messages with tool_calls + +### Performance Impact + +**Issue:** Concerned about performance overhead + +**Details:** Message sanitization has minimal performance impact: +- Runs in O(n) time where n = number of messages +- Only processes messages when `modify_params=True` +- Typically adds < 1ms to request processing time + +## FAQ + +**Q: Does sanitization modify my original messages?** + +A: No, sanitization creates a new list of messages. Your original messages remain unchanged. + +**Q: Can I disable specific sanitization cases?** + +A: Currently, all three cases are handled together when `modify_params=True`. To disable sanitization entirely, set `modify_params=False`. + +**Q: What happens to the dummy tool results?** + +A: Dummy tool results are sent to the LLM provider along with other messages. The model sees them as regular tool results with informative error messages. + +**Q: Does this work with streaming?** + +A: Yes, message sanitization works with both streaming and non-streaming requests. + +**Q: Is this related to `drop_params`?** + +A: No, they're separate features: +- `modify_params` - Modifies/fixes message content and structure +- `drop_params` - Removes unsupported API parameters + +Both can be enabled simultaneously. + +## See Also + +- [Reasoning Content with Tool Calling](../reasoning_content.md) +- [Function Calling Guide](./function_call.md) +- [Bedrock Provider Documentation](../providers/bedrock.md) +- [Anthropic Provider Documentation](../providers/anthropic.md) diff --git a/docs/my-website/sidebars.js b/docs/my-website/sidebars.js index 619bbed6808..ecfda1fd9b8 100644 --- a/docs/my-website/sidebars.js +++ b/docs/my-website/sidebars.js @@ -822,6 +822,7 @@ const sidebars = { "completion/knowledgebase", "guides/code_interpreter", "completion/message_trimming", + "completion/message_sanitization", "completion/model_alias", "completion/mock_requests", "completion/predict_outputs", From d721db2295f1f7cf7f763f27b131b9e38b19efa6 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 13:05:44 +0530 Subject: [PATCH 56/73] Fix : revert get_combined_tool_content --- .../streaming_chunk_builder_utils.py | 91 ++++++++++++++----- 1 file changed, 67 insertions(+), 24 deletions(-) diff --git a/litellm/litellm_core_utils/streaming_chunk_builder_utils.py b/litellm/litellm_core_utils/streaming_chunk_builder_utils.py index 47f5bdf73c0..02767a93121 100644 --- a/litellm/litellm_core_utils/streaming_chunk_builder_utils.py +++ b/litellm/litellm_core_utils/streaming_chunk_builder_utils.py @@ -132,7 +132,7 @@ class ChunkProcessor: ) return response - def get_combined_tool_content( # noqa: PLR0915 + def get_combined_tool_content( #noqa: PLR0915 self, tool_call_chunks: List[Dict[str, Any]] ) -> List[ChatCompletionMessageToolCall]: tool_calls_list: List[ChatCompletionMessageToolCall] = [] @@ -147,10 +147,26 @@ class ChunkProcessor: tool_calls = delta.get("tool_calls", []) for tool_call in tool_calls: - if not tool_call or not hasattr(tool_call, "function"): + # Handle both dict and object formats + if not tool_call: + continue + + # Check if tool_call has function (either as attribute or dict key) + has_function = False + if isinstance(tool_call, dict): + has_function = "function" in tool_call and tool_call["function"] is not None + else: + has_function = hasattr(tool_call, "function") and tool_call.function is not None + + if not has_function: continue - index = getattr(tool_call, "index", 0) + # Get index (handle both dict and object) + if isinstance(tool_call, dict): + index = tool_call.get("index", 0) + else: + index = getattr(tool_call, "index", 0) + if index not in tool_call_map: tool_call_map[index] = { "id": None, @@ -160,30 +176,56 @@ class ChunkProcessor: "provider_specific_fields": None, } - if hasattr(tool_call, "id") and tool_call.id: - tool_call_map[index]["id"] = tool_call.id - if hasattr(tool_call, "type") and tool_call.type: - tool_call_map[index]["type"] = tool_call.type - if hasattr(tool_call, "function"): - if ( - hasattr(tool_call.function, "name") - and tool_call.function.name - ): - tool_call_map[index]["name"] = tool_call.function.name - if ( - hasattr(tool_call.function, "arguments") - and tool_call.function.arguments - ): - tool_call_map[index]["arguments"].append( - tool_call.function.arguments - ) + # Extract id, type, and function data (handle both dict and object) + if isinstance(tool_call, dict): + if tool_call.get("id"): + tool_call_map[index]["id"] = tool_call["id"] + if tool_call.get("type"): + tool_call_map[index]["type"] = tool_call["type"] + + function = tool_call.get("function", {}) + if isinstance(function, dict): + if function.get("name"): + tool_call_map[index]["name"] = function["name"] + if function.get("arguments"): + tool_call_map[index]["arguments"].append(function["arguments"]) + else: + # function is an object + if hasattr(function, "name") and function.name: + tool_call_map[index]["name"] = function.name + if hasattr(function, "arguments") and function.arguments: + tool_call_map[index]["arguments"].append(function.arguments) + else: + # tool_call is an object + if hasattr(tool_call, "id") and tool_call.id: + tool_call_map[index]["id"] = tool_call.id + if hasattr(tool_call, "type") and tool_call.type: + tool_call_map[index]["type"] = tool_call.type + if hasattr(tool_call, "function"): + if ( + hasattr(tool_call.function, "name") + and tool_call.function.name + ): + tool_call_map[index]["name"] = tool_call.function.name + if ( + hasattr(tool_call.function, "arguments") + and tool_call.function.arguments + ): + tool_call_map[index]["arguments"].append( + tool_call.function.arguments + ) # Preserve provider_specific_fields from streaming chunks provider_fields = None - if hasattr(tool_call, "provider_specific_fields") and tool_call.provider_specific_fields: - provider_fields = tool_call.provider_specific_fields - elif hasattr(tool_call, "function") and hasattr(tool_call.function, "provider_specific_fields") and tool_call.function.provider_specific_fields: - provider_fields = tool_call.function.provider_specific_fields + if isinstance(tool_call, dict): + provider_fields = tool_call.get("provider_specific_fields") + if not provider_fields and isinstance(tool_call.get("function"), dict): + provider_fields = tool_call["function"].get("provider_specific_fields") + else: + if hasattr(tool_call, "provider_specific_fields") and tool_call.provider_specific_fields: + provider_fields = tool_call.provider_specific_fields + elif hasattr(tool_call, "function") and hasattr(tool_call.function, "provider_specific_fields") and tool_call.function.provider_specific_fields: + provider_fields = tool_call.function.provider_specific_fields if provider_fields: # Merge provider_specific_fields if multiple chunks have them @@ -222,6 +264,7 @@ class ChunkProcessor: return tool_calls_list + def get_combined_function_call_content( self, function_call_chunks: List[Dict[str, Any]] ) -> FunctionCall: From 3daab290f609661cf8947455efdf458b10bfa4ec Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 13:08:16 +0530 Subject: [PATCH 57/73] Fix : revert get_combined_tool_content --- litellm/litellm_core_utils/streaming_chunk_builder_utils.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/litellm/litellm_core_utils/streaming_chunk_builder_utils.py b/litellm/litellm_core_utils/streaming_chunk_builder_utils.py index 02767a93121..53252df0a28 100644 --- a/litellm/litellm_core_utils/streaming_chunk_builder_utils.py +++ b/litellm/litellm_core_utils/streaming_chunk_builder_utils.py @@ -132,7 +132,7 @@ class ChunkProcessor: ) return response - def get_combined_tool_content( #noqa: PLR0915 + def get_combined_tool_content( # noqa: PLR0915 self, tool_call_chunks: List[Dict[str, Any]] ) -> List[ChatCompletionMessageToolCall]: tool_calls_list: List[ChatCompletionMessageToolCall] = [] From bf99cea82fd8bb437188a25abf7c528d368c56d2 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 14:33:25 +0530 Subject: [PATCH 58/73] Fix malformed tool call tranform --- .../prompt_templates/factory.py | 20 ++- .../bedrock/chat/converse_transformation.py | 9 +- litellm/types/llms/bedrock.py | 2 +- .../test_bedrock_completion.py | 154 ++++++++++++++++++ 4 files changed, 175 insertions(+), 10 deletions(-) diff --git a/litellm/litellm_core_utils/prompt_templates/factory.py b/litellm/litellm_core_utils/prompt_templates/factory.py index 4320f756454..01bf18d79b2 100644 --- a/litellm/litellm_core_utils/prompt_templates/factory.py +++ b/litellm/litellm_core_utils/prompt_templates/factory.py @@ -3233,17 +3233,21 @@ def _convert_to_bedrock_tool_call_invoke( id = tool["id"] name = tool["function"].get("name", "") arguments = tool["function"].get("arguments", "") - arguments_dict = json.loads(arguments) if arguments else {} - # Ensure arguments_dict is always a dict (Bedrock requires toolUse.input to be an object) - # When some providers return arguments: '""' (JSON-encoded empty string), json.loads returns "" - if not isinstance(arguments_dict, dict): - arguments_dict = {} if not arguments or not arguments.strip(): - arguments_dict = {} + arguments_input = {} else: - arguments_dict = json.loads(arguments) + # Try to parse the arguments JSON + try: + arguments_input = json.loads(arguments) + except json.JSONDecodeError as e: + verbose_logger.warning( + f"Malformed JSON in tool call arguments for tool '{name}': {str(e)}. " + f"Storing as raw string to allow conversation to continue." + ) + arguments_input = arguments + bedrock_tool = BedrockToolUseBlock( - input=arguments_dict, name=name, toolUseId=id + input=arguments_input, name=name, toolUseId=id ) bedrock_content_block = BedrockContentBlock(toolUse=bedrock_tool) _parts_list.append(bedrock_content_block) diff --git a/litellm/llms/bedrock/chat/converse_transformation.py b/litellm/llms/bedrock/chat/converse_transformation.py index 59590e464fc..9bc1e8c85e2 100644 --- a/litellm/llms/bedrock/chat/converse_transformation.py +++ b/litellm/llms/bedrock/chat/converse_transformation.py @@ -1395,9 +1395,16 @@ class AmazonConverseConfig(BaseConfig): response_tool_name = get_bedrock_tool_name( response_tool_name=_response_tool_name ) + tool_input = content["toolUse"]["input"] + if isinstance(tool_input, str): + arguments_str = tool_input + else: + # Otherwise, serialize it to JSON + arguments_str = json.dumps(tool_input) + _function_chunk = ChatCompletionToolCallFunctionChunk( name=response_tool_name, - arguments=json.dumps(content["toolUse"]["input"]), + arguments=arguments_str, ) _tool_response_chunk = ChatCompletionToolCallChunk( diff --git a/litellm/types/llms/bedrock.py b/litellm/types/llms/bedrock.py index ef2f1ba4d5e..e0858898eae 100644 --- a/litellm/types/llms/bedrock.py +++ b/litellm/types/llms/bedrock.py @@ -62,7 +62,7 @@ class ToolResultBlock(TypedDict, total=False): class ToolUseBlock(TypedDict): - input: dict + input: Any # Per boto3 spec: document type can be dict, list, int, float, str, bool, or None name: str toolUseId: str diff --git a/tests/llm_translation/test_bedrock_completion.py b/tests/llm_translation/test_bedrock_completion.py index 7c0db41d13a..f08060214c5 100644 --- a/tests/llm_translation/test_bedrock_completion.py +++ b/tests/llm_translation/test_bedrock_completion.py @@ -3954,3 +3954,157 @@ def test_bedrock_openai_error_handling(): assert exc_info.value.status_code == 422 print("✓ Error handling works correctly") + + +def test_bedrock_malformed_tool_json_handling(): + """ + Test that Bedrock handles malformed JSON in tool call arguments gracefully. + + This test covers the issue where: + 1. LLM generates malformed JSON in tool call arguments + 2. Subsequent requests with conversation history should not crash + 3. The toolUse.input field should handle any JSON value type per boto3 spec + + Related issue: https://github.com/BerriAI/litellm/issues/[issue_number] + """ + from litellm.litellm_core_utils.prompt_templates.factory import ( + _convert_to_bedrock_tool_call_invoke, + ) + from litellm.llms.bedrock.chat.converse_transformation import AmazonConverseConfig + from litellm.types.llms.bedrock import ContentBlock + + # Test 1: Malformed JSON in tool call arguments + malformed_tool_calls = [ + { + "id": "call_123", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "Paris", "invalid_json', # Malformed JSON + }, + } + ] + + # Should not raise an exception, but store as raw string + result = _convert_to_bedrock_tool_call_invoke(malformed_tool_calls) + assert len(result) == 1 + assert result[0]["toolUse"]["name"] == "get_weather" + # The malformed JSON should be stored as a string + assert isinstance(result[0]["toolUse"]["input"], str) + assert result[0]["toolUse"]["input"] == '{"location": "Paris", "invalid_json' + print("✓ Malformed JSON stored as raw string") + + # Test 2: Valid JSON should still work normally + valid_tool_calls = [ + { + "id": "call_456", + "type": "function", + "function": { + "name": "get_weather", + "arguments": '{"location": "London"}', + }, + } + ] + + result = _convert_to_bedrock_tool_call_invoke(valid_tool_calls) + assert len(result) == 1 + assert result[0]["toolUse"]["name"] == "get_weather" + assert isinstance(result[0]["toolUse"]["input"], dict) + assert result[0]["toolUse"]["input"] == {"location": "London"} + print("✓ Valid JSON parsed correctly") + + # Test 3: Empty arguments should create empty dict + empty_tool_calls = [ + { + "id": "call_789", + "type": "function", + "function": { + "name": "no_args_function", + "arguments": "", + }, + } + ] + + result = _convert_to_bedrock_tool_call_invoke(empty_tool_calls) + assert len(result) == 1 + assert result[0]["toolUse"]["input"] == {} + print("✓ Empty arguments handled correctly") + + # Test 4: Bedrock to OpenAI conversion handles string input + converse_config = AmazonConverseConfig() + content_blocks = [ + ContentBlock( + toolUse={ + "name": "get_weather", + "toolUseId": "call_123", + "input": '{"location": "Paris", "invalid_json', # String input (malformed) + } + ) + ] + + content_str, tools, reasoning = converse_config._translate_message_content( + content_blocks + ) + assert len(tools) == 1 + assert tools[0]["function"]["name"] == "get_weather" + # Should return the string as-is + assert tools[0]["function"]["arguments"] == '{"location": "Paris", "invalid_json' + print("✓ Bedrock to OpenAI conversion handles string input") + + # Test 5: Bedrock to OpenAI conversion handles dict input + content_blocks_dict = [ + ContentBlock( + toolUse={ + "name": "get_weather", + "toolUseId": "call_456", + "input": {"location": "London"}, # Dict input (normal case) + } + ) + ] + + content_str, tools, reasoning = converse_config._translate_message_content( + content_blocks_dict + ) + assert len(tools) == 1 + assert tools[0]["function"]["name"] == "get_weather" + # Should serialize dict to JSON string + assert tools[0]["function"]["arguments"] == '{"location": "London"}' + print("✓ Bedrock to OpenAI conversion handles dict input") + + # Test 6: Round-trip conversion with malformed JSON + # Test that we can convert OpenAI -> Bedrock -> OpenAI with malformed JSON + malformed_tool_calls_roundtrip = [ + { + "id": "call_999", + "type": "function", + "function": { + "name": "test_function", + "arguments": '{"key": "value", "broken', # Malformed + }, + } + ] + + # Step 1: OpenAI to Bedrock (should store as string) + bedrock_blocks = _convert_to_bedrock_tool_call_invoke(malformed_tool_calls_roundtrip) + assert isinstance(bedrock_blocks[0]["toolUse"]["input"], str) + + # Step 2: Bedrock back to OpenAI (should preserve the string) + content_blocks_roundtrip = [ + ContentBlock( + toolUse={ + "name": bedrock_blocks[0]["toolUse"]["name"], + "toolUseId": bedrock_blocks[0]["toolUse"]["toolUseId"], + "input": bedrock_blocks[0]["toolUse"]["input"], + } + ) + ] + + content_str, tools_roundtrip, reasoning = converse_config._translate_message_content( + content_blocks_roundtrip + ) + + # Should preserve the malformed JSON string through the round trip + assert tools_roundtrip[0]["function"]["arguments"] == '{"key": "value", "broken' + print("✓ Round-trip conversion preserves malformed JSON") + + print("✓ All malformed JSON handling tests passed") From 4d45574fc567aa9a46a4fecf475bd59148ce74ea Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Tue, 6 Jan 