* fix(logging): pass provider response headers to callbacks on every endpoint
Custom callbacks only received kwargs["response_headers"] for chat
completions. Responses, image generation and edit, speech, and
transcription calls either never recorded the provider's headers or
recorded them in one place and not the other.
Every handler now records the provider's httpx headers on the response's
hidden params as "headers" (raw) and "additional_headers" (processed,
with LiteLLM's own entries winning on a clash), and the logging object
derives model_call_details["response_headers"] from those hidden params
before cost calculation on the non-stream and both streaming success
paths, keeping a handler-set value authoritative. Binary speech responses
expose their hidden params to the standard logging payload, and the sync
OpenAI transcription request always fetches the raw response.
* test(images): point the legacy image and speech fakes at the raw response surface
Image generation now goes through the SDK's raw response so the provider headers can be read, and the speech binary response now carries hidden params. The unit fakes in the image generation, xinference, proxy provider, image edit, Vertex speech, and otel suites still pinned the old call surface and the old "no hidden params" assertion, so they read an uncalled mock or a fake response without headers.
* test(images): drop the rewritten mock comments and the generated edit PNGs
* test(images): move the llm-span test's image fake to the raw response surface
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* feat(vertex): native batch JSONL passthrough with cost tracking
Add a per-request `passthrough=true` multipart field on `POST /v1/files`
(and the same kwarg on `litellm.create_file`) that uploads a native
Vertex AI batch JSONL to the deployment's GCS bucket unchanged, so rows
using `googleSearch` and other Gemini-only features run as written and
the output, `groundingMetadata` included, comes back untouched.
Passthrough is sticky through the GCS object path
(`litellm-vertex-files/passthrough/...`), so batch create and output
retrieval inherit it without new state. Native output rows are costed
from their `usageMetadata` with the deployment's model and model_info,
in the polling and retrieve paths and for the existing global
`disable_vertex_batch_output_transformation` flag, which billed $0
before.
The proxy requires the target to resolve to vertex_ai deployments only,
refuses `passthrough` with a non-batch purpose, a non-default
`target_storage`, or pre-call guardrails, and validates native rows on
`request` instead of the OpenAI batch keys.
* refactor(vertex): keep native batch row pricing inside the Vertex adapter
Moves native Vertex batch row detection, response parsing, and per-row
pricing from litellm/batches/batch_utils.py into
litellm/llms/vertex_ai/batches/transformation.py, so batch_utils only
aggregates the rows it gets back. Adds tests/test_litellm/files to the
misc unit shard so the new test directory is claimed by a shard.
* fix(files): say what a passthrough batch upload takes when a row is not native
The missing-key 400 listed bare key names, so an OpenAI-shaped row under
passthrough=true read "Each line must be a JSON object with keys request".
The batch line shape now carries its own hint, and the passthrough one says
a passthrough upload takes native Vertex batch rows with a request key
* fix(batches): bill native Vertex embedding batch rows on the native cost path
A native Vertex output row whose response holds an embedding was validated as a
generateContent response, so the documented tokenCount-only shape counted as a failed
row. Price embedding rows from their own usage (promptTokenCount, else tokenCount) with
the helper the transformed embeddings path already used, and drop the prompt-details
helper nothing calls anymore.
* fix(batches): keep modality batch rates on native Vertex embedding rows
An embedding row that carries usageMetadata was billed from promptTokenCount alone, so
its promptTokensDetails no longer reached the audio, image, and video batch rates the
way it did before the native cost path. Run every row with usageMetadata through the
Gemini usage parser and keep the flat tokenCount fallback for embedding rows without it.
* fix(batches): price native Vertex batch rows by modelVersion under a wildcard deployment
A `vertex_ai/*` deployment hands the batch cost path `*` as the deployment model, which
no cost map resolves, so every native (passthrough or flag-on) row was billed at $0. A
wildcard deployment model now defers to the row's own `modelVersion`, the way the
transformed path already prices by the row's `model`.
Also moves the native passthrough tests under tests/test_litellm, the tree codecov
reads, and covers the raw upload chunking, the embedding output translation, the
unpriceable-row path, and the flag-on dispatch.
* fix(batches): keep explicit deployment prices for native Vertex rows without a modelVersion
Under a wildcard deployment a native batch row that carries no modelVersion (an embedding
row, or a generateContent row Vertex returned without one) was billed at $0 even when the
deployment's model_info sets explicit batch prices, because the cost calculator was never
called. The row now falls back to the wildcard name, which the cost calculator prices from
the explicit model_info, and only a row with neither a modelVersion nor a deployment model
is billed at $0 with the warning
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
The new CircleCI tests pipeline (#42773) runs tests/unit under pytest-cov on
CPython 3.12.2, where coverage traces every line through sys.settrace. The two
tracemalloc peak comparisons in test_vertex_ai_files_streaming.py drive 8000-row
payloads through both pipelines and slow from ~10s to over 3 minutes under that
tracer, so both hit the 90s pytest-timeout on every run.
