* fix(otel v2): map rerank and search output and the OCR, image edit and search input onto the Langfuse generation
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(otel v2): summarize OCR data URIs by media type and size and log an empty document URL as empty
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(otel v2): keep URL-less search results, name OCR file streams and skip non-str query parts when logging
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(otel v2): drop the unused typing imports and the decorative section divider
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(cost): bill batch prompts above 272K at OpenAI's long-context batch tier
* fix(cost): mirror batch long-context keys on custom pricing params
Register the two *_above_272k_tokens_batches keys on CustomPricingLiteLLMParams so a per-deployment override stays out of the shared backend key, add them to the inline model-info schema and alias-count tests, and build LiteLLM_Params and GenericLiteLLMParams through model_validate at the two dict-splat call sites so basedpyright's reportArgumentType budget ratchets down instead of blocking the new fields.
* fix(cost): add the gpt-5.5-pro batch long-context tier and ignore malformed batch tier keys
* fix(cost): bill cached batch tokens at OpenAI's cached batch rate
Adds cache_read_input_token_cost_batches and
cache_read_input_token_cost_above_272k_tokens_batches for the tiered
OpenAI entries at half the standard cached rate, bills cached batch
tokens at that rate per output line, and parses string-valued batch
rates in deployment-level model_info.
* fix(cost): bill batch cache writes at the batch cache-write rate and carry published batch rates for one-sided deployments
OpenAI's Batch table prices cache writes for gpt-6-astra, gpt-5.6, gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna at half the standard cache-write rate, so the cost map gains cache_creation_input_token_cost_batches and its above_272k tier for those entries and batch cost pulls written tokens out of the input bucket at that rate; models without the key keep billing writes at the batch input rate.
A deployment declaring only one side of its batch pricing now carries every published batch rate of the other side (tier, cached, cache write), its own keys win, and a lone tier, cached or cache-write batch key counts as declared pricing instead of being ignored.
* fix(cost): select the batch long-context tier from any batch tier key
A deployment that declares its own flat standard input rate keeps every
published batch rate of the output direction, including the 272K output
tier, but the tier was only ever selected when an input tier key was also
present. Detect the crossed tier from any of the four batch tier keys so
the carried output, cache-read, and cache-write tiers bill at their tier
rate above 272K tokens.
* chore(proxy): keep the OpenAPI snapshot as CI generates it
* fix(cost): pick each batch price component's tier from its own keys
The batch rate picker crossed one threshold for every component, so a
deployment declaring only an output tier also moved its input, cached, and
cache-write rates to that cutoff. Each component now crosses its own
*_above_<N>k_tokens_batches keys and falls back to its flat key.
The JSON schema is regenerated with the generator as it is on main:
cost-map-guard renders the PR's cost map with the base branch's generator,
so the descriptions for the new batch cache keys move to a follow-up.