2026 11:55:41 +0530 Subject: [PATCH 59/73] fix Updated all 27 occurrences of mode: image_edit to mode: image_edits --- ...odel_prices_and_context_window_backup.json | 60 +++++++++++-------- model_prices_and_context_window.json | 58 +++++++++--------- 2 files changed, 64 insertions(+), 54 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 85661def27c..e58db912cf4 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -25383,7 +25383,7 @@ }, "stability/inpaint": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25391,7 +25391,7 @@ }, "stability/outpaint": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.004, "supported_endpoints": [ "/v1/images/edits" @@ -25399,7 +25399,7 @@ }, "stability/erase": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25407,7 +25407,7 @@ }, "stability/search-and-replace": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25415,7 +25415,7 @@ }, "stability/search-and-recolor": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25423,7 +25423,7 @@ }, "stability/remove-background": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25431,7 +25431,7 @@ }, "stability/replace-background-and-relight": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25439,7 +25439,7 @@ }, "stability/sketch": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25447,7 +25447,7 @@ }, "stability/structure": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25455,7 +25455,7 @@ }, "stability/style": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25463,7 +25463,7 @@ }, "stability/style-transfer": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25471,7 +25471,7 @@ }, "stability/fast": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.002, "supported_endpoints": [ "/v1/images/edits" @@ -25479,7 +25479,7 @@ }, "stability/conservative": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.04, "supported_endpoints": [ "/v1/images/edits" @@ -25487,7 +25487,7 @@ }, "stability/creative": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.06, "supported_endpoints": [ "/v1/images/edits" @@ -25525,79 +25525,89 @@ "stability.stable-conservative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, +<<<<<<< HEAD "mode": "image_edit", "output_cost_per_image": 0.4 +======= + "mode": "image_edits", + "output_cost_per_image": 0.40 +>>>>>>> b712575d64 (fix Updated all 27 occurrences of mode: image_edit to mode: image_edits) }, "stability.stable-creative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, +<<<<<<< HEAD "mode": "image_edit", "output_cost_per_image": 0.6 +======= + "mode": "image_edits", + "output_cost_per_image": 0.60 +>>>>>>> b712575d64 (fix Updated all 27 occurrences of mode: image_edit to mode: image_edits) }, "stability.stable-fast-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.03 }, "stability.stable-outpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.06 }, "stability.stable-image-control-sketch-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-control-structure-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-erase-object-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-inpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-remove-background-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-search-recolor-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-search-replace-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-style-guide-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-style-transfer-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.08 }, "stability.stable-image-core-v1:1": { diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 85661def27c..52b41a464eb 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -25383,7 +25383,7 @@ }, "stability/inpaint": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25391,7 +25391,7 @@ }, "stability/outpaint": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.004, "supported_endpoints": [ "/v1/images/edits" @@ -25399,7 +25399,7 @@ }, "stability/erase": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25407,7 +25407,7 @@ }, "stability/search-and-replace": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25415,7 +25415,7 @@ }, "stability/search-and-recolor": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25423,7 +25423,7 @@ }, "stability/remove-background": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25431,7 +25431,7 @@ }, "stability/replace-background-and-relight": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25439,7 +25439,7 @@ }, "stability/sketch": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25447,7 +25447,7 @@ }, "stability/structure": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25455,7 +25455,7 @@ }, "stability/style": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25463,7 +25463,7 @@ }, "stability/style-transfer": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25471,7 +25471,7 @@ }, "stability/fast": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.002, "supported_endpoints": [ "/v1/images/edits" @@ -25479,7 +25479,7 @@ }, "stability/conservative": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.04, "supported_endpoints": [ "/v1/images/edits" @@ -25487,7 +25487,7 @@ }, "stability/creative": { "litellm_provider": "stability", - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.06, "supported_endpoints": [ "/v1/images/edits" @@ -25525,79 +25525,79 @@ "stability.stable-conservative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", - "output_cost_per_image": 0.4 + "mode": "image_edits", + "output_cost_per_image": 0.40 }, "stability.stable-creative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", - "output_cost_per_image": 0.6 + "mode": "image_edits", + "output_cost_per_image": 0.60 }, "stability.stable-fast-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.03 }, "stability.stable-outpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.06 }, "stability.stable-image-control-sketch-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-control-structure-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-erase-object-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-inpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-remove-background-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-search-recolor-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-search-replace-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-image-style-guide-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.07 }, "stability.stable-style-transfer-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edit", + "mode": "image_edits", "output_cost_per_image": 0.08 }, "stability.stable-image-core-v1:1": { From 82fe942fd9f89d2fdfd2f814068b6513c040730b Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Tue, 6 Jan 2026 12:01:27 +0530 Subject: [PATCH 60/73] fix: image_edits request handling fails for Stability models --- litellm/llms/bedrock/image_edit/handler.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/litellm/llms/bedrock/image_edit/handler.py b/litellm/llms/bedrock/image_edit/handler.py index b4b6c8d7622..0f1dcff6294 100644 --- a/litellm/llms/bedrock/image_edit/handler.py +++ b/litellm/llms/bedrock/image_edit/handler.py @@ -261,7 +261,7 @@ class BedrockImageEdit(BaseAWSLLM): """ config_class = self.get_config_class(model=model) config_instance = config_class() - request_body = config_instance.transform_image_edit_request( + request_body, _ = config_instance.transform_image_edit_request( model=model, prompt=prompt, image=image[0] if image else None, From 3df0d45d58e80fe01b0b59a8b37f5a8cba416b9e Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Tue, 6 Jan 2026 12:14:40 +0530 Subject: [PATCH 61/73] fix documentation --- docs/my-website/docs/providers/stability.md | 1 - 1 file changed, 1 deletion(-) diff --git a/docs/my-website/docs/providers/stability.md b/docs/my-website/docs/providers/stability.md index 6b340267e69..62a8ab43cd8 100644 --- a/docs/my-website/docs/providers/stability.md +++ b/docs/my-website/docs/providers/stability.md @@ -416,7 +416,6 @@ response = image_edit( image=open("original_image.png", "rb"), mask=open("mask_image.png", "rb"), prompt="Add flowers in the masked area", - size="1024x1024", ) print(response) ``` From e289dfc09489fb743764359c4e9055580d755879 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Tue, 6 Jan 2026 13:44:32 +0530 Subject: [PATCH 62/73] Fix mypy issues --- litellm/llms/custom_httpx/llm_http_handler.py | 4 ++-- litellm/llms/custom_llm.py | 4 ++-- litellm/llms/recraft/image_edit/transformation.py | 4 ++-- .../llms/vertex_ai/image_edit/vertex_imagen_transformation.py | 2 ++ 4 files changed, 8 insertions(+), 6 deletions(-) diff --git a/litellm/llms/custom_httpx/llm_http_handler.py b/litellm/llms/custom_httpx/llm_http_handler.py index 1da6e61252f..2f6d74eb7a7 100644 --- a/litellm/llms/custom_httpx/llm_http_handler.py +++ b/litellm/llms/custom_httpx/llm_http_handler.py @@ -4453,7 +4453,7 @@ class BaseLLMHTTPHandler: self, model: str, image: Any, - prompt: str, + prompt: Optional[str], image_edit_provider_config: BaseImageEditConfig, image_edit_optional_request_params: Dict, custom_llm_provider: str, @@ -4572,7 +4572,7 @@ class BaseLLMHTTPHandler: self, model: str, image: FileTypes, - prompt: str, + prompt: Optional[str], image_edit_provider_config: BaseImageEditConfig, image_edit_optional_request_params: Dict, custom_llm_provider: str, diff --git a/litellm/llms/custom_llm.py b/litellm/llms/custom_llm.py index d235df30f25..a820ac7f345 100644 --- a/litellm/llms/custom_llm.py +++ b/litellm/llms/custom_llm.py @@ -201,7 +201,7 @@ class CustomLLM(BaseLLM): self, model: str, image: Any, - prompt: str, + prompt: Optional[str], model_response: ImageResponse, api_key: Optional[str], api_base: Optional[str], @@ -216,7 +216,7 @@ class CustomLLM(BaseLLM): self, model: str, image: Any, - prompt: str, + prompt: Optional[str], model_response: ImageResponse, api_key: Optional[str], api_base: Optional[str], diff --git a/litellm/llms/recraft/image_edit/transformation.py b/litellm/llms/recraft/image_edit/transformation.py index 94449257694..533a5108604 100644 --- a/litellm/llms/recraft/image_edit/transformation.py +++ b/litellm/llms/recraft/image_edit/transformation.py @@ -124,7 +124,7 @@ class RecraftImageEditConfig(BaseImageEditConfig): ######################################################### # Reuse OpenAI logic: Separate images as `files` and send other parameters as `data` ######################################################### - files_list = self._get_image_files_for_request(image=image) + files_list = self._get_image_files_for_request(image=image) if image is not None else [] data_without_images = {k: v for k, v in request_dict.items() if k != "image"} return data_without_images, files_list @@ -132,7 +132,7 @@ class RecraftImageEditConfig(BaseImageEditConfig): def _get_image_files_for_request( self, - image: FileTypes, + image: Optional[FileTypes], ) -> List[Tuple[str, Any]]: files_list: List[Tuple[str, Any]] = [] diff --git a/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py b/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py index b61af6ffd3a..1515e6cbe93 100644 --- a/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py +++ b/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py @@ -150,6 +150,8 @@ class VertexAIImagenImageEditConfig(BaseImageEditConfig, VertexLLM): headers: dict, ) -> Tuple[Dict[str, Any], Optional[RequestFiles]]: # Prepare reference images in the correct Imagen format + if image is None: + raise ValueError("Vertex AI Imagen image edit requires at least one reference image.") reference_images = self._prepare_reference_images(image, image_edit_optional_request_params) if not reference_images: raise ValueError("Vertex AI Imagen image edit requires at least one reference image.") From f8e25aa0166dcfdf2d47378b93943a2bb956d5be Mon Sep 17 00:00:00 2001 From: Yuta Saito Date: Fri, 16 Jan 2026 18:23:01 +0900 Subject: [PATCH 63/73] chore: add ALLOWED_CVES. Because Wolfi glibc still flagged even on 2.42-r5. --- ci_cd/security_scans.sh | 1 + 1 file changed, 1 insertion(+) diff --git a/ci_cd/security_scans.sh b/ci_cd/security_scans.sh index 42ae25026db..9931730b7ad 100755 --- a/ci_cd/security_scans.sh +++ b/ci_cd/security_scans.sh @@ -129,6 +129,7 @@ run_grype_scans() { "CVE-2025-13836" # Python 3.13 HTTP response reading OOM/DoS - no fix available in base image "CVE-2025-12084" # Python 3.13 xml.dom.minidom quadratic algorithm - no fix available in base image "CVE-2025-60876" # BusyBox wget HTTP request splitting - no fix available in Chainguard Wolfi base image + "CVE-2026-0861" # Wolfi glibc still flagged even on 2.42-r5; upstream patched build unavailable yet "CVE-2010-4756" # glibc glob DoS - awaiting patched Wolfi glibc build "CVE-2019-1010022" # glibc stack guard bypass - awaiting patched Wolfi glibc build "CVE-2019-1010023" # glibc ldd remap issue - awaiting patched Wolfi glibc build From ac5a4df72442ef6ee8f13df0c9c0f1ca921fa7f9 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Tue, 6 Jan 2026 15:17:35 +0530 Subject: [PATCH 64/73] Fix: vertex ai doesn't support structured output --- .../anthropic/transformation.py | 34 +++++ ...partner_models_anthropic_transformation.py | 144 ++++++++++++++++++ 2 files changed, 178 insertions(+) diff --git a/litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/transformation.py b/litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/transformation.py index 24425f08b56..1df07f405e6 100644 --- a/litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/transformation.py +++ b/litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/transformation.py @@ -69,6 +69,9 @@ class VertexAIAnthropicConfig(AnthropicConfig): data.pop("model", None) # vertex anthropic doesn't accept 'model' parameter + # VertexAI doesn't support output_format parameter, remove it if present + data.pop("output_format", None) + tools = optional_params.get("tools") tool_search_used = self.is_tool_search_used(tools) auto_betas = self.get_anthropic_beta_list( @@ -89,6 +92,37 @@ class VertexAIAnthropicConfig(AnthropicConfig): return data + def map_openai_params( + self, + non_default_params: dict, + optional_params: dict, + model: str, + drop_params: bool, + ) -> dict: + """ + Override parent method to ensure VertexAI always uses tool-based structured outputs. + VertexAI doesn't support the output_format parameter, so we force all models + to use the tool-based approach for structured outputs. + """ + # Temporarily override model name to force tool-based approach + # This ensures Claude Sonnet 4.5 uses tools instead of output_format + original_model = model + if "response_format" in non_default_params: + model = "claude-3-sonnet-20240229" # Use a model that will use tool-based approach + + # Call parent method with potentially modified model name + optional_params = super().map_openai_params( + non_default_params=non_default_params, + optional_params=optional_params, + model=model, + drop_params=drop_params, + ) + + # Restore original model name for any other processing + model = original_model + + return optional_params + def transform_response( self, model: str, diff --git a/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/test_vertex_ai_partner_models_anthropic_transformation.py b/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/test_vertex_ai_partner_models_anthropic_transformation.py index 5f2dd387b95..7b60a0a3369 100644 --- a/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/test_vertex_ai_partner_models_anthropic_transformation.py +++ b/tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/test_vertex_ai_partner_models_anthropic_transformation.py @@ -115,3 +115,147 @@ def test_vertex_ai_anthropic_structured_output_header_not_added(): "Non-Vertex request SHOULD have anthropic-beta header for structured output" assert result_non_vertex["anthropic-beta"] == "structured-outputs-2025-11-13", \ f"Expected 'structured-outputs-2025-11-13', got: {result_non_vertex.get('anthropic-beta')}" + + +def test_vertex_ai_claude_sonnet_4_5_structured_output_fix(): + """ + Test fix for issue #18625: Claude Sonnet 4.5 on VertexAI should use tool-based + structured outputs instead of output_format parameter. + + This test verifies that: + 1. Claude Sonnet 4.5 uses tool-based structured outputs on VertexAI + 2. output_format parameter is removed from the final request + 3. The fix prevents "Extra inputs are not permitted" error + """ + config = VertexAIAnthropicConfig() + + # Test data matching the issue report + response_format = { + "type": "json_schema", + "json_schema": { + "name": "questions", + "strict": True, + "schema": { + "type": "object", + "properties": { + "question": { + "type": "string" + }, + "response": { + "type": "string" + } + }, + "required": ["question", "response"], + "additionalProperties": False + } + } + } + + messages = [ + {"role": "user", "content": "Generate a question and answer about AI."