* fix(batches): apply the Vertex regional endpoint uplift to batch cost
batch_cost_calculator takes vertex_location and multiplies by the model's
regional_endpoint_uplift_multiplier through get_vertex_regional_endpoint_uplift,
and every batch cost path (aretrieve_batch pricing, completed-batch logging,
native Vertex batch rows, the proxy batch cost poll and its per-line items)
resolves the deployment's location and passes it down
Co-authored-by: Mrun25 <165297479+Mrun25@users.noreply.github.com>
* test(logging): point the resolver tests at resolve_vertex_location_for_cost
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Mrun25 <165297479+Mrun25@users.noreply.github.com>
Backport of #47466 to stable/1.105.x.
Cherry-picked from c6c303a618 (main).
* fix(anthropic): apply tool_config cache-control points on chat and /v1/messages
A cache_control_injection_points entry with location tool_config was spread into
the Anthropic request body on /v1/chat/completions, which Anthropic rejected with
400 Extra inputs are not permitted, and the native /v1/messages path dropped the
point silently. Both paths now put the point's control on the last tool Anthropic
caches and never forward the key.
Co-authored-by: Tan Nguyen <tan@pressingly.net>
* refactor(anthropic): validate tool_config inputs instead of casting them
Keeps the lint gates flat: the helper takes untyped tools and points and
narrows them with the existing object validators, the chat transform builds
the body with the marked tools instead of writing into optional_params, and
the handler reads a typed view of kwargs.
* fix(openai): keep tool_config cache points out of extra_body, add integration test
The cache hook hands a tool_config point on for the provider step, and for
providers that call through the OpenAI SDK every non-OpenAI param rides in
extra_body, so OpenAI answered 400 Unrecognized request argument supplied.
The new integration file drives a live proxy against scripted Anthropic,
Bedrock Converse and OpenAI upstreams for /v1/chat/completions and
/v1/messages.
* test(integration): cover tool_config cache points on bridged /v1/messages
Three new cases on /v1/messages for models without a native Anthropic
messages path: OpenAI through the Responses bridge and Gemini through the
chat bridge never see cache_control_injection_points in the body, and
Bedrock Nova through the chat bridge gets the cachePoint on toolConfig
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Tan Nguyen <tan@pressingly.net>
Backport of #47394 to stable/1.105.x.
Cherry-picked from 93b79d2587 (main).
Backport of #47419 to stable/1.105.x.
Cherry-picked from 3124797399 (main).
Adapted to the line, on top of the ported helpers in the previous commit:
- runtime_model_cost_registrations reads a per-cost-map store of the fields each register_model call set (_RegisteredPricingFields), which stands in for main's AliasedCostEntry.registered_fields (#41523) so alias entries stay plain dicts. A catalog reload resets it, as main's reload drops registered_fields.
- An inference-profile ARN in model_id is resolved to its profile name inside _bedrock_pricing_info with the line's extract_model_name_from_bedrock_arn, instead of main's get_potential_model_names ARN handling.
- completion_cost keeps the line's candidate list and passes the profile info through; optional_params is read directly (main's _LOGGED_LITELLM_PARAMS is not on the line).
- The router adapters _MODEL_INFO_ADAPTER and _LITELLM_PARAMS_ADAPTER are defined here.
- Tests use the line's private names. Dropped: the AliasedCostEntry routing test, the router cache-only pricing test (main-only fixtures), and the ignore_stamped_cost assertion (main-only parameter).
Backport of #43935 to stable/1.105.x.
Cherry-picked from d7620c77a1 (main).
Adapted to the line: the Responses bridge change lands in litellm/main.py (the bridge check moved to litellm/responses/bridge_check.py later on main), the line's _get_effort_level and _supports_reasoning_effort_level names are kept, and the test files keep the line's helpers with only this PR's hunks applied.
Backport of #43955 to stable/1.105.x.
Cherry-picked from eb5295b20c (main).