* chore(proxy): restore the lazy OpenAPI snapshot to what CI's Python 3.12 generates
The merge commit carried a snapshot regenerated on a Python 3.14 venv, which dedents
docstrings at compile time, so one description line differed from the file CI regenerates
on 3.12 and the schema.d.ts sync check went red. The snapshot is byte-identical to main again
* fix(bedrock/claude_platform): strip body params the AWS endpoint rejects
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(bedrock/claude_platform): assert exact bodies through a strict fake gateway for every workspace alias and auth mode
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): block Claude Platform workspace id aliases in request bodies without admin opt-in
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Venkata Donavalli <vdonavalli@live.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: shrey kharbanda <shreshth@berri.ai>
Adds the anthropic cost map entry for claude-opus-5-5 at $4/$20 per MTok
with $5 per MTok 5m cache writes, $8 per MTok 1h cache writes, $0.20 per
MTok cache reads (0.05x base), and fast mode at 2x. The entry sets
thinking_always_on (Opus 5.5 cannot turn thinking off) and
supports_forced_tool_use false (tool_choice required/named 400s, same as
Fable 5.1), mirrors that model by omitting thinking cache preservation,
and registers the model in the setup wizard provider list
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(unit): make bedrock collector and secret scan timing tests deterministic
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(unit): count interpreter calls instead of wall clock in the secret scan scaling test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(unit): profile the secret scan with cProfile, restore the outer profiler and tighten the scaling bound
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(arize): per-team success and error sampling rates for the Arize AX callback
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(arize): fail open on invalid sampling rates and type the new sampling code
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(arize): type the sampling test helpers and parametrized fixture
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(jwt): accept a team alias in x-litellm-team-id
The header only matched canonical team ids, so a JWT caller selecting one of their teams by its alias got a 403 even though they belonged to it. The header value is now resolved through the existing alias lookup before the JWT allowed-team check and the DB membership fallback, while a value that is already a team id never costs an alias lookup and denials keep naming the value the caller sent
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(jwt): only alias a header team id the database provably lacks
Under fallback_to_db_teams a header value whose team row read fails for any reason other than TeamNotFoundError now keeps the membership denial instead of falling through to the alias lookup, so a degraded read cannot select a different team that carries the value as an alias. Drops the HeaderTeam docstring that only restated its fields
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: ryan <ryan@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): store the masked output in spend logs when a post_call guardrail rewrites the response
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): record served output without re-narrowing the logging object
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): overlay the served output before message redaction so turn_off_message_logging still wins
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(logging): type the monkeypatch fixture in the redaction ordering regression
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(e2e): poll through raw card output until the guardrail reaches the serving worker
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): keep blanked, multi-choice and disconnected served output out of raw spend logs
Served text keeps empty strings and tracks unavailable choices as None so a guardrail that blanks
the output still overrides the raw provider text. Stream choices are sized from the highest choice
index, served chunks are recorded before a client disconnect or stream failure propagates, and
message-logging redaction drops the served text from callback kwargs
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Proxy bug reports now carry the request's stream flag and a config block
built from dotted paths like router_settings.routing_strategy. A line is
emitted only when its key is defined by a LiteLLM schema and its value is
a bool or a LiteLLM-defined value (providers, callbacks, routing
strategies, cache types, guardrail integrations and modes, key management
systems). Secrets, URLs, numbers and custom values leave no line
* fix(e2e-stack): print add-mask lines only under GitHub Actions
* refactor(e2e-stack): inline the add-mask lines into main
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): answer 503 no_db_connection on management routes when the caller's user read hits a database outage
Under allow_requests_on_db_unavailable, once the caller's key row lapses the
request runs as the restricted fallback identity, and its own user read fails
on the outage. /v2/team/list, /user/list, and /user/filter/ui answered a bare
500 for that; every route now answers the same 503 body auth gives, through
one shared builder consulted by the generic exception handler and by
ui_view_users' own catch-all
* fix(proxy): log the database outage before answering 503 on user search
* chore(proxy): drop the docstrings on the db outage 503 helper and its tests
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* test(e2e): add conversational matrix across chat, messages and responses
Parameterizes one behavioral contract (reply, stream, cost log, tool call,
tool round trip) across /v1/chat/completions, /v1/messages and /v1/responses,
OpenAI and Anthropic models, and env-ref vs stored-credential auth, with
record/replay fixtures.