} + ] + + # Test parameters that would trigger the issue + non_default_params = { + "response_format": response_format, + "max_tokens": 1000, + } + + # Test 1: Verify map_openai_params forces tool-based approach for Claude Sonnet 4.5 + optional_params = {} + result_params = config.map_openai_params( + non_default_params=non_default_params, + optional_params=optional_params, + model="claude-3-5-sonnet-20241022", # Claude Sonnet 4.5 model + drop_params=False, + ) + + # Should have tools and tool_choice (tool-based approach) + assert "tools" in result_params, "Tools should be present for structured output" + assert "tool_choice" in result_params, "Tool choice should be present for structured output" + assert "json_mode" in result_params, "JSON mode should be enabled" + + # Verify the tool is the response format tool + tools = result_params["tools"] + assert len(tools) == 1, "Should have exactly one tool for response format" + assert tools[0]["name"] == "json_tool_call", "Tool should be named json_tool_call" + + # Test 2: Verify transform_request removes output_format parameter + # Simulate what would happen if parent class added output_format + test_data = { + "model": "claude-3-5-sonnet-20241022", + "messages": messages, + "max_tokens": 1000, + "tools": tools, + "tool_choice": result_params["tool_choice"], + "output_format": { # This would be added by parent class for Sonnet 4.5 + "type": "json_schema", + "schema": response_format["json_schema"]["schema"] + } + } + + # Mock the parent transform_request to return data with output_format + original_transform = config.__class__.__bases__[0].transform_request + + def mock_transform_request(self, model, messages, optional_params, litellm_params, headers): + # Return test data that includes output_format + return test_data.copy() + + # Temporarily replace parent method + config.__class__.__bases__[0].transform_request = mock_transform_request + + try: + final_data = config.transform_request( + model="claude-3-5-sonnet-20241022", + messages=messages, + optional_params=result_params, + litellm_params={}, + headers={}, + ) + + # Verify that output_format was removed (fixes the "Extra inputs are not permitted" error) + assert "output_format" not in final_data, "output_format should be removed for VertexAI" + assert "model" not in final_data, "model should be removed for VertexAI" + assert "tools" in final_data, "tools should still be present" + assert "tool_choice" in final_data, "tool_choice should still be present" + + finally: + # Restore original method + config.__class__.__bases__[0].transform_request = original_transform + + +def test_vertex_ai_anthropic_other_models_still_use_tools(): + """ + Test that other Anthropic models (non-Sonnet 4.5) on VertexAI also use tool-based + structured outputs, ensuring consistency across all models. + """ + config = VertexAIAnthropicConfig() + + response_format = { + "type": "json_schema", + "json_schema": { + "name": "test_schema", + "schema": { + "type": "object", + "properties": { + "result": {"type": "string"} + } + } + } + } + + # Test with Claude 3 Sonnet (not 4.5) + non_default_params = {"response_format": response_format} + optional_params = {} + + result_params = config.map_openai_params( + non_default_params=non_default_params, + optional_params=optional_params, + model="claude-3-sonnet-20240229", + drop_params=False, + ) + + # Should still use tool-based approach + assert "tools" in result_params, "Claude 3 Sonnet should also use tool-based structured output" + assert "tool_choice" in result_params, "Tool choice should be present" + assert "json_mode" in result_params, "JSON mode should be enabled" From dcd66db4a8aa57935605155b75eefef7aa793ad5 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 15:21:10 +0530 Subject: [PATCH 65/73] Revert "fix: models loadbalancing billing issue by filter (#18891)" This reverts commit 41d8f799294bf2d5fe9122710c0091bb1cab7561. --- litellm/proxy/auth/model_checks.py | 25 +- litellm/proxy/litellm_pre_call_utils.py | 59 ++--- litellm/router.py | 12 +- litellm/router_utils/common_utils.py | 79 +----- ...est_filter_deployments_by_access_groups.py | 227 ------------------ 5 files changed, 20 insertions(+), 382 deletions(-) delete mode 100644 tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py diff --git a/litellm/proxy/auth/model_checks.py b/litellm/proxy/auth/model_checks.py index af2574d88ee..71ae1348f39 100644 --- a/litellm/proxy/auth/model_checks.py +++ b/litellm/proxy/auth/model_checks.py @@ -64,27 +64,6 @@ def _get_models_from_access_groups( return all_models -def get_access_groups_from_models( - model_access_groups: Dict[str, List[str]], - models: List[str], -) -> List[str]: - """ - Extract access group names from a models list. - - Given a models list like ["gpt-4", "beta-models", "claude-v1"] - and access groups like {"beta-models": ["gpt-5", "gpt-6"]}, - returns ["beta-models"]. - - This is used to pass allowed access groups to the router for filtering - deployments during load balancing (GitHub issue #18333). - """ - access_groups = [] - for model in models: - if model in model_access_groups: - access_groups.append(model) - return access_groups - - async def get_mcp_server_ids( user_api_key_dict: UserAPIKeyAuth, ) -> List[str]: @@ -101,6 +80,7 @@ async def get_mcp_server_ids( # Make a direct SQL query to get just the mcp_servers try: + result = await prisma_client.db.litellm_objectpermissiontable.find_unique( where={"object_permission_id": user_api_key_dict.object_permission_id}, ) @@ -196,7 +176,6 @@ def get_complete_model_list( """ unique_models = [] - def append_unique(models): for model in models: if model not in unique_models: @@ -209,7 +188,7 @@ def get_complete_model_list( else: append_unique(proxy_model_list) if include_model_access_groups: - append_unique(list(model_access_groups.keys())) # TODO: keys order + append_unique(list(model_access_groups.keys())) # TODO: keys order if user_model: append_unique([user_model]) diff --git a/litellm/proxy/litellm_pre_call_utils.py b/litellm/proxy/litellm_pre_call_utils.py index 7a49c1f6520..ad0ab6b7a38 100644 --- a/litellm/proxy/litellm_pre_call_utils.py +++ b/litellm/proxy/litellm_pre_call_utils.py @@ -173,12 +173,12 @@ def _get_dynamic_logging_metadata( user_api_key_dict: UserAPIKeyAuth, proxy_config: ProxyConfig ) -> Optional[TeamCallbackMetadata]: callback_settings_obj: Optional[TeamCallbackMetadata] = None - key_dynamic_logging_settings: Optional[dict] = ( - KeyAndTeamLoggingSettings.get_key_dynamic_logging_settings(user_api_key_dict) - ) - team_dynamic_logging_settings: Optional[dict] = ( - KeyAndTeamLoggingSettings.get_team_dynamic_logging_settings(user_api_key_dict) - ) + key_dynamic_logging_settings: Optional[ + dict + ] = KeyAndTeamLoggingSettings.get_key_dynamic_logging_settings(user_api_key_dict) + team_dynamic_logging_settings: Optional[ + dict + ] = KeyAndTeamLoggingSettings.get_team_dynamic_logging_settings(user_api_key_dict) ######################################################################################### # Key-based callbacks ######################################################################################### @@ -661,11 +661,11 @@ class LiteLLMProxyRequestSetup: ## KEY-LEVEL SPEND LOGS / TAGS if "tags" in key_metadata and key_metadata["tags"] is not None: - data[_metadata_variable_name]["tags"] = ( - LiteLLMProxyRequestSetup._merge_tags( - request_tags=data[_metadata_variable_name].get("tags"), - tags_to_add=key_metadata["tags"], - ) + data[_metadata_variable_name][ + "tags" + ] = LiteLLMProxyRequestSetup._merge_tags( + request_tags=data[_metadata_variable_name].get("tags"), + tags_to_add=key_metadata["tags"], ) if "disable_global_guardrails" in key_metadata and isinstance( key_metadata["disable_global_guardrails"], bool @@ -933,9 +933,9 @@ async def add_litellm_data_to_request( # noqa: PLR0915 data[_metadata_variable_name]["litellm_api_version"] = version if general_settings is not None: - data[_metadata_variable_name]["global_max_parallel_requests"] = ( - general_settings.get("global_max_parallel_requests", None) - ) + data[_metadata_variable_name][ + "global_max_parallel_requests" + ] = general_settings.get("global_max_parallel_requests", None) ### KEY-LEVEL Controls key_metadata = user_api_key_dict.metadata @@ -1002,37 +1002,6 @@ async def add_litellm_data_to_request( # noqa: PLR0915 "user_api_key_model_max_budget" ] = user_api_key_dict.model_max_budget - # Extract allowed access groups for router filtering (GitHub issue #18333) - # This allows the router to filter deployments based on key's and team's access groups - # NOTE: We keep key and team access groups SEPARATE because a key doesn't always - # inherit all team access groups (per maintainer feedback). - if llm_router is not None: - from litellm.proxy.auth.model_checks import get_access_groups_from_models - - model_access_groups = llm_router.get_model_access_groups() - - # Key-level access groups (from user_api_key_dict.models) - key_models = list(user_api_key_dict.models) if user_api_key_dict.models else [] - key_allowed_access_groups = get_access_groups_from_models( - model_access_groups=model_access_groups, models=key_models - ) - if key_allowed_access_groups: - data[_metadata_variable_name][ - "user_api_key_allowed_access_groups" - ] = key_allowed_access_groups - - # Team-level access groups (from user_api_key_dict.team_models) - team_models = ( - list(user_api_key_dict.team_models) if user_api_key_dict.team_models else [] - ) - team_allowed_access_groups = get_access_groups_from_models( - model_access_groups=model_access_groups, models=team_models - ) - if team_allowed_access_groups: - data[_metadata_variable_name][ - "user_api_key_team_allowed_access_groups" - ] = team_allowed_access_groups - data[_metadata_variable_name]["user_api_key_metadata"] = user_api_key_dict.metadata _headers = dict(request.headers) _headers.pop( diff --git a/litellm/router.py b/litellm/router.py index f73d907c8c7..dc07280ea16 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -86,7 +86,6 @@ from litellm.router_utils.clientside_credential_handler import ( is_clientside_credential, ) from litellm.router_utils.common_utils import ( - filter_deployments_by_access_groups, filter_team_based_models, filter_web_search_deployments, ) @@ -7847,17 +7846,10 @@ class Router: request_kwargs=request_kwargs, ) - verbose_router_logger.debug(f"healthy_deployments after web search filter: {healthy_deployments}") - - # Filter by allowed access groups (GitHub issue #18333) - # This prevents cross-team load balancing when teams have models with same name in different access groups - healthy_deployments = filter_deployments_by_access_groups( - healthy_deployments=healthy_deployments, - request_kwargs=request_kwargs, + verbose_router_logger.debug( + f"healthy_deployments after web search filter: {healthy_deployments}" ) - verbose_router_logger.debug(f"healthy_deployments after access group filter: {healthy_deployments}") - if isinstance(healthy_deployments, dict): return healthy_deployments diff --git a/litellm/router_utils/common_utils.py b/litellm/router_utils/common_utils.py index 2c0ea5976d6..10acc343abd 100644 --- a/litellm/router_utils/common_utils.py +++ b/litellm/router_utils/common_utils.py @@ -75,7 +75,6 @@ def filter_team_based_models( if deployment.get("model_info", {}).get("id") not in ids_to_remove ] - def _deployment_supports_web_search(deployment: Dict) -> bool: """ Check if a deployment supports web search. @@ -113,7 +112,7 @@ def filter_web_search_deployments( is_web_search_request = False tools = request_kwargs.get("tools") or [] for tool in tools: - # These are the two websearch tools for OpenAI / Azure. + # These are the two websearch tools for OpenAI / Azure. if tool.get("type") == "web_search" or tool.get("type") == "web_search_preview": is_web_search_request = True break @@ -122,82 +121,8 @@ def filter_web_search_deployments( return healthy_deployments # Filter out deployments that don't support web search - final_deployments = [ - d for d in healthy_deployments if _deployment_supports_web_search(d) - ] + final_deployments = [d for d in healthy_deployments if _deployment_supports_web_search(d)] if len(healthy_deployments) > 0 and len(final_deployments) == 0: verbose_logger.warning("No deployments support web search for request") return final_deployments - -def filter_deployments_by_access_groups( - healthy_deployments: Union[List[Dict], Dict], - request_kwargs: Optional[Dict] = None, -) -> Union[List[Dict], Dict]: - """ - Filter deployments to only include those matching the user's allowed access groups. - - Reads from TWO separate metadata fields (per maintainer feedback): - - `user_api_key_allowed_access_groups`: Access groups from the API Key's models. - - `user_api_key_team_allowed_access_groups`: Access groups from the Team's models. - - A deployment is included if its access_groups overlap with EITHER the key's - or the team's allowed access groups. Deployments with no access_groups are - always included (not restricted). - - This prevents cross-team load balancing when multiple teams have models with - the same name but in different access groups (GitHub issue #18333). - """ - if request_kwargs is None: - return healthy_deployments - - if isinstance(healthy_deployments, dict): - return healthy_deployments - - metadata = request_kwargs.get("metadata") or {} - litellm_metadata = request_kwargs.get("litellm_metadata") or {} - - # Gather key-level allowed access groups - key_allowed_access_groups = ( - metadata.get("user_api_key_allowed_access_groups") - or litellm_metadata.get("user_api_key_allowed_access_groups") - or [] - ) - - # Gather team-level allowed access groups - team_allowed_access_groups = ( - metadata.get("user_api_key_team_allowed_access_groups") - or litellm_metadata.get("user_api_key_team_allowed_access_groups") - or [] - ) - - # Combine both for the final allowed set - combined_allowed_access_groups = list(key_allowed_access_groups) + list( - team_allowed_access_groups - ) - - # If no access groups specified from either source, return all deployments (backwards compatible) - if not combined_allowed_access_groups: - return healthy_deployments - - allowed_set = set(combined_allowed_access_groups) - filtered = [] - for deployment in healthy_deployments: - model_info = deployment.get("model_info") or {} - deployment_access_groups = model_info.get("access_groups") or [] - - # If deployment has no access groups, include it (not restricted) - if not deployment_access_groups: - filtered.append(deployment) - continue - - # Include if any of deployment's groups overlap with allowed groups - if set(deployment_access_groups) & allowed_set: - filtered.append(deployment) - - if len(healthy_deployments) > 0 and len(filtered) == 0: - verbose_logger.warning( - f"No deployments match allowed access groups {combined_allowed_access_groups}" - ) - - return filtered diff --git a/tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py b/tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py deleted file mode 100644 index 9ac5072c5d8..00000000000 --- a/tests/test_litellm/router_unit_tests/test_filter_deployments_by_access_groups.py +++ /dev/null @@ -1,227 +0,0 @@ -""" -Unit tests for filter_deployments_by_access_groups function. - -Tests the fix for GitHub issue #18333: Models loadbalanced outside of Model Access Group. -""" - -import pytest - -from litellm.router_utils.common_utils import filter_deployments_by_access_groups - - -class TestFilterDeploymentsByAccessGroups: - """Tests for the filter_deployments_by_access_groups function.""" - - def test_no_filter_when_no_access_groups_in_metadata(self): - """When no allowed_access_groups in metadata, return all deployments.""