Handler import conflict resolved by keeping the line's imports plus aload_beta_headers_config; the CLI-beta wire assertion change is dropped since that test is not on the line.
Backport of #45686 to stable/1.105.x.
Cherry-picked from 1aee1a21ed (main).
Test-file import conflict resolved by keeping the line's imports plus this PR's own.
Backport of #45659 to stable/1.105.x.
Cherry-picked from fa7ad80a62 (main).
Test-file conflict resolved by keeping the line's file plus this PR's own additions.
The #46342 backport carried block_binding through every subclass of the
Anthropic and Bedrock Invoke configs, so Bedrock Mantle, Claude Platform and
OpenAI-compatible Anthropic endpoints started sending it. Each of the six chat
and messages configs now opts in by exact class, and the Bedrock Invoke
display updates and clear_thinking rebuilds carry block_binding only when the
config opts in. Every other config sends the same thinking as 1.105.1
* fix(anthropic): treat adaptive thinking as enabled when mapping response_format
is_thinking_enabled only recognised thinking.type=enabled, so a request that
paired thinking.type=adaptive with response_format took the synthetic JSON
tool path with a forced tool_choice, which Anthropic rejects alongside
thinking. The same check also gates the Bedrock Converse and Databricks
forced tool_choice downgrades, so they now see adaptive thinking too.
* fix(anthropic): keep signed thinking blocks with empty text on chat completions
Models that default to display: omitted return signed thinking blocks whose
text is empty, and Anthropic rejects the next turn when they are not replayed
unchanged. is_unsignable_thinking_block dropped them along with unsigned
blocks, and the inline content path required non-empty text, so a tool-use
turn on Opus 4.7+ or the 5 families lost its thinking blocks and 400ed on the
tool-result turn. Only a missing signature now drops a block, and the
mid-conversation system placement mirrors that rule. The /v1/messages strip
helper is left to #41362.
* fix(anthropic): skip the thinking-strip retry when block_binding asks for an error
The /v1/messages retry loop strips thinking blocks and resends on any
invalid-thinking 400, including the binding mismatch a caller opted into
with thinking.block_binding.prefix_mismatch_behavior: error. The retry
decision now sees the request body and surfaces that 400 instead of
overriding the caller's choice. #42888 covers the litellm_params side.
* fix(bedrock): carry thinking.block_binding through the Invoke thinking rebuilds
Bedrock Invoke rebuilt the thinking param from scratch when display: updates
was requested, on both the /v1/messages and chat completions paths, and the
legacy enabled-to-adaptive translation plus the clear_thinking injection did
the same, so thinking.block_binding never reached Bedrock. Every rebuild now
spreads the caller's thinking and only overrides the keys it owns, and
AnthropicThinkingParam types block_binding. The thinking-binding-controls
beta was already on the Bedrock allowlist.