Adds general_settings.disable_model_info_refresh so the proxy fronting a
replay fixture does not poll every OpenAI-compatible deployment's /v1/models
in the background, which otherwise leaves unconsumed interactions in the
recorded bundle.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(e2e): force the weather tool on the first turn and rename Provider to Deployment
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: mateo <mateo@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(errors): prefilled GitHub issue link on unmapped internal errors
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(errors): scope bug report link to unmapped exceptions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(errors): cover bug report link on unmapped SDK exceptions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(errors): exercise unmapped branch without changing openai mapping
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style: format bug report changes
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): guard bug report generation for invalid exceptions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): keep bug report notice out of client error bodies
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(errors): strip only the notice separator from client messages
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(errors): only link bug reports for exceptions without a provider status
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(errors): allowlist-only bug report link, drop exception message and model from URL
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): read request.url.path for the unhandled-exception bug report route
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(bug_report): drop non-string provider values before the allowlist lookup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(bug_report): drop the OS line from the prefilled issue
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: ryan <ryan@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(cost): honor per-second custom pricing on chat completions for every provider
* test(cost): assert a per-second priced deployment bills instead of staying at $0
The zero-cost diagnostic tests from #42345 used a per-second-only entry as their
misconfigured fixture, which this branch now bills. Switch that fixture to a
per-query-only entry, which is still selected as the deployment's own pricing and
still prices chat usage at $0, and add a per-second test asserting the call
duration is billed with no diagnostic
Also let a caller's explicit total_time outrank the logging window in
completion_cost, so the SDK precedence stays stamped response, caller, logging
* test(response_metadata): move the per-second pricing regression into the mapped tests/unit file
* fix(cost_calculator): keep media-mode per-second rates off the wall-clock path
A video, transcription, speech, or realtime entry's per-second rates price media seconds, which
their dedicated cost paths bill from the media itself. The generic per-second branch now skips
those modes, so a video status poll on a per-second video model bills nothing instead of the
seconds the poll took to answer.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(otel v2): map completions, images, speech, transcription and moderation output onto the generation output
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(redaction): redact text completion choices in the standard logging payload
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(e2e): compare decoded generation output text and follow the live moderation verdict
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(otel v2): compare logged byte counts with the received media and move e2e schemas into models.py
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci(e2e): keep the otel_v2 Langfuse output e2e file out of the stage-mirror gate it cannot run in
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): mask PII in streaming /v1/messages output
Raw Anthropic SSE frames were passed through the post_call output masking
callback untouched, and ProxyLogging rerouted the callback to the unified
apply_guardrail path on /v1/messages because mask_response_content was
false. Buffer the raw frames, assemble them with the shared Anthropic SSE
helpers, mask through Presidio, and re-emit the masked frames.
Resolves LIT-8288
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): replay raw SSE frames when masking fails mid-stream
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): propagate upstream stream errors instead of returning an empty stream
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(presidio): extract buffered stream masking to satisfy complexity budget
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): let BLOCK on generated PII refuse the streaming /v1/messages response
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(presidio): fold the BLOCK re-raise into the existing except to stay within the complexity budget
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(presidio): move the blocked stream consumption into a helper so pytest.raises holds one statement
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(presidio): fail closed when output masking of a raw SSE stream errors
A Presidio outage on streaming /v1/messages replayed the unscanned frames
to the caller. Propagate the error instead, matching the non streaming
path and the merge base
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(presidio): cover structured chat stream output masking and trailing bytes passthrough
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci: benchmark and gate an installed release wheel
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci: simplify installed-wheel benchmark check
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(rust): add native tokenizer codec
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(tokenizer): route Python tokenization through the Rust extension
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style(lint): format tokenizer call
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(packaging): restore runtime dependencies and native images
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(tokenizer): preserve Python SDK behavior with Rust tokenizers
* fix(tokenizer): restore compatibility paths
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(tokenizer): count custom tokenizers directly
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(tokenizer): preserve caller-supplied Python tokenizer counts
* fix(tokenizer): reuse packaged vocabularies in the native wheel
* refactor(rust_bridge): route token counting through the catalog as RUST_OPT_IN
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(spend_tracking): compare tokenizer groups by value
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(deps): re-resolve filelock under the <4.0 pin
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(llms): align transformation override signatures with base configs
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* build(rust): use fat LTO to keep the native wheel under the 35 MB limit
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(tokenizer): preserve Python defaults with opt-in Rust dispatch
* test(proxy): tolerate missing litellm.utils.Tokenizer when patching it
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): patch the tokenizer dispatch function instead of the removed alias
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(tokenizer): give the Rust wrappers the tiktoken and tokenizers surface
Callers of litellm.encoding and litellm.create_tokenizer must see the same
read-only API whichever backend the catalog selects.
