add_provider_specific_params_to_optional_params built the extra_body dropped-key
set as frozenset(additional_drop_params), so one non-string entry raised
TypeError: unhashable type: 'list' and every openai-compatible call carrying one
failed with a 500 before it reached the transport. The set now takes only the
string entries, the element type every other signature in this chain already
declares as list[str].
A list-form entry still drops nothing: is_nested_path() tests a string, so
delete_nested_value() has never applied one on any provider. This removes the
crash only, so a working string path such as "tools[*].function.x" sitting
beside a malformed list entry is applied instead of taking the request down.
* 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(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>
Register s3_access_key_id, s3_secret_access_key and s3_encryption_key_id as
LiteLLM-owned batch params so they are no longer forwarded to Bedrock as
additionalModelRequestFields (which 400s ordinary chat on a batch-configured
deployment), keep them on CredentialLiteLLMParams so the batch/file paths
still receive them, and redact the S3 credential key names in debug logs.
Resolves LIT-8290
Co-authored-by: yucheng <yucheng@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A caller-supplied extra_body model overrode the authorized model in the request the shared HTTP handler sends upstream. Strip it before dispatch
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Register s3_access_key_id, s3_secret_access_key and s3_encryption_key_id as
LiteLLM-owned batch params so they are no longer forwarded to Bedrock as
additionalModelRequestFields (which 400s ordinary chat on a batch-configured
deployment), keep them on CredentialLiteLLMParams so the batch/file paths
still receive them, and redact the S3 credential key names in debug logs.
Resolves LIT-8290
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Mantle serves anthropic.claude-haiku-4-5 without the dated -20251001-v1:0
suffix the Bedrock row carries, so the native route billed it at 0. Add a
bedrock_mantle/anthropic.claude-haiku-4-5 row and let a
bedrock_mantle/<region>/<model> name fall back to the region-free
bedrock_mantle/<model> row before the provider-prefixed lookup. Also
satisfy the mutable-collection gate in the native messages transformation.
The azure row of test_get_model_info_falls_back_from_dated_snapshot_to_undated_entry used gpt-5.6-luna-2026-07-09, which main's cost map carries as an exact azure key, so the lookup returned the dated key and the required misc test job failed on main. All three dated snapshot tests now use a 2099-01-01 snapshot date, so they keep exercising the strip path whatever real snapshots the map gains
DeepSeek charges half the listed rate outside 01:00-04:00 and 06:00-10:00 UTC
Monday to Friday, so every deepseek-flash, deepseek-v4-flash,
deepseek-v4-flash-vision-exp, and deepseek-v4-pro entry now carries an
off_peak_pricing block with those windows and the halved input, output, and
cache-hit rates. The generated cost map schema picks up the block, and the
regression tests pin the peak and off-peak cost of one call at fixed moments.
Fifty six of the deleted tests turn out to assert the output of litellm code rather than the catalog lookup itself, things like map_openai_params, get_supported_openai_params, should_fake_stream, transform_request bodies, cost_per_token arithmetic, get_llm_provider routing, and provider config dispatch. They only happen to read shipped entries as inputs, so they belong in the later rewrite that injects a local model_cost, not in this deletion
Each one is restored verbatim from origin/main along with the fixtures, helpers, constants and imports it needs, and tests/test_litellm/test_sambanova_model_metadata.py is restored wholesale
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Regenerated every touched file from origin/main applying only the B1 test deletions and the unused import and helper cleanup they leave behind, without running the formatter across untouched code. CI only checks ruff format under litellm/, so the earlier reflows of test files were pure diff noise for reviewers
Also drops the tests/local_testing/test_prompt_caching.py entry from the caching-local shard in test-unit.yml since that file is deleted
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The repo rule is that a test must only fail when litellm code changes, never when a vendor updates a price, renames a field, or drops a model. These tests asserted shipped catalog entries directly, comparing lookup results to literals copied from model_prices_and_context_window.json or requiring named entries to exist or be absent, so every cost map sync could break them without any litellm code changing
Tests that exercise real litellm behavior with an injected local model_cost, invariants like backup parity, and assertions on non-lookup code paths are untouched
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
The Responses id security hook keeps the id a client addressed under
`_litellm_addressed_response_id` in the request body so internal retries can
re-authorize it. On a model without a native Responses config that body is
bridged into `completion()` kwargs, the key was treated as a provider param,
and providers rejected it, so every follow-up turn carrying
`previous_response_id` returned 400.
Register the key in `all_litellm_params` so it is dropped before any provider
request, and share one constant between the hook and the param list.
The deepgram/streaming/* rows added for the Deepgram WebSocket passthrough declare /v1/listen as their endpoint, so the registry validation test needs it in the enum, the same way /vertex_ai/live was added for that passthrough
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Keeps the original CustomStreamWrapper so response headers and the
correlation-context cleanup in __del__ are untouched when a deployment
hook rewrites the converted response. Covers the early-return branches
for real provider streams and unmapped call types
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A pre-call deployment hook can turn a requested stream into a non-streaming provider call and the result is wrapped back into a fake stream. The async client wrapper treated that wrapper like a caller-requested stream and returned before async_post_call_success_deployment_hook, so SDK callers lost post-call deployment processing (including CustomGuardrail post_call enforcement) on converted streams. Run the hook on the complete ModelResponse behind the wrapper and rewrap a modified response so it reaches the emitted chunks.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
