* feat(realtime guardrails): end_session_after_n_fails + Endpoint Settings wizard step
Adds per-session violation thresholds and an optional endpoint-settings step
to the guardrail wizard for /v1/realtime.
Backend:
- Add end_session_after_n_fails, on_violation, realtime_violation_message fields
to BaseLitellmParams (no DB migration — stored in existing JSON column)
- Store same fields on CustomGuardrail instance attrs
- Pass through in litellm_content_filter initializer
- Track _violation_count per RealTimeStreaming session; close backend_ws when
on_violation=end_session OR violation count >= end_session_after_n_fails
- Use realtime_violation_message as the spoken text (falls back to guardrail
error string if not configured)
UI (add_guardrail_form.tsx):
- Rename "Default Categories" step to "Topics"
- Add step 5 "Endpoint Settings (Optional)" for content filter guardrails
- Call type dropdown shows /v1/realtime
- Settings are in a collapsed accordion (closed by default)
- "End session after X violations" + on_violation radio + spoken message field
Tests: 2 new tests in test_realtime_streaming.py
- test_end_session_after_n_fails_closes_connection
- test_on_violation_end_session_closes_on_first_fail
* fix(test): move inline imports to module level in realtime streaming tests
* Update ui/litellm-dashboard/src/components/guardrails/add_guardrail_form.tsx
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
---------
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* fix(realtime): guardrails with pre_call/post_call mode now work on realtime WebSocket; return error directly to consumer
* fix(realtime guardrails): address code review feedback
- Restore session.update injection for audio/VAD path, but only when
realtime_input_transcription guardrails are configured (not pre_call).
Forward session.created to the client first so no error arrives before
the client sees the session.
- Change _swallow_next_response_create bool to int counter so consecutive
blocked items are handled correctly.
- Extract _build_litellm_metadata() helper to eliminate duplicated
metadata-building logic across OpenAI/Azure/XAI provider branches.
- Plumb litellm_metadata and user_api_key_dict to Azure and XAI handlers
so guardrails work for those providers too.
- Add tests for session.update injection, no-inject for pre_call-only,
and consecutive-block counter.
* simplify: remove response.create swallowing after guardrail block
When an item is blocked, the error event is already sent to the client.
The subsequent response.create from the client is fine to forward through —
the LLM may respond to previous context which is acceptable behavior.
Removing the swallow counter eliminates unnecessary state tracking.
* feat: add pretty view for realtime API logs in dashboard
- Create RealtimePrettyView component that renders structured session
config, conversation turns with transcripts, and token breakdowns
- Update PrettyMessagesView to detect realtime responses (via
isRealtimeResponse helper) and delegate to the new component
- Session card shows model, voice, modalities, temperature, instructions
in a collapsible panel
- Conversation turns show status, per-turn token usage, and audio/text
transcripts with appropriate icons
- Add 24 tests for RealtimePrettyView and 3 tests for PrettyMessagesView
- All 75 LogDetailsDrawer tests pass
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* chore: remove dev_config.yaml from tracked files
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* feat: show turn count in realtime pretty view session header and output header
- Add purple 'N turns' tag to Session card header for at-a-glance turn count
- Add 'Turns: N' to the Output section header next to tokens/cost
- Extend SectionHeader to accept optional turnCount prop
- Add 3 new tests for turn count display (singular, plural, output header)
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* fix: address Greptile review feedback
- Remove response.audio.done and conversation.item.created from
isRealtimeResponse() detection since the view doesn't render them;
prevents misleading fallback for responses with only those events
- Remove dead code: index >= 0 is always true in .map() callback
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* fix(adapter): populate cache_read_input_tokens from prompt_tokens_details
The Anthropic adapter's translate_openai_response_to_anthropic checked
only the private _cache_read_input_tokens attr (set by Anthropic/DeepSeek)
but not prompt_tokens_details.cached_tokens (set by OpenAI/Azure).
Use prompt_tokens_details.cached_tokens directly — it is already extracted
and is the standard field populated by all providers.
Fixes#22089
* fix(adapter): apply same cache_read_input_tokens fix to streaming path
The streaming path in translate_streaming_openai_response_to_anthropic
had the same bug — relying on _cache_read_input_tokens instead of
prompt_tokens_details.cached_tokens.