- OpenAIEncoding mirrors tiktoken.Encoding: n_vocab, max_token_value,
token_byte_values, encode_single_token, encode_with_unstable,
encode_to_numpy, decode_with_offsets, is_special_token, repr; the Rust
tiktoken crate keeps a Vocabulary beside each CoreBPE and reports the
requested encoding name (gpt2 stays gpt2).
- HuggingFaceTokenizer mirrors the read-only tokenizers.Tokenizer surface
(token_to_id, id_to_token, get_vocab, get_vocab_size,
get_added_tokens_decoder, num_special_tokens_to_add, padding, truncation,
encode_special_tokens, from_buffer); HuggingFaceEncoding gains the
char/word/token lookups, pad, truncate, set_sequence_id and merge.
Mutators stay on the Python tokenizer.
- from_json/from_pretrained claim the fork gate only when the huggingface
feature is compiled in; the surrogate fallback matches on the Codec.
- Tokenizer caching is keyed on the same catalog Context the dispatch runs
on; rust_tokenizer reads the encoding name without loading an encoding;
LITELLM_RUST parsing is cached.
- Drop the unused tiktoken_encoding_for_model export and Error::Download.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* fix(tokenizer): close the exhaustive matches with assert_never
CodeQL reads a `match` over a Literal with no default arm as an implicit
`None` return. `assert_never` makes the exhaustiveness explicit for both the
HuggingFace tokenizer loader and the Rust token-counter factory.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* feat(tokenizer): derive the fast counter from the shared tokenizer
The count-only counter (`fast` feature) and the codec each parsed the same
artifact: TokenCounter took the Anthropic JSON and the tiktoken rank files
from Python while Tokenizer loaded them again. One parse now serves both.
- FastTokenizer builds from a model another loader holds: `from_shared`
takes the Arc<tokenizers::Tokenizer> the HF codec keeps, and
`from_*_pairs` take the ranks the tiktoken vocabulary already parsed.
- `FastCounter::fast_counter` in the core crate derives it from either codec;
encodings the fast scanner does not reproduce are refused.
- Native `Tokenizer.count(text, fast=False)` opts into that counter, built
once per tokenizer on first use; `TokenCounter.from_tokenizer(tokenizer,
fast=False)` replaces the JSON and rank-file constructors.
- The Python route counts over the native tokenizers the codec path shares
(`native_encoding`, `native_anthropic`) and no longer reads rank files;
the packaged Anthropic tokenizer has one loader, `tokenizer_dispatch.anthropic`.
- Public wrappers gain `count(text, fast=False)`.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---------
Co-authored-by: Yujong Lee <yujong@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* feat(proxy): add TinyFish Agent API passthrough with per-step billing
* chore(ui): regenerate dashboard API types for /tinyfish passthrough
* fix(proxy): satisfy strict lint budget for tinyfish passthrough
* style: ruff format tinyfish passthrough handler
* test(proxy): exercise tinyfish route through the app with a faked upstream
* refactor(proxy): make cross-module tinyfish billing hooks public
* fix(proxy): tolerate transient tinyfish poll failures instead of dropping the charge
* fix(proxy): defer billing for disconnected tinyfish SSE runs to the background poller
* Revert "fix(proxy): defer billing for disconnected tinyfish SSE runs to the background poller"
This reverts commit ef0bcfb4a0.