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1"]}}, - {"model_info": {"id": "2", "access_groups": ["AG2"]}}, - ] - request_kwargs = {"metadata": {"user_api_key_team_id": "team-1"}} - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - assert len(result) == 2 # All deployments returned - - def test_filter_to_single_access_group(self): - """Filter to only deployments matching allowed access group.""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1"]}}, - {"model_info": {"id": "2", "access_groups": ["AG2"]}}, - ] - request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - assert len(result) == 1 - assert result[0]["model_info"]["id"] == "2" - - def test_filter_with_multiple_allowed_groups(self): - """Filter with multiple allowed access groups.""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1"]}}, - {"model_info": {"id": "2", "access_groups": ["AG2"]}}, - {"model_info": {"id": "3", "access_groups": ["AG3"]}}, - ] - request_kwargs = { - "metadata": {"user_api_key_allowed_access_groups": ["AG1", "AG2"]} - } - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - assert len(result) == 2 - ids = [d["model_info"]["id"] for d in result] - assert "1" in ids - assert "2" in ids - assert "3" not in ids - - def test_deployment_with_multiple_access_groups(self): - """Deployment with multiple access groups should match if any overlap.""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1", "AG2"]}}, - {"model_info": {"id": "2", "access_groups": ["AG3"]}}, - ] - request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - assert len(result) == 1 - assert result[0]["model_info"]["id"] == "1" - - def test_deployment_without_access_groups_included(self): - """Deployments without access groups should be included (not restricted).""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1"]}}, - {"model_info": {"id": "2"}}, # No access_groups - {"model_info": {"id": "3", "access_groups": []}}, # Empty access_groups - ] - request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - # Should include deployments 2 and 3 (no restrictions) - assert len(result) == 2 - ids = [d["model_info"]["id"] for d in result] - assert "2" in ids - assert "3" in ids - - def test_dict_deployment_passes_through(self): - """When deployment is a dict (specific deployment), pass through.""" - deployment = {"model_info": {"id": "1", "access_groups": ["AG1"]}} - request_kwargs = {"metadata": {"user_api_key_allowed_access_groups": ["AG2"]}} - - result = filter_deployments_by_access_groups( - healthy_deployments=deployment, - request_kwargs=request_kwargs, - ) - - assert result == deployment # Unchanged - - def test_none_request_kwargs_passes_through(self): - """When request_kwargs is None, return deployments unchanged.""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1"]}}, - ] - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=None, - ) - - assert result == deployments - - def test_litellm_metadata_fallback(self): - """Should also check litellm_metadata for allowed access groups.""" - deployments = [ - {"model_info": {"id": "1", "access_groups": ["AG1"]}}, - {"model_info": {"id": "2", "access_groups": ["AG2"]}}, - ] - request_kwargs = { - "litellm_metadata": {"user_api_key_allowed_access_groups": ["AG1"]} - } - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - assert len(result) == 1 - assert result[0]["model_info"]["id"] == "1" - - -def test_filter_deployments_by_access_groups_issue_18333(): - """ - Regression test for GitHub issue #18333. - - Scenario: Two models named 'gpt-5' in different access groups (AG1, AG2). - Team2 has access to AG2 only. When Team2 requests 'gpt-5', only the AG2 - deployment should be available for load balancing. - """ - deployments = [ - { - "model_name": "gpt-5", - "litellm_params": {"model": "gpt-4.1", "api_key": "key-1"}, - "model_info": {"id": "ag1-deployment", "access_groups": ["AG1"]}, - }, - { - "model_name": "gpt-5", - "litellm_params": {"model": "gpt-4o", "api_key": "key-2"}, - "model_info": {"id": "ag2-deployment", "access_groups": ["AG2"]}, - }, - ] - - # Team2's request with allowed access groups - request_kwargs = { - "metadata": { - "user_api_key_team_id": "team-2", - "user_api_key_allowed_access_groups": ["AG2"], - } - } - - result = filter_deployments_by_access_groups( - healthy_deployments=deployments, - request_kwargs=request_kwargs, - ) - - # Only AG2 deployment should be returned - assert len(result) == 1 - assert result[0]["model_info"]["id"] == "ag2-deployment" - assert result[0]["litellm_params"]["model"] == "gpt-4o" - - -def test_get_access_groups_from_models(): - """ - Test the helper function that extracts access group names from models list. - This is used by the proxy to populate user_api_key_allowed_access_groups. - """ - from litellm.proxy.auth.model_checks import get_access_groups_from_models - - # Setup: access groups definition - model_access_groups = { - "AG1": ["gpt-4", "gpt-5"], - "AG2": ["claude-v1", "claude-v2"], - "beta-models": ["gpt-5-turbo"], - } - - # Test 1: Extract access groups from models list - models = ["gpt-4", "AG1", "AG2", "some-other-model"] - result = get_access_groups_from_models( - model_access_groups=model_access_groups, models=models - ) - assert set(result) == {"AG1", "AG2"} - - # Test 2: No access groups in models list - models = ["gpt-4", "claude-v1", "some-model"] - result = get_access_groups_from_models( - model_access_groups=model_access_groups, models=models - ) - assert result == [] - - # Test 3: Empty models list - result = get_access_groups_from_models( - model_access_groups=model_access_groups, models=[] - ) - assert result == [] - - # Test 4: All access groups - models = ["AG1", "AG2", "beta-models"] - result = get_access_groups_from_models( - model_access_groups=model_access_groups, models=models - ) - assert set(result) == {"AG1", "AG2", "beta-models"} From 09fb1581cbd44913afce8e7052024aa7a57eb5c2 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 16:37:44 +0530 Subject: [PATCH 66/73] Fix:add async_get_available_deployment_for_pass_through in code tests --- .../test_router_get_deployments.py | 202 ++++++++++++++++++ 1 file changed, 202 insertions(+) diff --git a/tests/local_testing/test_router_get_deployments.py b/tests/local_testing/test_router_get_deployments.py index 358ed74f55c..8df04b4f1d3 100644 --- a/tests/local_testing/test_router_get_deployments.py +++ b/tests/local_testing/test_router_get_deployments.py @@ -592,3 +592,205 @@ async def test_weighted_selection_router_async(rpm_list, tpm_list): except Exception as e: traceback.print_exc() pytest.fail(f"Error occurred: {e}") + + +def test_get_available_deployment_for_pass_through(): + """ + Test get_available_deployment_for_pass_through function + - Tests that only deployments with use_in_pass_through=True are returned + - Tests that BadRequestError is raised when no pass-through deployments exist + """ + try: + litellm.set_verbose = False + model_list = [ + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "gpt-3.5-turbo", + "api_key": os.getenv("OPENAI_API_KEY"), + "use_in_pass_through": True, + }, + }, + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "azure/gpt-4.1-mini", + "api_key": os.getenv("AZURE_API_KEY"), + "api_base": os.getenv("AZURE_API_BASE"), + "api_version": os.getenv("AZURE_API_VERSION"), + "use_in_pass_through": False, + }, + }, + ] + router = Router( + model_list=model_list, + ) + + # Test that only pass-through deployment is returned + selected_model = router.get_available_deployment_for_pass_through( + "gpt-3.5-turbo" + ) + assert selected_model["litellm_params"]["model"] == "gpt-3.5-turbo" + assert selected_model["litellm_params"]["use_in_pass_through"] is True + + router.reset() + except Exception as e: + traceback.print_exc() + pytest.fail(f"Error occurred: {e}") + + +def test_get_available_deployment_for_pass_through_no_deployments(): + """ + Test get_available_deployment_for_pass_through raises BadRequestError + when no deployments have use_in_pass_through=True + """ + try: + litellm.set_verbose = False + model_list = [ + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "gpt-3.5-turbo", + "api_key": os.getenv("OPENAI_API_KEY"), + "use_in_pass_through": False, + }, + }, + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "azure/gpt-4.1-mini", + "api_key": os.getenv("AZURE_API_KEY"), + "api_base": os.getenv("AZURE_API_BASE"), + "api_version": os.getenv("AZURE_API_VERSION"), + "use_in_pass_through": False, + }, + }, + ] + router = Router( + model_list=model_list, + ) + + # Test that BadRequestError is raised when no pass-through deployments exist + try: + router.get_available_deployment_for_pass_through("gpt-3.5-turbo") + pytest.fail( + "Expected BadRequestError when no pass-through deployments exist" + ) + except litellm.BadRequestError as e: + assert "use_in_pass_through=True" in str(e) + + router.reset() + except Exception as e: + if isinstance(e, litellm.BadRequestError): + pass # Expected error + else: + traceback.print_exc() + pytest.fail(f"Error occurred: {e}") + + +@pytest.mark.asyncio +async def test_async_get_available_deployment_for_pass_through(): + """ + Test async_get_available_deployment_for_pass_through function + - Tests that only deployments with use_in_pass_through=True are returned + - Tests async version works correctly + """ + try: + litellm.set_verbose = False + model_list = [ + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "gpt-3.5-turbo", + "api_key": os.getenv("OPENAI_API_KEY"), + "use_in_pass_through": True, + }, + }, + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "azure/gpt-4.1-mini", + "api_key": os.getenv("AZURE_API_KEY"), + "api_base": os.getenv("AZURE_API_BASE"), + "api_version": os.getenv("AZURE_API_VERSION"), + "use_in_pass_through": False, + }, + }, + ] + router = Router( + model_list=model_list, + ) + + # Test that only pass-through deployment is returned + selected_model = await router.async_get_available_deployment_for_pass_through( + model="gpt-3.5-turbo", request_kwargs={} + ) + assert selected_model["litellm_params"]["model"] == "gpt-3.5-turbo" + assert selected_model["litellm_params"]["use_in_pass_through"] is True + + router.reset() + except Exception as e: + traceback.print_exc() + pytest.fail(f"Error occurred: {e}") + + +def test_filter_pass_through_deployments(): + """ + Test _filter_pass_through_deployments function + - Tests that it correctly filters deployments with use_in_pass_through=True + """ + try: + litellm.set_verbose = False + model_list = [ + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "gpt-3.5-turbo", + "api_key": os.getenv("OPENAI_API_KEY"), + "use_in_pass_through": True, + }, + }, + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "azure/gpt-4.1-mini", + "api_key": os.getenv("AZURE_API_KEY"), + "api_base": os.getenv("AZURE_API_BASE"), + "api_version": os.getenv("AZURE_API_VERSION"), + "use_in_pass_through": False, + }, + }, + { + "model_name": "gpt-3.5-turbo", + "litellm_params": { + "model": "azure/gpt-35-turbo", + "api_key": os.getenv("AZURE_API_KEY"), + "api_base": os.getenv("AZURE_API_BASE"), + "api_version": os.getenv("AZURE_API_VERSION"), + "use_in_pass_through": True, + }, + }, + ] + router = Router( + model_list=model_list, + ) + + # Get all healthy deployments + healthy_deployments = router.get_model_list() + + # Filter pass-through deployments + pass_through_deployments = router._filter_pass_through_deployments( + healthy_deployments + ) + + # Should only have 2 deployments with use_in_pass_through=True + assert len(pass_through_deployments) == 2 + + # Verify all returned deployments have use_in_pass_through=True + for deployment in pass_through_deployments: + assert deployment["litellm_params"]["use_in_pass_through"] is True + + router.reset() + except Exception as e: + traceback.print_exc() + pytest.fail(f"Error occurred: {e}") From 84974d5745149b7c1fecc7e365570b7d2fad8cb3 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 16:55:12 +0530 Subject: [PATCH 67/73] Fix boto3 conflicting dependency --- .circleci/config.yml | 26 ++++++++++---------- poetry.lock | 24 +++++++++--------- pyproject.toml | 2 +- requirements.txt | 2 +- tests/code_coverage_tests/license_cache.json | 2 +- 5 files changed, 28 insertions(+), 28 deletions(-) diff --git a/.circleci/config.yml b/.circleci/config.yml index dc3e6d64e98..2f21cc4481f 100644 --- a/.circleci/config.yml +++ b/.circleci/config.yml @@ -144,7 +144,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 @@ -260,7 +260,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 @@ -367,7 +367,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 @@ -637,7 +637,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install "langfuse>=2.0.0" @@ -759,7 +759,7 @@ jobs: pip install "google-cloud-aiplatform==1.43.0" pip install "google-genai==1.22.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 @@ -865,7 +865,7 @@ jobs: pip install "google-cloud-aiplatform==1.43.0" pip install "google-genai==1.22.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 @@ -972,7 +972,7 @@ jobs: pip install "google-cloud-aiplatform==1.43.0" pip install "google-genai==1.22.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install lunary==0.2.5 @@ -1198,7 +1198,7 @@ jobs: pip install "pytest-asyncio==0.21.1" pip install "respx==0.22.0" pip install "pydantic==2.10.2" - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" # Run pytest and generate JUnit XML report - run: name: Run tests @@ -1879,7 +1879,7 @@ jobs: pip install aiohttp pip install openai pip install click - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install jinja2 pip install "tokenizers==0.20.0" pip install "uvloop==0.21.0" @@ -2176,7 +2176,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install "langfuse>=2.0.0" @@ -2316,7 +2316,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install "langchain_mcp_adapters==0.0.5" @@ -2462,7 +2462,7 @@ jobs: pip install "google-generativeai==0.3.2" pip install "google-cloud-aiplatform==1.43.0" pip install pyarrow - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "aioboto3==15.5.0" pip install langchain pip install "langfuse>=2.0.0" @@ -3118,7 +3118,7 @@ jobs: pip install "pytest==7.3.1" pip install "pytest-mock==3.12.0" pip install "pytest-asyncio==0.21.1" - pip install "boto3==1.40.15" + pip install "boto3==1.40.61" pip install "mypy==1.18.2" pip install pyarrow pip install numpydoc diff --git a/poetry.lock b/poetry.lock index 249933b2ae1..35e97766189 100644 --- a/poetry.lock +++ b/poetry.lock @@ -525,21 +525,21 @@ files = [ [[package]] name = "boto3" -version = "1.40.15" +version = "1.40.61" description = "The AWS SDK for Python" optional = true python-versions = ">=3.9" groups = ["main"] markers = "extra == \"proxy\"" files = [ - {file = "boto3-1.40.15-py3-none-any.whl", hash = "sha256:52b8aa78c9906c4e49dcec6817c041df33c9825073bf66e7df8fc00afbe47b4b"}, - {file = "boto3-1.40.15.tar.gz", hash = "sha256:271b379ce5ad35ca82f1009e917528a182eed0e2de197ccffb0c51acadec5c79"}, + {file = "boto3-1.40.61-py3-none-any.whl", hash = "sha256:6b9c57b2a922b5d8c17766e29ed792586a818098efe84def27c8f582b33f898c"}, + {file = "boto3-1.40.61.tar.gz", hash = "sha256:d6c56277251adf6c2bdd25249feae625abe4966831676689ff23b4694dea5b12"}, ] [package.dependencies] -botocore = ">=1.40.15,<1.41.0" +botocore = ">=1.40.61,<1.41.0" jmespath = ">=0.7.1,<2.0.0" -s3transfer = ">=0.13.0,<0.14.0" +s3transfer = ">=0.14.0,<0.15.0" [package.extras] crt = ["botocore[crt] (>=1.21.0,<2.0a0)"] @@ -2375,7 +2375,7 @@ description = "WSGI HTTP Server for UNIX" optional = true python-versions = ">=3.7" groups = ["main"] -markers = "python_version >= \"3.10\" and platform_system != \"Windows\" and (extra == \"proxy\" or extra == \"mlflow\") or extra == \"proxy\"" +markers = "extra == \"proxy\" or (extra == \"mlflow\" or extra == \"proxy\") and platform_system != \"Windows\" and python_version >= \"3.10\"" files = [ {file = "gunicorn-23.0.0-py3-none-any.whl", hash = "sha256:ec400d38950de4dfd418cff8328b2c8faed0edb0d517d3394e457c317908ca4d"}, {file = "gunicorn-23.0.0.tar.gz", hash = "sha256:f014447a0101dc57e294f6c18ca6b40227a4c90e9bdb586042628030cba004ec"}, @@ -3433,8 +3433,8 @@ files = [ [package.dependencies] numpy = [ {version = ">=1.23.3", markers = "python_version >= \"3.11\""}, + {version = ">1.20"}, {version = ">=1.21.2", markers = "python_version >= \"3.10\""}, - {version = ">1.20", markers = "python_version < \"3.10\""}, {version = ">=1.26.0", markers = "python_version >= \"3.12\""}, ] @@ -6255,15 +6255,15 @@ files = [ [[package]] name = "s3transfer" -version = "0.13.1" +version = "0.14.0" description = "An Amazon S3 Transfer Manager" optional = true python-versions = ">=3.9" groups = ["main"] markers = "extra == \"proxy\"" files = [ - {file = "s3transfer-0.13.1-py3-none-any.whl", hash = "sha256:a981aa7429be23fe6dfc13e80e4020057cbab622b08c0315288758d67cabc724"}, - {file = "s3transfer-0.13.1.tar.gz", hash = "sha256:c3fdba22ba1bd367922f27ec8032d6a1cf5f10c934fb5d68cf60fd5a23d936cf"}, + {file = "s3transfer-0.14.0-py3-none-any.whl", hash = "sha256:ea3b790c7077558ed1f02a3072fb3cb992bbbd253392f4b6e9e8976941c7d456"}, + {file = "s3transfer-0.14.0.tar.gz", hash = "sha256:eff12264e7c8b4985074ccce27a3b38a485bb7f7422cc8046fee9be4983e4125"}, ] [package.dependencies] @@ -7201,7 +7201,7 @@ files = [ {file = "tomli-2.3.0-py3-none-any.whl", hash = "sha256:e95b1af3c5b07d9e643909b5abbec77cd9f1217e6d0bca72b0234736b9fb1f1b"}, {file = "tomli-2.3.0.tar.gz", hash = "sha256:64be704a875d2a59753d80ee8a533c3fe183e3f06807ff7dc2232938ccb01549"}, ] -markers = {main = "python_version == \"3.10\" and (extra == \"utils\" or extra == \"mlflow\") or extra == \"utils\" and python_version == \"3.9\"", dev = "python_version < \"3.11\"", proxy-dev = "python_version < \"3.11\""} +markers = {main = "extra == \"utils\" and python_version == \"3.9\" or python_version == \"3.10\" and (extra == \"utils\" or extra == \"mlflow\")", dev = "python_version < \"3.11\"", proxy-dev = "python_version < \"3.11\""} [[package]] name = "tomlkit" @@ -7981,4 +7981,4 @@ utils = ["numpydoc"] [metadata] lock-version = "2.1" python-versions = ">=3.9,<4.0" -content-hash = "a0d4bdda2742911291e79bab30faaaede14463f738c239425afcfe0f6b886d55" +content-hash = "f391c702cf58ef2ba7641acdc3ae13d7c8e672faede68c0a624bd2ba0fb46b12" diff --git a/pyproject.toml b/pyproject.toml index ceb8a9d5d0d..a5071353d6b 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -56,7 +56,7 @@ google-cloud-iam = {version = "^2.19.1", optional = true} resend = {version = ">=0.8.0", optional = true} pynacl = {version = "^1.5.0", optional = true} websockets = {version = "^15.0.1", optional = true} -boto3 = {version = "1.40.15", optional = true} +boto3 = {version = "1.40.61", optional = true} redisvl = {version = "^0.4.1", optional = true, markers = "python_version >= '3.9' and python_version < '3.14'"} mcp = {version = "^1.21.2", optional = true, python = ">=3.10"} litellm-proxy-extras = {version = "0.4.21", optional = true} diff --git a/requirements.txt b/requirements.txt index 6cc93d0351f..e98e295de30 100644 --- a/requirements.txt +++ b/requirements.txt @@ -10,7 +10,7 @@ uvicorn==0.31.1 # server dep gunicorn==23.0.0 # server dep fastuuid==0.13.5 # for uuid4 uvloop==0.21.0 # uvicorn dep, gives us much better performance under load -boto3==1.40.15 # aws bedrock/sagemaker calls +boto3==1.40.61 # aws bedrock/sagemaker calls redis==5.2.1 # redis caching prisma==0.11.0 # for db nodejs-wheel-binaries==24.12.0 ## required by prisma for migrations, prevents runtime download (updated from nodejs-bin for security fixes) diff --git a/tests/code_coverage_tests/license_cache.json b/tests/code_coverage_tests/license_cache.json index 21f74e26520..bd6c2be9ace 100644 --- a/tests/code_coverage_tests/license_cache.json +++ b/tests/code_coverage_tests/license_cache.json @@ -4,7 +4,7 @@ "pyyaml:6.0.2": "MIT", "gunicorn:22.0.0": "MIT", "uvloop:0.21.0": "MIT License", - "boto3:1.40.15": "Apache License 2.0", + "boto3:1.40.61": "Apache License 2.0", "redis:5.0.0": "MIT", "numpy:2.1.1": "Copyright (c) 2005-2024, NumPy Developers. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of the NumPy Developers nor the names of any contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. ---- The NumPy repository and source distributions bundle several libraries that are compatibly licensed. We list these here. Name: lapack-lite Files: numpy/linalg/lapack_lite/* License: BSD-3-Clause For details, see numpy/linalg/lapack_lite/LICENSE.txt Name: dragon4 Files: numpy/_core/src/multiarray/dragon4.c License: MIT For license text, see numpy/_core/src/multiarray/dragon4.c Name: libdivide Files: numpy/_core/include/numpy/libdivide/* License: Zlib For license text, see numpy/_core/include/numpy/libdivide/LICENSE.txt Note that the following files are vendored in the repository and sdist but not installed in built numpy packages: Name: Meson Files: vendored-meson/meson/* License: Apache 2.0 For license text, see vendored-meson/meson/COPYING Name: spin Files: .spin/cmds.py License: BSD-3 For license text, see .spin/LICENSE ---- This binary distribution of NumPy also bundles the following software: Name: OpenBLAS Files: numpy/.dylibs/libscipy_openblas*.so Description: bundled as a dynamically linked library Availability: https://github.com/OpenMathLib/OpenBLAS/ License: BSD-3-Clause Copyright (c) 2011-2014, The OpenBLAS Project All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. Neither the name of the OpenBLAS project nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Name: LAPACK Files: numpy/.dylibs/libscipy_openblas*.so Description: bundled in OpenBLAS Availability: https://github.com/OpenMathLib/OpenBLAS/ License: BSD-3-Clause-Attribution Copyright (c) 1992-2013 The University of Tennessee and The University of Tennessee Research Foundation. All rights reserved. Copyright (c) 2000-2013 The University of California Berkeley. All rights reserved. Copyright (c) 2006-2013 The University of Colorado Denver. All rights reserved. $COPYRIGHT$ Additional copyrights may follow $HEADER$ Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: - Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer listed in this license in the documentation and/or other materials provided with the distribution. - Neither the name of the copyright holders nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. The copyright holders provide no reassurances that the source code provided does not infringe any patent, copyright, or any other intellectual property rights of third parties. The copyright holders disclaim any liability to any recipient for claims brought against recipient by any third party for infringement of that parties intellectual property rights. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS \"AS IS\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. Name: GCC runtime library Files: numpy/.dylibs/libgfortran*, numpy/.dylibs/libgcc* Description: dynamically linked to files compiled with gcc Availability: https://gcc.gnu.org/git/?p=gcc.git;a=tree;f=libgfortran License: GPL-3.0-with-GCC-exception Copyright (C) 2002-2017 Free Software Foundation, Inc. Libgfortran is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version. Libgfortran is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. Under Section 7 of GPL version 3, you are granted additional permissions described in the GCC Runtime Library Exception, version 3.1, as published by the Free Software Foundation. You should have received a copy of the GNU General Public License and a copy of the GCC Runtime Library Exception along with this program; see the files COPYING3 and COPYING.RUNTIME respectively. If not, see . ---- Full text of license texts referred to above follows (that they are listed below does not necessarily imply the conditions apply to the present binary release): ---- GCC RUNTIME LIBRARY EXCEPTION Version 3.1, 31 March 2009 Copyright (C) 2009 Free Software Foundation, Inc. Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. This GCC Runtime Library Exception (\"Exception\") is an additional permission under section 7 of the GNU General Public License, version 3 (\"GPLv3\"). It applies to a given file (the \"Runtime Library\") that bears a notice placed by the copyright holder of the file stating that the file is governed by GPLv3 along with this Exception. When you use GCC to compile a program, GCC may combine portions of certain GCC header files and runtime libraries with the compiled program. The purpose of this Exception is to allow compilation of non-GPL (including proprietary) programs to use, in this way, the header files and runtime libraries covered by this Exception. 0. Definitions. A file is an \"Independent Module\" if it either requires the Runtime Library for execution after a Compilation Process, or makes use of an interface provided by the Runtime Library, but is not otherwise based on the Runtime Library. \"GCC\" means a version of the GNU Compiler Collection, with or without modifications, governed by version 3 (or a specified later version) of the GNU General Public License (GPL) with the option of using any subsequent versions published by the FSF. \"GPL-compatible Software\" is software whose conditions of propagation, modification and use would permit combination with GCC in accord with the license of GCC. \"Target Code\" refers to output from any compiler for a real or virtual target processor architecture, in executable form or suitable for input to an assembler, loader, linker and/or execution phase. Notwithstanding that, Target Code does not include data in any format that is used as a compiler intermediate representation, or used for producing a compiler intermediate representation. The \"Compilation Process\" transforms code entirely represented in non-intermediate languages designed for human-written code, and/or in Java Virtual Machine byte code, into Target Code. Thus, for example, use of source code generators and preprocessors need not be considered part of the Compilation Process, since the Compilation Process can be understood as starting with the output of the generators or preprocessors. A Compilation Process is \"Eligible\" if it is done using GCC, alone or with other GPL-compatible software, or if it is done without using any work based on GCC. For example, using non-GPL-compatible Software to optimize any GCC intermediate representations would not qualify as an Eligible Compilation Process. 1. Grant of Additional Permission. You have permission to propagate a work of Target Code formed by combining the Runtime Library with Independent Modules, even if such propagation would otherwise violate the terms of GPLv3, provided that all Target Code was generated by Eligible Compilation Processes. You may then convey such a combination under terms of your choice, consistent with the licensing of the Independent Modules. 2. No Weakening of GCC Copyleft. The availability of this Exception does not imply any general presumption that third-party software is unaffected by the copyleft requirements of the license of GCC. ---- GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007 Copyright (C) 2007 Free Software Foundation, Inc. Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. Preamble The GNU General Public License is a free, copyleft license for software and other kinds of works. The licenses for most software and other practical works are designed to take away your freedom to share and change the works. By contrast, the GNU General Public License is intended to guarantee your freedom to share and change all versions of a program--to make sure it remains free software for all its users. We, the Free Software Foundation, use the GNU General Public License for most of our software; it applies also to any other work released this way by its authors. You can apply it to your programs, too. When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for them if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs, and that you know you can do these things. To protect your rights, we need to prevent others from denying you these rights or asking you to surrender the rights. Therefore, you have certain responsibilities if you distribute copies of the software, or if you modify it: responsibilities to respect the freedom of others. For example, if you distribute copies of such a program, whether gratis or for a fee, you must pass on to the recipients the same freedoms that you received. You must make sure that they, too, receive or can get the source code. And you must show them these terms so they know their rights. Developers that use the GNU GPL protect your rights with two steps: (1) assert copyright on the software, and (2) offer you this License giving you legal permission to copy, distribute and/or modify it. For the developers' and authors' protection, the GPL clearly explains that there is no warranty for this free software. For both users' and authors' sake, the GPL requires that modified versions be marked as changed, so that their problems will not be attributed erroneously to authors of previous versions. Some devices are designed to deny users access to install or run modified versions of the software inside them, although the manufacturer can do so. This is fundamentally incompatible with the aim of protecting users' freedom to change the software. The systematic pattern of such abuse occurs in the area of products for individuals to use, which is precisely where it is most unacceptable. Therefore, we have designed this version of the GPL to prohibit the practice for those products. If such problems arise substantially in other domains, we stand ready to extend this provision to those domains in future versions of the GPL, as needed to protect the freedom of users. Finally, every program is threatened constantly by software patents. States should not allow patents to restrict development and use of software on general-purpose computers, but in those that do, we wish to avoid the special danger that patents applied to a free program could make it effectively proprietary. To prevent this, the GPL assures that patents cannot be used to render the program non-free. The precise terms and conditions for copying, distribution and modification follow. TERMS AND CONDITIONS 0. Definitions. \"This License\" refers to version 3 of the GNU General Public License. \"Copyright\" also means copyright-like laws that apply to other kinds of works, such as semiconductor masks. \"The Program\" refers to any copyrightable work licensed under this License. Each licensee is addressed as \"you\". \"Licensees\" and \"recipients\" may be individuals or organizations. To \"modify\" a work means to copy from or adapt all or part of the work in a fashion requiring copyright permission, other than the making of an exact copy. The resulting work is called a \"modified version\" of the earlier work or a work \"based on\" the earlier work. A \"covered work\" means either the unmodified Program or a work based on the Program. To \"propagate\" a work means to do anything with it that, without permission, would make you directly or secondarily liable for infringement under applicable copyright law, except executing it on a computer or modifying a private copy. Propagation includes copying, distribution (with or without modification), making available to the public, and in some countries other activities as well. To \"convey\" a work means any kind of propagation that enables other parties to make or receive copies. Mere interaction with a user through a computer network, with no transfer of a copy, is not conveying. An interactive user interface displays \"Appropriate Legal Notices\" to the extent that it includes a convenient and prominently visible feature that (1) displays an appropriate copyright notice, and (2) tells the user that there is no warranty for the work (except to the extent that warranties are provided), that licensees may convey the work under this License, and how to view a copy of this License. If the interface presents a list of user commands or options, such as a menu, a prominent item in the list meets this criterion. 