* fix(proxy): improve auth exception logging levels and add structured context
Downgrade expected auth failures (ProxyException, HTTPException < 500,
BudgetExceededError) from ERROR to WARNING log level to reduce noise from
routine rejected requests (e.g. missing/invalid API keys on polled endpoints
like /schedule/model_cost_map_reload/status).
Unexpected exceptions and HTTPException with status >= 500 still log at
ERROR with full traceback.
Enrich log messages with structured context: route, HTTP method, masked
API key (using existing abbreviate_api_key), error type, and error code.
All fields also passed via log extra dict for log aggregation tools.
Fixes#21293
* Update tests/test_litellm/proxy/auth/test_auth_exception_handler.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
---------
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* feat(vertex_ai): add Vertex AI Gemini Live support via unified /realtime endpoint
Adds VertexAIRealtimeConfig which translates the OpenAI Realtime WebSocket
protocol to Vertex AI BidiGenerateContent. Supports voice in/voice out
(16 kHz mic → 24 kHz speaker) and text in/text out through the proxy's
/realtime endpoint.
Key changes:
- New litellm/llms/vertex_ai/realtime/transformation.py with VertexAIRealtimeConfig
- Builds correct wss:// URL (regional + global)
- OAuth2 Bearer token auth (not API key)
- Full model path (projects/.../publishers/google/models/...)
- Ignores session.update (Vertex AI only accepts one setup message)
- realtime_api/main.py: vertex_ai branch resolves OAuth token + constructs config
- llm_http_handler.py: auto-sends session setup before bidirectional_forward
- gemini/realtime/transformation.py: fix crashes on empty turnComplete events
- realtime_streaming.py: try/except guard so bad messages don't kill the loop
- proxy_server.py: add missing websockets.exceptions import
* docs: add vertex_realtime to sidebars
* fix: drop unknown event types in Gemini transform; add vertex_ai health check
* fix: propagate UUID fallback IDs from transform_content_done_event to return_additional_content_done_events
* fix: route guardrail backend sends through provider transform; fix str.strip misuse for model prefix
* fix: handle Vertex AI full resource path in session.created; route guardrail block sends through _send_to_backend
* fix: remove unused VertexBase in transformation.py; apply UUID fallback in return_additional_content_done_events
Adds a new block_code_execution guardrail that detects markdown fenced code blocks
in request/response content and blocks or masks them by language. Includes full
UI integration, type definitions, compliance test dataset, and 26 unit tests.
Key guardrail capabilities:
- Regex-based fenced code block detection with configurable blocked languages
- Confidence scoring with tunable threshold
- Execution-intent heuristics (request-side only) with conflict resolution
- Block or mask actions for detected code
- Support for pre_call, post_call, and during_call event hooks
Security hardening:
- Response-side blocking skips intent heuristics (LLM output doesn't contain
user intent phrases, so checking would silently disable post_call blocking)
- No-execution short-circuit includes conflict resolution: if both no-execution
and execution phrases match, execution intent wins
- Tightened overly broad phrases to prevent trivial bypass
- _normalize_escaped_newlines only applies to pure-escaped payloads to avoid
corrupting content that discusses escape sequences
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Replace the non-paginated Key Alias filter with a new PaginatedKeyAliasSelect component that mirrors the existing PaginatedModelSelect pattern. This aligns the UI with the paginated /key/aliases endpoint from PR #22137.
Changes:
- Added useInfiniteKeyAliases hook for paginated key alias fetching
- Created PaginatedKeyAliasSelect component with infinite scroll (80% threshold)
- Updated keyAliasesCall in networking to accept page/size/search params
- Replaced Key Alias filter in Request Logs and Virtual Keys tables to use customComponent
- Removed fetchAllKeyAliases helper and related upfront fetching logic
- Added 22 tests for new component and hook; all existing tests pass (54 tests)
Fixes the issue where the UI was fetching all key aliases at once, causing performance issues with large key sets.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
* feat(realtime): add guardrails query param to /v1/realtime WebSocket endpoint
- Add 'guardrails' query param (comma-separated) to realtime_websocket_endpoint
- Import websockets and websockets.exceptions at module level (fixes NameError in except clause)
- Split try/except into Phase 1 (pre-call) and Phase 2 (routing) so guardrail
errors send back a typed error event before closing, while upstream errors
close silently with 1011
* feat(ui): pass selectedGuardrails from sidebar to RealtimePlayground WebSocket URL
* docs(realtime): add guardrails section with dynamic passing examples
* Add claims agent guardrails with 243-case eval dataset
5 new category guardrails for healthcare claims agent chatbots:
- claims_fraud_coaching: fraud coaching, exaggeration, document forgery
- claims_phi_disclosure: unauthorized PHI access, bulk data extraction
- claims_prior_auth_gaming: code manipulation, medical necessity misrepresentation
- claims_system_override: system injection, rule bypass, role impersonation
- claims_medical_advice: medical advice (claims-context-aware)
Plus claims_agent_safety.yaml policy template combining all 5.