* fix(proxy): bill tinyfish SSE runs via detached poller and only COMPLETED runs
Disconnected run-sse clients previously left completed runs unbilled: the stream-end
handler saw a still-RUNNING run and logged $0. The poller now spawns from the streaming
path on the first run_id frame, outlives the disconnect, and writes the one spend row when
the run turns terminal; the stream-end path only logs the $0 fallback for run_id-less
streams. Costs now apply only to COMPLETED runs ($0 for FAILED/CANCELLED, matching the
upstream invoice), spend rows carry the request's litellm_call_id (previously NULL
request_ids collided and were silently dropped), and the GET /v1/runs listing is blocked
so callers behind the shared key cannot discover each other's runs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* fix(proxy): drop GET /v1/runs from the tinyfish allowlist error message
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* chore(proxy): sync openapi artifacts for tinyfish docstring, suppress LIT011 on flag write
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* style(proxy): ruff-format the sse poller flag write
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* style(proxy): keep the rebind-ok suppression on the flag write's own line
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* fix(proxy): harden tinyfish billing paths from review findings
Skip failure dispatch when the SSE poller owns billing (a failure row collided with the
poller's billed row on request_id and dropped the charge), late-spawn the poller for
run_ids that arrive in unterminated frames instead of mispricing RUNNING runs at $0,
thread litellm_params into poller-billed standard logging objects so SLO consumers see
attribution, untype the run error field so upstream error-shape drift cannot void a
billable run, normalize a schemeless TINYFISH_AGENT_API_BASE, extend the poll budget to
cover queue wait (3600s) with ~60s outage tolerance, and log poller cancellation.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* chore(proxy): satisfy ratcheted BLE001/LIT002 budgets from main in tinyfish handler
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* style(proxy): drop stray blank line from merge resolution
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* fix(proxy): reject passthrough envelope controls on tinyfish route, raise blocking-run timeout
The generic passthrough unwraps a caller-supplied custom_body as the forwarded request and
honors a caller stream flag, so custom_body.use_vault bypassed the credentialed-run 403 and
stream: true flipped a blocking run into the streaming pipeline. The route now 400s the
envelope fields (custom_body, stream, query_params); streaming comes from the endpoint.
Blocking runs also get a 1500s default timeout covering the upstream 1200s run cap, unless
the operator configured pass_through_request_timeout.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* style(proxy): resolve operator timeout without a dict-literal default
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Pxz11TQKYvCXvKawNUa2g4
* test(passthrough): list the TinyFish route among protocol-constrained pass-through routes
* chore(proxy): regenerate the lazy OpenAPI snapshot after merging main
* chore(proxy): keep the lazy OpenAPI snapshot as CI's Python 3.12 renders it
* fix(proxy): reject TinyFish POST bodies that are not a JSON object
A form-encoded or text body carried stream and use_vault past both field
gates, because the gates only saw fields the body parsed to as JSON. The
route now checks the content type before reading the body and answers 400
for anything that is not a JSON object.
* fix(tinyfish): reject submit paths with extra slashes so run-async always bills
The allowlist dropped empty path segments, so POST /v1/automation/run-async/
was forwarded upstream while the billing dispatch only recognises the exact
path and would have logged the submit at $0 without starting the poller.
Any path with a trailing or doubled slash now returns 403 before forwarding.
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(anthropic_adapter): keep reasoning_effort a string for targets that stay on chat completions
* fix(anthropic_adapter): judge the summary bridge with the deployment's api_base
The adapter's bridge check now resolves the provider and base the same
way completion() does, passing the deployment's api_base and api_key
into get_llm_provider and the resolved base into the bridge check, so a
Foundry deployment lands on the same route in both places and a bare
model name routed by its api_base still gets the plain tier.
A litellm_proxy target keeps the dict, since the upstream gateway makes
its own bridge decision and needs the summary to make it.
* refactor(anthropic_adapter): return the plain effort instead of writing it inside the helper
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* test(pricing): lock supports_prompt_caching on cache-priced grok rows
Check the catalog field itself so a price sync cannot drop the Vertex
Grok flag while the helper still passes via the xai/ fallback. Also
cover get_model_info on the full vertex_ai/xai/grok-4.6 key and keep
the backup map in lockstep.