1. Source Code. The \"source code\" for a work means the preferred form of the work for making modifications to it. \"Object code\" means any non-source form of a work. A \"Standard Interface\" means an interface that either is an official standard defined by a recognized standards body, or, in the case of interfaces specified for a particular programming language, one that is widely used among developers working in that language. The \"System Libraries\" of an executable work include anything, other than the work as a whole, that (a) is included in the normal form of packaging a Major Component, but which is not part of that Major Component, and (b) serves only to enable use of the work with that Major Component, or to implement a Standard Interface for which an implementation is available to the public in source code form. A \"Major Component\", in this context, means a major essential component (kernel, window system, and so on) of the specific operating system (if any) on which the executable work runs, or a compiler used to produce the work, or an object code interpreter used to run it. The \"Corresponding Source\" for a work in object code form means all the source code needed to generate, install, and (for an executable work) run the object code and to modify the work, including scripts to control those activities. However, it does not include the work's System Libraries, or general-purpose tools or generally available free programs which are used unmodified in performing those activities but which are not part of the work. For example, Corresponding Source includes interface definition files associated with source files for the work, and the source code for shared libraries and dynamically linked subprograms that the work is specifically designed to require, such as by intimate data communication or control flow between those subprograms and other parts of the work. The Corresponding Source need not include anything that users can regenerate automatically from other parts of the Corresponding Source. The Corresponding Source for a work in source code form is that same work. 2. Basic Permissions. All rights granted under this License are granted for the term of copyright on the Program, and are irrevocable provided the stated conditions are met. This License explicitly affirms your unlimited permission to run the unmodified Program. The output from running a covered work is covered by this License only if the output, given its content, constitutes a covered work. This License acknowledges your rights of fair use or other equivalent, as provided by copyright law. You may make, run and propagate covered works that you do not convey, without conditions so long as your license otherwise remains in force. You may convey covered works to others for the sole purpose of having them make modifications exclusively for you, or provide you with facilities for running those works, provided that you comply with the terms of this License in conveying all material for which you do not control copyright. Those thus making or running the covered works for you must do so exclusively on your behalf, under your direction and control, on terms that prohibit them from making any copies of your copyrighted material outside their relationship with you. Conveying under any other circumstances is permitted solely under the conditions stated below. Sublicensing is not allowed; section 10 makes it unnecessary. 3. Protecting Users' Legal Rights From Anti-Circumvention Law. No covered work shall be deemed part of an effective technological measure under any applicable law fulfilling obligations under article 11 of the WIPO copyright treaty adopted on 20 December 1996, or similar laws prohibiting or restricting circumvention of such measures. When you convey a covered work, you waive any legal power to forbid circumvention of technological measures to the extent such circumvention is effected by exercising rights under this License with respect to the covered work, and you disclaim any intention to limit operation or modification of the work as a means of enforcing, against the work's users, your or third parties' legal rights to forbid circumvention of technological measures. 4. Conveying Verbatim Copies. You may convey verbatim copies of the Program's source code as you receive it, in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate copyright notice; keep intact all notices stating that this License and any non-permissive terms added in accord with section 7 apply to the code; keep intact all notices of the absence of any warranty; and give all recipients a copy of this License along with the Program. You may charge any price or no price for each copy that you convey, and you may offer support or warranty protection for a fee. 5. Conveying Modified Source Versions. You may convey a work based on the Program, or the modifications to produce it from the Program, in the form of source code under the terms of section 4, provided that you also meet all of these conditions: a) The work must carry prominent notices stating that you modified it, and giving a relevant date. b) The work must carry prominent notices stating that it is released under this License and any conditions added under section 7. This requirement modifies the requirement in section 4 to \"keep intact all notices\". c) You must license the entire work, as a whole, under this License to anyone who comes into possession of a copy. This License will therefore apply, along with any applicable section 7 additional terms, to the whole of the work, and all its parts, regardless of how they are packaged. This License gives no permission to license the work in any other way, but it does not invalidate such permission if you have separately received it. d) If the work has interactive user interfaces, each must display Appropriate Legal Notices; however, if the Program has interactive interfaces that do not display Appropriate Legal Notices, your work need not make them do so. A compilation of a covered work with other separate and independent works, which are not by their nature extensions of the covered work, and which are not combined with it such as to form a larger program, in or on a volume of a storage or distribution medium, is called an \"aggregate\" if the compilation and its resulting copyright are not used to limit the access or legal rights of the compilation's users beyond what the individual works permit. Inclusion of a covered work in an aggregate does not cause this License to apply to the other parts of the aggregate. 6. Conveying Non-Source Forms. You may convey a covered work in object code form under the terms of sections 4 and 5, provided that you also convey the machine-readable Corresponding Source under the terms of this License, in one of these ways: a) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by the Corresponding Source fixed on a durable physical medium customarily used for software interchange. b) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by a written offer, valid for at least three years and valid for as long as you offer spare parts or customer support for that product model, to give anyone who possesses the object code either (1) a copy of the Corresponding Source for all the software in the product that is covered by this License, on a durable physical medium customarily used for software interchange, for a price no more than your reasonable cost of physically performing this conveying of source, or (2) access to copy the Corresponding Source from a network server at no charge. c) Convey individual copies of the object code with a copy of the written offer to provide the Corresponding Source. This alternative is allowed only occasionally and noncommercially, and only if you received the object code with such an offer, in accord with subsection 6b. d) Convey the object code by offering access from a designated place (gratis or for a charge), and offer equivalent access to the Corresponding Source in the same way through the same place at no further charge. You need not require recipients to copy the Corresponding Source along with the object code. If the place to copy the object code is a network server, the Corresponding Source may be on a different server (operated by you or a third party) that supports equivalent copying facilities, provided you maintain clear directions next to the object code saying where to find the Corresponding Source. Regardless of what server hosts the Corresponding Source, you remain obligated to ensure that it is available for as long as needed to satisfy these requirements. e) Convey the object code using peer-to-peer transmission, provided you inform other peers where the object code and Corresponding Source of the work are being offered to the general public at no charge under subsection 6d. A separable portion of the object code, whose source code is excluded from the Corresponding Source as a System Library, need not be included in conveying the object code work. A \"User Product\" is either (1) a \"consumer product\", which means any tangible personal property which is normally used for personal, family, or household purposes, or (2) anything designed or sold for incorporation into a dwelling. In determining whether a product is a consumer product, doubtful cases shall be resolved in favor of coverage. For a particular product received by a particular user, \"normally used\" refers to a typical or common use of that class of product, regardless of the status of the particular user or of the way in which the particular user actually uses, or expects or is expected to use, the product. A product is a consumer product regardless of whether the product has substantial commercial, industrial or non-consumer uses, unless such uses represent the only significant mode of use of the product. \"Installation Information\" for a User Product means any methods, procedures, authorization keys, or other information required to install and execute modified versions of a covered work in that User Product from a modified version of its Corresponding Source. The information must suffice to ensure that the continued functioning of the modified object code is in no case prevented or interfered with solely because modification has been made. If you convey an object code work under this section in, or with, or specifically for use in, a User Product, and the conveying occurs as part of a transaction in which the right of possession and use of the User Product is transferred to the recipient in perpetuity or for a fixed term (regardless of how the transaction is characterized), the Corresponding Source conveyed under this section must be accompanied by the Installation Information. But this requirement does not apply if neither you nor any third party retains the ability to install modified object code on the User Product (for example, the work has been installed in ROM). The requirement to provide Installation Information does not include a requirement to continue to provide support service, warranty, or updates for a work that has been modified or installed by the recipient, or for the User Product in which it has been modified or installed. Access to a network may be denied when the modification itself materially and adversely affects the operation of the network or violates the rules and protocols for communication across the network. Corresponding Source conveyed, and Installation Information provided, in accord with this section must be in a format that is publicly documented (and with an implementation available to the public in source code form), and must require no special password or key for unpacking, reading or copying. 7. Additional Terms. \"Additional permissions\" are terms that supplement the terms of this License by making exceptions from one or more of its conditions. Additional permissions that are applicable to the entire Program shall be treated as though they were included in this License, to the extent that they are valid under applicable law. If additional permissions apply only to part of the Program, that part may be used separately under those permissions, but the entire Program remains governed by this License without regard to the additional permissions. When you convey a copy of a covered work, you may at your option remove any additional permissions from that copy, or from any part of it. (Additional permissions may be written to require their own removal in certain cases when you modify the work.) You may place additional permissions on material, added by you to a covered work, for which you have or can give appropriate copyright permission. Notwithstanding any other provision of this License, for material you add to a covered work, you may (if authorized by the copyright holders of that material) supplement the terms of this License with terms: a) Disclaiming warranty or limiting liability differently from the terms of sections 15 and 16 of this License; or b) Requiring preservation of specified reasonable legal notices or author attributions in that material or in the Appropriate Legal Notices displayed by works containing it; or c) Prohibiting misrepresentation of the origin of that material, or requiring that modified versions of such material be marked in reasonable ways as different from the original version; or d) Limiting the use for publicity purposes of names of licensors or authors of the material; or e) Declining to grant rights under trademark law for use of some trade names, trademarks, or service marks; or f) Requiring indemnification of licensors and authors of that material by anyone who conveys the material (or modified versions of it) with contractual assumptions of liability to the recipient, for any liability that these contractual assumptions directly impose on those licensors and authors. All other non-permissive additional terms are considered \"further restrictions\" within the meaning of section 10. If the Program as you received it, or any part of it, contains a notice stating that it is governed by this License along with a term that is a further restriction, you may remove that term. If a license document contains a further restriction but permits relicensing or conveying under this License, you may add to a covered work material governed by the terms of that license document, provided that the further restriction does not survive such relicensing or conveying. If you add terms to a covered work in accord with this section, you must place, in the relevant source files, a statement of the additional terms that apply to those files, or a notice indicating where to find the applicable terms. Additional terms, permissive or non-permissive, may be stated in the form of a separately written license, or stated as exceptions; the above requirements apply either way. 8. Termination. You may not propagate or modify a covered work except as expressly provided under this License. Any attempt otherwise to propagate or modify it is void, and will automatically terminate your rights under this License (including any patent licenses granted under the third paragraph of section 11). However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation. Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice. Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, you do not qualify to receive new licenses for the same material under section 10. 