All 5 eval suites pass at 100% precision/recall/F1 (243 test cases).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add claims agent chatbot safety policy template
Combines the 5 claims guardrails into a single deployable policy template:
fraud coaching, PHI disclosure, prior-auth gaming, system override, and medical advice.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Add guardrail benchmark results and UI compliance prompts
Adds benchmark results for claims, discrimination, and content filter guardrails.
Updates UI compliance prompt data.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Remove duplicate "file an appeal" exception in claims_prior_auth_gaming.yaml
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* Remove unused claims_agent_safety.yaml policy template
The claims-agent-safety template in policy_templates.json references
individual category files in categories/, not this combined file.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* Add post-call hook for Lakera guardrail and mask PII in responses
* Add post-call hook for Lakera and mask PII in responses
* Fix post-call hook: pass event_type to call_v2_guard
* Address Greptile review: return ModelResponse, fix mutation, add header, test location, mask order
- PII masking path: return ModelResponse instead of dict so deployment hook accepts it
- Avoid mutating request data: deep copy original_messages and messages in _mask_pii_in_messages
- Add guardrail header in PII-only return path
- Add test in tests/test_litellm/ (test_lakera_ai_v2.py) per PR checklist
- Sort PII payload spans by (start,end) descending so multiple spans in one message mask correctly
Co-authored-by: Cursor <cursoragent@cursor.com>
* Updated ponteital for index mismatch when choices have null content and inconsistent on_flagged access pattern
* Update litellm/proxy/guardrails/guardrail_hooks/lakera_ai_v2.py
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
* Update to explicitly state supported endpoints - chat completions
* Fix minor lint error on masked_entity_count
---------
Co-authored-by: Steve <steve.giguere@lakera.ai>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
The backend validator and frontend form both enforced auth_value as
required when auth_type is api_key, bearer_token, or basic. Users who
want to provide auth dynamically (via per-request headers or OAuth2
flows) could not skip the field.
- Remove required validation from auth_value in create_mcp_server.tsx
(keep whitespace-only rejection, matching the edit flow)
- Remove validate_credentials_requirements in NewMCPServerRequest
(all downstream code already treats auth_value as optional)
- Add tests for the create MCP server component
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Use the backend-provided request_duration_ms field instead of computing
duration client-side from startTime/endTime. Add sort support for the
Duration column, which sends sortBy=request_duration_ms to the API.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a prominent warning banner to the UI dashboard when detailed debug
mode (LITELLM_LOG=DEBUG) is enabled. This alerts users to significant
performance degradation caused by extensive diagnostic logging.