Co-authored-by: Techboy bebop <kumarpriyanshu09@users.noreply.github.com>
* test(pricing): assert grok cache rows via get_model_info
The helper can still pass via the bare xai/ grok row, so lock the
catalog flag through get_model_info instead of reading the JSON files
Co-authored-by: Techboy bebop <kumarpriyanshu09@users.noreply.github.com>
* test(pricing): assert grok cache flags via get_model_info
Address Greptile P2 by checking supports_prompt_caching through
get_model_info on full catalog keys instead of raw JSON fields, so
the check cannot pass via the bare xai/grok-* helper path. Keep
backup/primary parity for vertex_ai/xai/grok-* entries.
* test(pricing): assert grok cache flags via runtime APIs only
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Techboy bebop <kumarpriyanshu09@users.noreply.github.com>
get_user_object wrapped every failed read, a refused connection included, in
ValueError("User doesn't exist in db ..."), so JWT callers got a 401 naming a
missing user while Postgres was down and virtual-key callers got 503
no_db_connection for the same outage. A connection or transport error now
propagates as-is and the auth exception mapper answers 503 no_db_connection;
a genuinely missing row and query-level errors still answer 401.
The MCP auth and token-exchange docstrings and the exception-chain helper's
docstring described the old wrap and are updated to the new contract.
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* docs(rust): plan Python interop foundation
* fix(rust): preserve Python settings coercion at the native boundary
* chore(rust): drop interop planning note
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(rust): resolve OCR provider secrets through an async SecretSource before transformation
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(rust): project the Python secret manager into the bridge and resolve OCR secrets through it
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(rust): drop premium_user from the secret manager snapshot
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(rust-bridge): read the private key management globals once in the settings snapshot
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(rust): bound the bridge secret manager state cache to the active snapshot
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(rust): inline coercion unit tests
* fix(rust): preserve Python secret manager bindings
* refactor(rust-bridge): let settings projectors own their contract specs
Each settings group now declares its SettingSpec rows next to the projector
that reads them, and the manifest test derives python_settings.json from those
tables instead of a hand-copied duplicate. Field carries (group, name) instead
of a dotted path, and coercion gains the dict-item reader plus the Redis
Boolean, certificate-requirement, non-empty string, and numeric adapters that
the cache configuration projection adopts next.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(rust-bridge): capture the secret manager binding in one settings read
The secret_manager accessor now carries the live client and settings objects,
so the bridge classifies the binding from a single snapshot instead of
re-reading litellm globals. The unreachable native arm and the service alias
go away, the binding-to-state mapping moves next to the snapshot, and the
Python callback precomputes its key_manager name.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(rust-bridge): execute typed settings field declarations
* refactor(rust-bridge): compare cache backends by identity behind one exact trait
cache-response gains an object-safe ExactResponseCache so every exact-match
backend sits behind one pointer; WriteBuffer flushes through it. The bridge's
NativeResponseCache shrinks from nine variants and fifteen per-backend
accessors to an exact service plus the three semantic backends, and facade
mismatch detection compares BackendIdentity values instead of matching on
each backend type. Request projections move next to NativeRequest.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* refactor(rust-bridge): drive both Python-embedded semantic caches through one execution
Redis-semantic and Valkey-semantic operations now share one SemanticExecution
body: await the Python embedder, seed the task-local vector, run the native
backend, repeat per batch entry. Valkey drops its with_embedder path in favor
of the same seeded embedder, and each backend keeps its own embedding-failure
policy. PythonEmbedder exposes one call shape. Redis-semantic thresholds are
compared at the backend's f32 width, which un-breaks the redis-stack parity
tests that a 0.8 facade threshold failed before this branch.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* wip
* feat(rust-bridge): complete response cache runtime surface
* fix(rust-bridge): preserve secret manager callback exceptions
* refactor(rust-bridge): unify route cache and secret rollout catalog
---------
Co-authored-by: Yujong Lee <yujong@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
The gpt-5.6-responses_previous_response_id case sent a literal id the proxy never issued, which the Responses id security hook refuses with a 403 at production defaults. The case now primes a response through the proxy and chains the id it hands back, so the harness drops allow_unmanaged_response_ids and the security hook stays exercised
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* test(e2e-ui): check the MCP Tools tab against the upstream's own tools/list
DeepWiki renamed ask_question to ask_wiki_question, and the spec hardcoded the old name, so
e2e_ui_testing went red on main for something that is not a litellm regression. The spec now asks
the upstream server for its tool list with the official MCP TypeScript SDK and expects the tab to
show exactly those cards, so a vendor rename cannot turn the job red again.