9. Acceptance Not Required for Having Copies. You are not required to accept this License in order to receive or run a copy of the Program. Ancillary propagation of a covered work occurring solely as a consequence of using peer-to-peer transmission to receive a copy likewise does not require acceptance. However, nothing other than this License grants you permission to propagate or modify any covered work. These actions infringe copyright if you do not accept this License. Therefore, by modifying or propagating a covered work, you indicate your acceptance of this License to do so. 10. Automatic Licensing of Downstream Recipients. Each time you convey a covered work, the recipient automatically receives a license from the original licensors, to run, modify and propagate that work, subject to this License. You are not responsible for enforcing compliance by third parties with this License. An \"entity transaction\" is a transaction transferring control of an organization, or substantially all assets of one, or subdividing an organization, or merging organizations. If propagation of a covered work results from an entity transaction, each party to that transaction who receives a copy of the work also receives whatever licenses to the work the party's predecessor in interest had or could give under the previous paragraph, plus a right to possession of the Corresponding Source of the work from the predecessor in interest, if the predecessor has it or can get it with reasonable efforts. You may not impose any further restrictions on the exercise of the rights granted or affirmed under this License. For example, you may not impose a license fee, royalty, or other charge for exercise of rights granted under this License, and you may not initiate litigation (including a cross-claim or counterclaim in a lawsuit) alleging that any patent claim is infringed by making, using, selling, offering for sale, or importing the Program or any portion of it. 11. Patents. A \"contributor\" is a copyright holder who authorizes use under this License of the Program or a work on which the Program is based. The work thus licensed is called the contributor's \"contributor version\". A contributor's \"essential patent claims\" are all patent claims owned or controlled by the contributor, whether already acquired or hereafter acquired, that would be infringed by some manner, permitted by this License, of making, using, or selling its contributor version, but do not include claims that would be infringed only as a consequence of further modification of the contributor version. For purposes of this definition, \"control\" includes the right to grant patent sublicenses in a manner consistent with the requirements of this License. Each contributor grants you a non-exclusive, worldwide, royalty-free patent license under the contributor's essential patent claims, to make, use, sell, offer for sale, import and otherwise run, modify and propagate the contents of its contributor version. In the following three paragraphs, a \"patent license\" is any express agreement or commitment, however denominated, not to enforce a patent (such as an express permission to practice a patent or covenant not to sue for patent infringement). To \"grant\" such a patent license to a party means to make such an agreement or commitment not to enforce a patent against the party. If you convey a covered work, knowingly relying on a patent license, and the Corresponding Source of the work is not available for anyone to copy, free of charge and under the terms of this License, through a publicly available network server or other readily accessible means, then you must either (1) cause the Corresponding Source to be so available, or (2) arrange to deprive yourself of the benefit of the patent license for this particular work, or (3) arrange, in a manner consistent with the requirements of this License, to extend the patent license to downstream recipients. \"Knowingly relying\" means you have actual knowledge that, but for the patent license, your conveying the covered work in a country, or your recipient's use of the covered work in a country, would infringe one or more identifiable patents in that country that you have reason to believe are valid. If, pursuant to or in connection with a single transaction or arrangement, you convey, or propagate by procuring conveyance of, a covered work, and grant a patent license to some of the parties receiving the covered work authorizing them to use, propagate, modify or convey a specific copy of the covered work, then the patent license you grant is automatically extended to all recipients of the covered work and works based on it. A patent license is \"discriminatory\" if it does not include within the scope of its coverage, prohibits the exercise of, or is conditioned on the non-exercise of one or more of the rights that are specifically granted under this License. You may not convey a covered work if you are a party to an arrangement with a third party that is in the business of distributing software, under which you make payment to the third party based on the extent of your activity of conveying the work, and under which the third party grants, to any of the parties who would receive the covered work from you, a discriminatory patent license (a) in connection with copies of the covered work conveyed by you (or copies made from those copies), or (b) primarily for and in connection with specific products or compilations that contain the covered work, unless you entered into that arrangement, or that patent license was granted, prior to 28 March 2007. Nothing in this License shall be construed as excluding or limiting any implied license or other defenses to infringement that may otherwise be available to you under applicable patent law. 12. No Surrender of Others' Freedom. If conditions are imposed on you (whether by court order, agreement or otherwise) that contradict the conditions of this License, they do not excuse you from the conditions of this License. If you cannot convey a covered work so as to satisfy simultaneously your obligations under this License and any other pertinent obligations, then as a consequence you may not convey it at all. For example, if you agree to terms that obligate you to collect a royalty for further conveying from those to whom you convey the Program, the only way you could satisfy both those terms and this License would be to refrain entirely from conveying the Program. 13. Use with the GNU Affero General Public License. Notwithstanding any other provision of this License, you have permission to link or combine any covered work with a work licensed under version 3 of the GNU Affero General Public License into a single combined work, and to convey the resulting work. The terms of this License will continue to apply to the part which is the covered work, but the special requirements of the GNU Affero General Public License, section 13, concerning interaction through a network will apply to the combination as such. 14. Revised Versions of this License. The Free Software Foundation may publish revised and/or new versions of the GNU General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. Each version is given a distinguishing version number. If the Program specifies that a certain numbered version of the GNU General Public License \"or any later version\" applies to it, you have the option of following the terms and conditions either of that numbered version or of any later version published by the Free Software Foundation. If the Program does not specify a version number of the GNU General Public License, you may choose any version ever published by the Free Software Foundation. If the Program specifies that a proxy can decide which future versions of the GNU General Public License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Program. Later license versions may give you additional or different permissions. However, no additional obligations are imposed on any author or copyright holder as a result of your choosing to follow a later version. 15. Disclaimer of Warranty. THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM \"AS IS\" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 16. Limitation of Liability. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. 17. Interpretation of Sections 15 and 16. If the disclaimer of warranty and limitation of liability provided above cannot be given local legal effect according to their terms, reviewing courts shall apply local law that most closely approximates an absolute waiver of all civil liability in connection with the Program, unless a warranty or assumption of liability accompanies a copy of the Program in return for a fee. END OF TERMS AND CONDITIONS How to Apply These Terms to Your New Programs If you develop a new program, and you want it to be of the greatest possible use to the public, the best way to achieve this is to make it free software which everyone can redistribute and change under these terms. To do so, attach the following notices to the program. It is safest to attach them to the start of each source file to most effectively state the exclusion of warranty; and each file should have at least the \"copyright\" line and a pointer to where the full notice is found. Copyright (C) This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . Also add information on how to contact you by electronic and paper mail. If the program does terminal interaction, make it output a short notice like this when it starts in an interactive mode: Copyright (C) This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. This is free software, and you are welcome to redistribute it under certain conditions; type `show c' for details. The hypothetical commands `show w' and `show c' should show the appropriate parts of the General Public License. Of course, your program's commands might be different; for a GUI interface, you would use an \"about box\". You should also get your employer (if you work as a programmer) or school, if any, to sign a \"copyright disclaimer\" for the program, if necessary. For more information on this, and how to apply and follow the GNU GPL, see . The GNU General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Lesser General Public License instead of this License. But first, please read . Name: libquadmath Files: numpy/.dylibs/libquadmath*.so Description: dynamically linked to files compiled with gcc Availability: https://gcc.gnu.org/git/?p=gcc.git;a=tree;f=libquadmath License: LGPL-2.1-or-later GCC Quad-Precision Math Library Copyright (C) 2010-2019 Free Software Foundation, Inc. Written by Francois-Xavier Coudert This file is part of the libquadmath library. Libquadmath is free software; you can redistribute it and/or modify it under the terms of the GNU Library General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. Libquadmath is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. https://www.gnu.org/licenses/old-licenses/lgpl-2.1.html", "prisma:0.11.0": "APACHE", From 95f98a4c521f953490223cfb07ebe029266cb77e Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 17:03:09 +0530 Subject: [PATCH 68/73] Potential fix for code scanning alert no. 3990: Clear-text logging of sensitive information Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com> --- litellm/litellm_core_utils/prompt_templates/factory.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/litellm/litellm_core_utils/prompt_templates/factory.py b/litellm/litellm_core_utils/prompt_templates/factory.py index 2f57ad9e813..c530ae7964f 100644 --- a/litellm/litellm_core_utils/prompt_templates/factory.py +++ b/litellm/litellm_core_utils/prompt_templates/factory.py @@ -2145,7 +2145,7 @@ def _is_orphaned_tool_result( if not found_matching_tool_call: verbose_logger.debug( - f"_is_orphaned_tool_result: Found orphaned tool result with tool_call_id={tool_call_id}" + "_is_orphaned_tool_result: Found orphaned tool result with redacted tool_call_id" ) return True From c76b527281813fd45c298f374c71019d80979836 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 17:17:21 +0530 Subject: [PATCH 69/73] Fix model map --- litellm/model_prices_and_context_window_backup.json | 12 +----------- model_prices_and_context_window.json | 2 +- 2 files changed, 2 insertions(+), 12 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index e58db912cf4..7c410a4ab30 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -25525,24 +25525,14 @@ "stability.stable-conservative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, -<<<<<<< HEAD - "mode": "image_edit", - "output_cost_per_image": 0.4 -======= "mode": "image_edits", "output_cost_per_image": 0.40 ->>>>>>> b712575d64 (fix Updated all 27 occurrences of mode: image_edit to mode: image_edits) }, "stability.stable-creative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, -<<<<<<< HEAD - "mode": "image_edit", - "output_cost_per_image": 0.6 -======= "mode": "image_edits", "output_cost_per_image": 0.60 ->>>>>>> b712575d64 (fix Updated all 27 occurrences of mode: image_edit to mode: image_edits) }, "stability.stable-fast-upscale-v1:0": { "litellm_provider": "bedrock", @@ -33940,4 +33930,4 @@ "litellm_provider": "llamagate", "mode": "embedding" } -} \ No newline at end of file +} diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 52b41a464eb..7c410a4ab30 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -33930,4 +33930,4 @@ "litellm_provider": "llamagate", "mode": "embedding" } -} \ No newline at end of file +} From ce105abdfbf1fc69feea4cf0d4ae8c2ed2472f9f Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 17:41:42 +0530 Subject: [PATCH 70/73] Fix all mypy issues --- litellm/llms/azure_ai/image_edit/flux2_transformation.py | 5 ++++- litellm/llms/base_llm/image_edit/transformation.py | 2 +- .../llms/bedrock/image_edit/stability_transformation.py | 9 ++++++--- litellm/llms/gemini/image_edit/transformation.py | 5 ++++- litellm/llms/openai/image_edit/dalle2_transformation.py | 7 +++++-- litellm/llms/openai/image_edit/transformation.py | 5 ++++- litellm/llms/recraft/image_edit/transformation.py | 5 ++++- litellm/llms/stability/image_edit/transformations.py | 9 ++++++--- .../vertex_ai/image_edit/vertex_gemini_transformation.py | 5 ++++- .../vertex_ai/image_edit/vertex_imagen_transformation.py | 5 ++++- 10 files changed, 42 insertions(+), 15 deletions(-) diff --git a/litellm/llms/azure_ai/image_edit/flux2_transformation.py b/litellm/llms/azure_ai/image_edit/flux2_transformation.py index caa39056675..87bae59ba0f 100644 --- a/litellm/llms/azure_ai/image_edit/flux2_transformation.py +++ b/litellm/llms/azure_ai/image_edit/flux2_transformation.py @@ -87,7 +87,7 @@ class AzureFoundryFlux2ImageEditConfig(OpenAIImageEditConfig): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, @@ -99,6 +99,9 @@ class AzureFoundryFlux2ImageEditConfig(OpenAIImageEditConfig): FLUX 2 uses the same endpoint for generation and editing, with the image passed as base64 in the JSON body. """ + if prompt is None: + raise ValueError("FLUX 2 image edit requires a prompt.") + image_b64 = self._convert_image_to_base64(image) # Build request body with required params diff --git a/litellm/llms/base_llm/image_edit/transformation.py b/litellm/llms/base_llm/image_edit/transformation.py index d522675296f..cc723480371 100644 --- a/litellm/llms/base_llm/image_edit/transformation.py +++ b/litellm/llms/base_llm/image_edit/transformation.py @@ -92,7 +92,7 @@ class BaseImageEditConfig(ABC): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, diff --git a/litellm/llms/bedrock/image_edit/stability_transformation.py b/litellm/llms/bedrock/image_edit/stability_transformation.py index bcaf0923f69..e8b77812988 100644 --- a/litellm/llms/bedrock/image_edit/stability_transformation.py +++ b/litellm/llms/bedrock/image_edit/stability_transformation.py @@ -21,18 +21,18 @@ Supported models: API Reference: https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters.html """ -import json import base64 +import json from typing import TYPE_CHECKING, Any, Dict, Optional, Tuple import httpx from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig from litellm.types.images.main import ImageEditOptionalRequestParams -from litellm.types.router import GenericLiteLLMParams from litellm.types.llms.stability import ( OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO, ) +from litellm.types.router import GenericLiteLLMParams from litellm.types.utils import FileTypes, ImageObject, ImageResponse from litellm.utils import get_model_info @@ -153,7 +153,7 @@ class BedrockStabilityImageEditConfig(BaseImageEditConfig): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, @@ -164,6 +164,9 @@ class BedrockStabilityImageEditConfig(BaseImageEditConfig): Returns the request body dict that will be JSON-encoded by the handler. """ + if prompt is None: + raise ValueError("Bedrock Stability image edit requires a prompt.") + # Build Bedrock Stability request data: Dict[str, Any] = { "prompt": prompt, diff --git a/litellm/llms/gemini/image_edit/transformation.py b/litellm/llms/gemini/image_edit/transformation.py index 78a7ff9546f..0015155b47f 