Backend changes:
- Enhanced /health/readiness endpoint to include log_level and
is_detailed_debug fields
- Added detection using verbose_logger.getEffectiveLevel()
- Backward compatible - old clients ignore new fields
Frontend changes:
- Updated useHealthReadiness TypeScript interface
- Created DebugWarningBanner component using Ant Design Alert
- Integrated banner into dashboard layout below navbar
- Banner only shows when DEBUG level is active
- Non-dismissible to ensure users are aware of performance impact
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Backend - Spend Log Storage for Realtime Calls:
- Collect user voice transcripts and text input during WebSocket sessions
- Store collected messages in spend logs when store_prompts_in_spend_logs enabled
- Capture tool definitions from session.update and tool calls from response.done
- Enrich proxy_server_request with tools and response with tool_calls for UI
Backend - WebSocket Auth:
- Support browser-based auth via Sec-WebSocket-Protocol subprotocol
- Echo back subprotocol on WebSocket accept
UI - Realtime Playground:
- New RealtimePlayground component with WebSocket voice+text chat
- Mic recording (PCM16 24kHz), server VAD, audio playback, text input
- Handle binary WebSocket frames (Blob/ArrayBuffer decoding)
- Add /v1/realtime endpoint option to playground endpoint selector
UI - Tools Section for Realtime Logs:
- Extract tool calls from realtime response format (response.tool_calls
and response.results[].response.output[].type=function_call)
Tests:
- 15 new backend tests for realtime streaming and spend log storage
- 4 new UI tests for realtime tool call extraction
Fixes pre-existing build errors:
- ToolPolicies.tsx: duplicate import, antd styles type
- create_key_button.tsx: missing message import
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* docs: document x-litellm-tags header and request body tags parameter
- Add documentation for x-litellm-tags header (comma-separated or array)
- Add documentation for tags in request body
- Clarify that dynamic tags override config tags
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* docs: consolidate tag documentation and improve cross-references
- Make request_tags.md the single source of truth for all tag options
- Add cross-reference from cost_tracking.md to request_tags.md
- Document both direct tags and metadata.tags formats
- Add key/team tag setup and custom header tracking to request_tags.md
- Reduce duplication and make navigation clearer
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* docs: use generic examples instead of specific company names
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* docs: clarify x-litellm-tags header format is comma-separated string
HTTP headers are always strings, not arrays. Remove misleading
array format documentation for the header parameter.
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
* Update docs/my-website/docs/proxy/request_tags.md
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Ishaan Jaff <ishaan-jaff@users.noreply.github.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
Replace literal Base64-encoded example value in docstring with a
placeholder to prevent GitGuardian/ggshield from flagging it as a
leaked Basic Authentication credential in container scans.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
- Add per-command error check in _pipeline_lpop_helper to match _pipeline_rpush_helper, preventing silent data loss on WRONGTYPE errors
- Fix pre-existing bug: org spend queue metric was using REDIS_DAILY_SPEND_UPDATE_QUEUE instead of REDIS_DAILY_ORG_SPEND_UPDATE_QUEUE
- Add test for per-command LPOP pipeline error propagation
* feat(proxy): add max_iterations limiter for agent session loops (#22058)
Adds a new proxy hook that enforces a per-session cap on the number of
LLM calls an agentic loop can make. Callers send a session_id with each
request, and the hook counts calls per session, returning 429 when the
configured max_iterations limit is exceeded.
- Uses Redis Lua script for atomic increment (multi-instance safe)
- Falls back to in-memory cache when Redis unavailable
- Follows parallel_request_limiter_v3 pattern
- Configurable via key metadata: {"max_iterations": 25}
- Session counters auto-expire via TTL (default 1hr)
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
* feat: add new code execution dataset
* feat(agent_endpoints/): allow giving agents keys
* fix: ui fixes
* feat: allow assigning mcp servers to agents
* fix: eliminate duplicate DB queries in MCP agent auth and N+1 in agent listing (#22110)
- Extract _get_agent_object_permission helper so _get_allowed_mcp_servers_for_agent
and _get_agent_tool_permissions_for_server share a single DB fetch instead of
each independently querying the same agent row (was 1+N queries per MCP request)
- Use include={"object_permission": True} on find_many in get_all_agents_from_db
to eagerly load permissions in one query instead of N+1
- Use include={"object_permission": True} on create/update/find_unique in all
agent CRUD operations, removing attach_object_permission_to_dict follow-up calls
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Add new document explaining automatic credential usage tracking and tagging. When models use reusable credentials, LiteLLM automatically injects a Credential: <name> tag on requests, enabling credential-level spend tracking on the Usage page with no additional configuration.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
Added unit tests for all pricing calculator components with 64 passing tests across 5 test files:
- multi_export_utils.test.ts (16 tests for PDF/CSV export functions)
- use_multi_cost_estimate.test.ts (15 tests for cost estimation hook)
- multi_export_dropdown.test.tsx (8 tests for export dropdown component)
- multi_cost_results.test.tsx (15 tests for results display and UI states)
- index.test.tsx (10 tests for main calculator component)
Also fixed missing page description for tool-policies page in page_metadata.ts.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