* test(e2e-ui): cite the pinned DeepWiki tool name and drop the helper docstring
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(responses): drop client_metadata before bridging to chat completions
Codex CLI sends client_metadata on every /v1/responses call. For a
provider with no native Responses config the chat-completions bridge
forwarded the raw kwargs, so client_metadata reached the provider as a
chat body field and Databricks rejected the request with an unknown
field 400. The bridge now drops the Responses-only request fields
before calling completion while still passing every other kwarg
through, so deployment-level params such as chat_template_kwargs keep
reaching providers without a native config.
* fix(databricks): merge consecutive system messages for chat-template models
Codex sends instructions plus a leading developer item, which the Responses
bridge and the developer-to-system translation turn into two consecutive
system messages that Databricks chat-template models reject with "System
message must be at the beginning". Each run of consecutive system messages
is now merged into one before the request is built for non-Claude models.
Also keep client_metadata out of the bridged chat request even when
allowed_openai_params names it, so both bridge branches drop the same set.
* fix(databricks): skip empty system messages when merging consecutive ones
Databricks drops empty content before the merge, so a system message in a
run could carry no content key and the merge iterated None. Those messages
are now skipped; a run with no content at all keeps its first message.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* refactor(agentic-loop): build follow-up kwargs in one place so no executor can repeat a request param
The Responses and both chat completions follow-up executors each rebuilt the follow-up kwargs by hand and then expanded them next to the request params, so a plan whose kwargs repeated a request param raised a duplicate keyword TypeError. They now share build_agentic_followup_kwargs, which drops any key already sent as a request param (and the explicitly passed model/input/messages) from both the request kwargs and the plan kwargs. Each executor keeps its own internal-key filter unchanged, and the /v1/messages executor is untouched because it merges into a single dict and cannot hit this.
* test(agentic-loop): move follow-up regressions into their mapped test files
Greptile review: the executor regressions belong in test_llm_http_handler.py and test_chat_completion_agentic_loop.py rather than a split-off file, and the builder test helper returned a read-only mapping while promising a dict. The Responses overlap test is dropped because #41560 already added the same one to the mapped file.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): release unclaimed budget reservations at request end
* fix(proxy): release unclaimed budget reservations of websocket sessions too
* test(proxy): drop the structural middleware inheritance check
* fix(proxy): claim the budget reservation on streaming pass-through before its cost callback
The SSE chunk processor hands its success handler to the logging worker
after the response, so the request-end release freed the reservation
first and left the key unguarded until the worker drained. Claim it at
both end-of-stream hand-offs, the immediate enqueue and the coroutine
parked for deferred dispatch.
Give the xai realtime test double the litellm_params attribute every
real Logging object carries, since the wrapper now reads it.
* test(pass-through): give the vertex streaming test doubles a litellm_params dict
The spec'd Logging mocks in test_vertex_ai_anthropic_streaming_cost_injection.py
lacked the instance attribute the chunk processor now reads to claim the budget
reservation. Also restores main's _lazy_openapi_snapshot.json: the branch's copy
had been regenerated under Python 3.14, which dedents one docstring description
that the CI regeneration on Python 3.12 keeps indented, and the PR adds no lazily
loaded route, so main's file is the correct one.