Mark them no_cover so pytest-cov pauses tracing for just these two. Their
assertions are unchanged and every other test in the file still reports coverage.
* test(integration): add read-replica routing harness
* refactor(integration): hoist the maintenance url imports
* fix(integration): keep per-test databases and the witness sequence readable under replica roles
* fix(integration): opt bespoke database and pool tests out of the injected read replica
* test(integration): commit recorded replica routing expectations
* fix(integration): judge routing by role containment so shrinking role sets do not fail
* fix(integration): run the pool-limit shutdown choreography on the superuser database url
* ci(integration): add the mcp group to the replica matrix
* fix(integration): judge routing by exact role sets with a named either-role allowlist
* test(integration): drop containment-era routing expectations for re-recording
* chore(integration): drop docstrings from the replica harness scripts
* docs(integration): describe exact routing matching and the either-role list
* test(integration): record exact replica routing expectations
* test(integration): allow the SELECT 1 health probe on either role
* test(integration): replace committed routing expectations with an on-demand base-vs-head parity run
* test(integration): fix parity env scope, readme wording, and seed-deterministic serialization test
* test(integration): make the sorted-role serialization test deterministic in-process
* test(integration): swap all product code in parity runs and pin role gains
* ci(integration): force tracked-file removal before parity checkout
---------
Co-authored-by: yuneng <yuneng@berri.ai>
* fix(fireworks_ai): route firerouter short names and bill pass-through legs at the routed model's rates
fireworks_ai/firerouter and fireworks_ai/firerouter/<slug> resolve to
accounts/fireworks/routers/... instead of a models/ path, and the cost
calculator falls back to the routed model's own catalog entry before the
Fireworks size buckets so a Claude leg is no longer priced at $0
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fireworks_ai): bill routed legs under the routed model's own provider
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(fireworks_ai): require the k suffix when parsing tiered input fields
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(params): carry stream_chunk_size through litellm_params instead of provider params
* test(integration): fence stream_chunk_size out of every provider request body
* test(bedrock): type parametrized stream chunk test params
* test(integration): drop the contracts manifest resurrected by the main merge
* test(bedrock): type the stream_chunk_size test helpers
* test(params): finish AGENTS.md typing pass on stream_chunk_size tests
* test(integration): drop the covers marker from the stream_chunk_size wire test
---------
Co-authored-by: shrey kharbanda <shreshth@berri.ai>
* fix(bedrock): send json_schema as a forced tool on Claude Opus 4.7 and 4.8 Converse
Bedrock rejects outputConfig.textFormat on Opus 4.7 and 4.8 with
"output_config.format: Extra inputs are not permitted", and the AWS
model cards list structured outputs as not supported for both, so
their cost-map entries no longer claim supports_native_structured_output
and json_schema requests fall back to the json_tool_call tool.
Fixes#27846
* test(bedrock): assert Opus 4.7 and 4.8 inline the schema on Invoke, move the native case to Sonnet 4.6
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* test: point CircleCI-only suites at models still in the cost map
#42435 removed cost map entries past their deprecation date and #42437 added
litellm_uisettings to the config-synced tables, but both only updated
tests/test_litellm. The CircleCI-only suites (local_testing, llm_translation,
logging_callback_tests, litellm_utils_tests, unit) kept using the removed
models or the old table list and went red on main.
Each test keeps its assertions and swaps the removed model for a current one
with the same provider and capabilities. The fireworks tests pick a vision
model from the cost map because #34941 set supports_vision false on
minimax-m3, and the vertex image provider test injects the image model set
because #42435 removed every vertex_ai-image-models entry.
* test(vertex_ai): register the image model through add_known_models in the provider test
* chore(cost-map): remove models past their deprecation date
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(cost-calc): drop the empty parametrize left behind by the gemini web search removal
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(cost-map): drop merge base block left by conflict resolution
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(cost-calc): drop gemini image cost tests pinned on removed model
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(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
* 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>
* 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>
* 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>
* 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>
Cherry-pick of merge commit b3882d8e43 (PRs #39321, #39562, #40107), which landed on litellm_internal_staging instead of main.
Co-authored-by: ojensen-berri <ojensen@berri.ai>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