100644 --- a/litellm/llms/gemini/image_edit/transformation.py +++ b/litellm/llms/gemini/image_edit/transformation.py @@ -80,7 +80,7 @@ class GeminiImageEditConfig(BaseImageEditConfig): def transform_image_edit_request( # type: ignore[override] self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict[str, Any], litellm_params: GenericLiteLLMParams, @@ -90,6 +90,9 @@ class GeminiImageEditConfig(BaseImageEditConfig): if not inline_parts: raise ValueError("Gemini image edit requires at least one image.") + if prompt is None: + raise ValueError("Gemini image edit requires a prompt.") + contents = [ { "parts": inline_parts + [{"text": prompt}], diff --git a/litellm/llms/openai/image_edit/dalle2_transformation.py b/litellm/llms/openai/image_edit/dalle2_transformation.py index 37e92be17a8..13531546d2e 100644 --- a/litellm/llms/openai/image_edit/dalle2_transformation.py +++ b/litellm/llms/openai/image_edit/dalle2_transformation.py @@ -1,5 +1,5 @@ from io import BufferedReader -from typing import TYPE_CHECKING, Any, Dict, List, Tuple, cast +from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, cast from httpx._types import RequestFiles @@ -30,7 +30,7 @@ class DallE2ImageEditConfig(OpenAIImageEditConfig): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, @@ -41,6 +41,9 @@ class DallE2ImageEditConfig(OpenAIImageEditConfig): DALL-E-2 only accepts a single image with field name "image" (not "image[]"). """ + if prompt is None: + raise ValueError("DALL-E-2 image edit requires a prompt.") + request = ImageEditRequestParams( model=model, image=image, diff --git a/litellm/llms/openai/image_edit/transformation.py b/litellm/llms/openai/image_edit/transformation.py index 1b90d96fa92..9edad9ee2c9 100644 --- a/litellm/llms/openai/image_edit/transformation.py +++ b/litellm/llms/openai/image_edit/transformation.py @@ -79,7 +79,7 @@ class OpenAIImageEditConfig(BaseImageEditConfig): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, @@ -91,6 +91,9 @@ class OpenAIImageEditConfig(BaseImageEditConfig): Handles multipart/form-data for images. Uses "image[]" field name to support multiple images (e.g., for gpt-image-1). """ + if prompt is None: + raise ValueError("OpenAI image edit requires a prompt.") + request = ImageEditRequestParams( model=model, image=image, diff --git a/litellm/llms/recraft/image_edit/transformation.py b/litellm/llms/recraft/image_edit/transformation.py index 533a5108604..9bf46704ed1 100644 --- a/litellm/llms/recraft/image_edit/transformation.py +++ b/litellm/llms/recraft/image_edit/transformation.py @@ -101,7 +101,7 @@ class RecraftImageEditConfig(BaseImageEditConfig): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, @@ -114,6 +114,9 @@ class RecraftImageEditConfig(BaseImageEditConfig): https://www.recraft.ai/docs#image-to-image """ + if prompt is None: + raise ValueError("Recraft image edit requires a prompt.") + request_body: RecraftImageEditRequestParams = RecraftImageEditRequestParams( model=model, prompt=prompt, diff --git a/litellm/llms/stability/image_edit/transformations.py b/litellm/llms/stability/image_edit/transformations.py index 173fae2d6fd..013e3f27a02 100644 --- a/litellm/llms/stability/image_edit/transformations.py +++ b/litellm/llms/stability/image_edit/transformations.py @@ -14,11 +14,11 @@ from httpx._types import RequestFiles from litellm.llms.base_llm.image_edit.transformation import BaseImageEditConfig from litellm.secret_managers.main import get_secret_str from litellm.types.images.main import ImageEditOptionalRequestParams -from litellm.types.router import GenericLiteLLMParams from litellm.types.llms.stability import ( OPENAI_SIZE_TO_STABILITY_ASPECT_RATIO, STABILITY_EDIT_ENDPOINTS, ) +from litellm.types.router import GenericLiteLLMParams from litellm.types.utils import FileTypes, ImageObject, ImageResponse from litellm.utils import get_model_info @@ -170,7 +170,7 @@ class StabilityImageEditConfig(BaseImageEditConfig): def transform_image_edit_request( self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict, litellm_params: GenericLiteLLMParams, @@ -186,9 +186,12 @@ class StabilityImageEditConfig(BaseImageEditConfig): # Populate multipart form-data as separate text fields (data) and files. # Stability expects prompt/output_format/etc. as normal form fields, not file parts. data: Dict[str, Any] = { - "prompt": prompt, "output_format": "png", # Default to PNG } + + # Add prompt only if provided (some Stability endpoints don't require it) + if prompt is not None: + data["prompt"] = prompt # Handle image parameter - could be a single file or list image_file = image[0] if isinstance(image, list) else image # type: ignore files: Dict[str, Any] = {"image": image_file} diff --git a/litellm/llms/vertex_ai/image_edit/vertex_gemini_transformation.py b/litellm/llms/vertex_ai/image_edit/vertex_gemini_transformation.py index 174d05cf7cf..154d5669eb8 100644 --- a/litellm/llms/vertex_ai/image_edit/vertex_gemini_transformation.py +++ b/litellm/llms/vertex_ai/image_edit/vertex_gemini_transformation.py @@ -151,7 +151,7 @@ class VertexAIGeminiImageEditConfig(BaseImageEditConfig, VertexLLM): def transform_image_edit_request( # type: ignore[override] self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict[str, Any], litellm_params: GenericLiteLLMParams, @@ -161,6 +161,9 @@ class VertexAIGeminiImageEditConfig(BaseImageEditConfig, VertexLLM): if not inline_parts: raise ValueError("Vertex AI Gemini image edit requires at least one image.") + if prompt is None: + raise ValueError("Vertex AI Gemini image edit requires a prompt.") + # Correct format for Vertex AI Gemini image editing contents = { "role": "USER", diff --git a/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py b/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py index 1515e6cbe93..337a4bd4dd6 100644 --- a/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py +++ b/litellm/llms/vertex_ai/image_edit/vertex_imagen_transformation.py @@ -143,7 +143,7 @@ class VertexAIImagenImageEditConfig(BaseImageEditConfig, VertexLLM): def transform_image_edit_request( # type: ignore[override] self, model: str, - prompt: str, + prompt: Optional[str], image: FileTypes, image_edit_optional_request_params: Dict[str, Any], litellm_params: GenericLiteLLMParams, @@ -156,6 +156,9 @@ class VertexAIImagenImageEditConfig(BaseImageEditConfig, VertexLLM): if not reference_images: raise ValueError("Vertex AI Imagen image edit requires at least one reference image.") + if prompt is None: + raise ValueError("Vertex AI Imagen image edit requires a prompt.") + # Correct Imagen instances format instances = [ { From 48d1e769a8423552090b4f7f14ffd6f296f26fe8 Mon Sep 17 00:00:00 2001 From: Sameer Kankute Date: Fri, 16 Jan 2026 18:37:05 +0530 Subject: [PATCH 71/73] Add azure/gpt-5.2-codex --- ...odel_prices_and_context_window_backup.json | 73 +++++++++++++++++++ model_prices_and_context_window.json | 31 ++++++++ 2 files changed, 104 insertions(+) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 7c410a4ab30..4862bd4a738 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -3634,6 +3634,37 @@ "supports_tool_choice": true, "supports_vision": true }, + "azure/gpt-5.2-codex": { + "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_token": 1.75e-06, + "litellm_provider": "azure", + "max_input_tokens": 128000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.4e-05, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true + }, "azure/gpt-5.2-pro": { "input_cost_per_token": 2.1e-05, "litellm_provider": "azure", @@ -10170,6 +10201,48 @@ "mode": "completion", "output_cost_per_token": 5e-07 }, + "deepseek-v3-2-251201": { + "input_cost_per_token": 0.0, + "litellm_provider": "volcengine", + "max_input_tokens": 98304, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 0.0, + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "glm-4-7-251222": { + "input_cost_per_token": 0.0, + "litellm_provider": "volcengine", + "max_input_tokens": 204800, + "max_output_tokens": 131072, + "max_tokens": 131072, + "mode": "chat", + "output_cost_per_token": 0.0, + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, + "kimi-k2-thinking-251104": { + "input_cost_per_token": 0.0, + "litellm_provider": "volcengine", + "max_input_tokens": 229376, + "max_output_tokens": 32768, + "max_tokens": 32768, + "mode": "chat", + "output_cost_per_token": 0.0, + "supports_assistant_prefill": true, + "supports_function_calling": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_tool_choice": true + }, "doubao-embedding": { "input_cost_per_token": 0.0, "litellm_provider": "volcengine", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 9604db2ee05..4862bd4a738 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -3634,6 +3634,37 @@ "supports_tool_choice": true, "supports_vision": true }, + "azure/gpt-5.2-codex": { + "cache_read_input_token_cost": 1.75e-07, + "input_cost_per_token": 1.75e-06, + "litellm_provider": "azure", + "max_input_tokens": 128000, + "max_output_tokens": 16384, + "max_tokens": 16384, + "mode": "chat", + "output_cost_per_token": 1.4e-05, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_function_calling": true, + "supports_native_streaming": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true + }, "azure/gpt-5.2-pro": { "input_cost_per_token": 2.1e-05, "litellm_provider": "azure", From 7d06216b7727e9ca9dd9491233e4ce23e6c0d139 Mon Sep 17 00:00:00 2001 From: Ishaan Jaffer Date: Fri, 16 Jan 2026 08:32:30 -0800 Subject: [PATCH 72/73] ci/cd fixes --- litellm/__init__.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/litellm/__init__.py b/litellm/__init__.py index 2b7d26de129..9eb3f075d5e 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -9,7 +9,7 @@ warnings.filterwarnings("ignore", message=".*conflict with protected namespace.* warnings.filterwarnings( "ignore", message=".*Accessing the.*attribute on the instance is deprecated.*" ) -### INIT VARIABLES ######################## +### INIT VARIABLES ######################### import threading import os from typing import ( From b86aae02121b33e54165ad8209c634a84d811b60 Mon Sep 17 00:00:00 2001 From: Ishaan Jaffer Date: Fri, 16 Jan 2026 08:54:48 -0800 Subject: [PATCH 73/73] fix stability mode --- ...odel_prices_and_context_window_backup.json | 54 +++++++++---------- model_prices_and_context_window.json | 54 +++++++++---------- 2 files changed, 54 insertions(+), 54 deletions(-) diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 4862bd4a738..4abbddb0d50 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -25456,7 +25456,7 @@ }, "stability/inpaint": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25464,7 +25464,7 @@ }, "stability/outpaint": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.004, "supported_endpoints": [ "/v1/images/edits" @@ -25472,7 +25472,7 @@ }, "stability/erase": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25480,7 +25480,7 @@ }, "stability/search-and-replace": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25488,7 +25488,7 @@ }, "stability/search-and-recolor": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25496,7 +25496,7 @@ }, "stability/remove-background": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25504,7 +25504,7 @@ }, "stability/replace-background-and-relight": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25512,7 +25512,7 @@ }, "stability/sketch": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25520,7 +25520,7 @@ }, "stability/structure": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25528,7 +25528,7 @@ }, "stability/style": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25536,7 +25536,7 @@ }, "stability/style-transfer": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25544,7 +25544,7 @@ }, "stability/fast": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.002, "supported_endpoints": [ "/v1/images/edits" @@ -25552,7 +25552,7 @@ }, "stability/conservative": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.04, "supported_endpoints": [ "/v1/images/edits" @@ -25560,7 +25560,7 @@ }, "stability/creative": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.06, "supported_endpoints": [ "/v1/images/edits" @@ -25598,79 +25598,79 @@ "stability.stable-conservative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.40 }, "stability.stable-creative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.60 }, "stability.stable-fast-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.03 }, "stability.stable-outpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.06 }, "stability.stable-image-control-sketch-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-control-structure-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-erase-object-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-inpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-remove-background-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-search-recolor-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-search-replace-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-style-guide-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-style-transfer-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.08 }, "stability.stable-image-core-v1:1": { diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 4862bd4a738..4abbddb0d50 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -25456,7 +25456,7 @@ }, "stability/inpaint": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25464,7 +25464,7 @@ }, "stability/outpaint": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.004, "supported_endpoints": [ "/v1/images/edits" @@ -25472,7 +25472,7 @@ }, "stability/erase": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25480,7 +25480,7 @@ }, "stability/search-and-replace": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25488,7 +25488,7 @@ }, "stability/search-and-recolor": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25496,7 +25496,7 @@ }, "stability/remove-background": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25504,7 +25504,7 @@ }, "stability/replace-background-and-relight": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25512,7 +25512,7 @@ }, "stability/sketch": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25520,7 +25520,7 @@ }, "stability/structure": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25528,7 +25528,7 @@ }, "stability/style": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.005, "supported_endpoints": [ "/v1/images/edits" @@ -25536,7 +25536,7 @@ }, "stability/style-transfer": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.008, "supported_endpoints": [ "/v1/images/edits" @@ -25544,7 +25544,7 @@ }, "stability/fast": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.002, "supported_endpoints": [ "/v1/images/edits" @@ -25552,7 +25552,7 @@ }, "stability/conservative": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.04, "supported_endpoints": [ "/v1/images/edits" @@ -25560,7 +25560,7 @@ }, "stability/creative": { "litellm_provider": "stability", - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.06, "supported_endpoints": [ "/v1/images/edits" @@ -25598,79 +25598,79 @@ "stability.stable-conservative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.40 }, "stability.stable-creative-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.60 }, "stability.stable-fast-upscale-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.03 }, "stability.stable-outpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.06 }, "stability.stable-image-control-sketch-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-control-structure-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-erase-object-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-inpaint-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-remove-background-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-search-recolor-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-search-replace-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-image-style-guide-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.07 }, "stability.stable-style-transfer-v1:0": { "litellm_provider": "bedrock", "max_input_tokens": 77, - "mode": "image_edits", + "mode": "image_edit", "output_cost_per_image": 0.08 }, "stability.stable-image-core-v1:1": {