* fix(pass-through): claim the budget reservation only after its cost callback is enqueued
Every pass-through success hand-off stamped callback_bound before handing the
coroutine to the logging worker. When that enqueue raised, the reservation stayed
claimed with no callback left to reconcile it, so the request-end release skipped it
and the reserved cost stayed pinned on the key's counter. Enqueue first, then claim,
so a failed hand-off leaves the reservation for the request-end release.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* feat(cost): warn and count $0 cost on billable requests
A request that carries usage but prices to $0 on a model whose pricing
entry has a non-zero rate now logs one warning naming the model, the
pricing entry, and the missing rate, and increments
litellm_zero_cost_requests_total{requested_model, model, model_id,
api_provider, reason}. Free models (every used rate is 0), requests
without usage, and unmapped models stay silent. The diagnostic rides on
the standard logging payload as zero_cost_diagnostic
* fix(cost): keep the zero-cost diagnostic importable on 3.10 and recursion-free
* fix(cost): warn once per request when a $0 result is priced again
* fix(cost): judge a free deployment by its own pricing and keep it silent on calculator errors
* fix(cost): warn once per request when a usage-less evaluation sits between two zero-cost findings
* fix(cost): judge zero-cost findings by the priced entry, skip cache hits, count failure rows
* test(cost): type the zero-cost diagnostic test helpers
* test(logging): flag a $0 terminal Responses stream event by its inner response
* chore: restore the lazy OpenAPI snapshot as CI's Python 3.12 generates it
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* fix(proxy): drop cost-map metadata echoed back on model save
Filter unchanged cost-map fields from model-info save echoes while preserving edited overrides.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): drop a stored override when an echoed save resets it to the cost-map value
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): compare model_info echo against the deployment's cost-map lookup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): decrypt the stored model before the cost-map lookup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): treat a reset to the bundled catalog value as an echo even after router registration
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): compare the reset against the catalog as loaded, not only the bundled backup
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(types): type the catalog snapshot and echo filter parameters as Mapping[str, object]
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): inject the loaded catalog into update_db_model instead of patching the class
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): use contextlib.suppress for cost-map lookup miss to stay under BLE001 budget
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: ryan <ryan@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): price non-canonical image sizes from the nearest row and honour dump options
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(fal_ai): drop monkeypatched mixed pricing case
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(fal_ai): use the default dimensions constant directly
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): forward nested include and exclude when dumping image data
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): honour pydantic item selectors in image data serializer
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): match negative item selectors in image data serializer
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(proxy): opt-in litellm_call_id in JSON error bodies
Add general_settings.include_call_id_in_error_body. When true, the value
already on the x-litellm-call-id response header is copied into JSON error
bodies: as error.litellm_call_id on the OpenAI-shaped routes, /v1/messages,
and streaming first-chunk errors, and as a top-level litellm_call_id on
pass-through routes. Off by default, so error bodies stay byte-identical
unless an admin opts in
* chore(proxy): drop helper docstring and restore lazy OpenAPI snapshot
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* feat(fal_ai): add flux-lora-depth image edits and moondream3 chat completions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(fal_ai): retrigger codecov processing
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): reject multi-turn and system messages for moondream3 chat
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): return 400 for invalid moondream3 chat requests
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): reject moondream3 responses missing output or usage
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fal_ai): reject streaming moondream3 requests before dispatch
stream never reaches optional_params, so the transform_request check could not fire; reject in _complete_fal_ai on ctx.stream
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(bedrock): sign batch S3 requests with s3_access_key_id and s3_secret_access_key
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(bedrock): keep S3 signer test additions scoped to new cases
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(bedrock): drop e2e suite changes from the S3 signing fix
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(bedrock): build S3 credentials directly from the s3_* pair so ambient AWS_* env never mixes in
Restores the split-identity e2e coverage
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
- fix(bedrock): send every Mantle beta in the anthropic-beta header on the bedrock/mantle route
- refactor(bedrock): type the Mantle header helper and build the header fields in one comprehension