* fix(anthropic): keep forcing the response_format tool under adaptive thinking
This reverts ad9e861ae7. Anthropic, Bedrock Invoke and Vertex all accept a
forced tool_choice next to thinking.type=adaptive and only reject it with
type=enabled, so counting adaptive as enabled fixed no live failure. It did
drop the forced JSON tool, so Claude answered response_format requests in
prose, and it downgraded a required tool_choice to auto on Bedrock Converse
* test(e2e): drop the adaptive thinking tool round trip case
With adaptive thinking the model often skips thinking on the weather prompt, so
turn 1 carries no signed empty block and the case fails before the replay it is
meant to cover. Anthropic also answers the replay with a 200 whether or not the
block is forwarded, so the case could not tell the fix from main. The factory
unit tests pin the outgoing shape instead
* fix(anthropic): only skip the thinking-strip retry for a binding mismatch 400
prefix_mismatch_behavior: error asks Anthropic to reject a block bound to
another conversation. It says nothing about unrelated bad blocks such as a
null signature, so those still get the strip-and-retry self-heal. Also trims
the empty-thinking docstrings to what the code guarantees and types the new
test parameters
* refactor(anthropic): narrow thinking fields with the mapping validator instead of typing.cast
typing.cast is banned by the strict ruff gate (TID251), so the three new casts pushed that rule one over its merge-base count. The existing pydantic mapping validator gives the same typed view without an unchecked cast, and it now has a generic name since it is no longer only used for Claude Code payloads
* refactor(anthropic): read thinking settings through the validated mapping so basedpyright stays at its merge-base count
The Bedrock clear_thinking rewrite still narrowed with a cast and an untyped dict fallback, which added two reportUnknownVariableType errors. It now uses the same pydantic mapping validator, made public for that reuse, and the retry hook validates the request body itself so the handler passes a typed dict instead of an untyped one
---------
Co-authored-by: yuneng-jiang <yuneng@berri.ai>
(cherry picked from commit f470e418ea)
The request-side sanitizer on /v1/messages dropped every thinking block whose
text is empty, even with a valid signature. Claude Opus 4.8 / Fable 5 / Fable 5.1
return exactly that shape by default (thinking.display omitted): the signature
is the preserved reasoning. Replaying the assistant turn as returned therefore
lost the model's reasoning on every multi-turn request, silently (the provider
answers 200 without it), and left preserved-thinking enforcement
(thinking.block_binding) nothing to bind.
Anthropic accepts an empty thinking block that carries a signature and binds it
to its prefix; only an empty block with no signature (the LIT-6357 case, a turn
a non-Anthropic model served through the bridge) is rejected with "each
thinking block must contain thinking". Use the signature-aware predicate
is_empty_unsigned_thinking_block, which #38809 already applied on the emit
side, for the request side too. The unsigned case is still dropped.
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
(cherry picked from commit 897bfea38d)
* feat(sdk): add litellm.systemone and trim litellm.decisions to the OpenAI shape
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* ci: run tests/unit/systemone in the core-utils unit shard
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(sdk): keep litellm.decisions(state=...) working with a DeprecationWarning
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test: make tests/unit/systemone a package
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(sdk): surface the litellm.decisions(state=...) deprecation as a UserWarning
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>
(cherry picked from commit 25d9b05386)
* feat(strands_decider): call a Strands Decider on an AgentCore runtime ARN
A strands_decider deployment whose api_base is a bedrock-agentcore runtime ARN now
calls InvokeAgentRuntime on that runtime, signed with SigV4 from the deployment's aws_*
params or the default chain, reuses one session per runtime, and maps the runtime's
200 error envelopes to 400, 503, or 500
* ci: run the strands_decider unit tests in the llms job
* fix(strands_decider): take the AgentCore session id from the deployment, never the caller
A caller could pick the runtime session through extra_headers, and every new id starts another microVM. The session id now comes from the deployment's agentcore_runtime_session_id (banned in request bodies) or the ARN default, and caller-supplied host, x-amz-*, and x-amzn-bedrock-agentcore-runtime-* headers are not forwarded
* fix(strands_decider): map runtime error replies only on AgentCore ARN calls
* fix(strands_decider): sign plain decisions calls inline, off the AWS signing executor
* test(strands_decider): cover AgentCore runtime ARN calls in integration wire tests
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
(cherry picked from commit ed9c31cfa8)
#45214 routed both decision formats through the shared decisions IR in
litellm/llms/base_llm/decisions/transformation.py and litellm/types/decisions.py.
That left litellm/llms/base_llm/decisions/systemone.py (to_system_one_request,
question_keys, to_decisions_response, the SystemOne* models and their adapter)
and litellm/types/openai_decisions.py with no importer outside their own unit
tests, so both modules and both test files go.
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
(cherry picked from commit 0e48048bd5)
* refactor(decisions): dispatch /v1/decisions through provider configs and the shared HTTP handler
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(decisions): map OpenRouter connection failures to APIConnectionError and drop explanatory docstrings
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(decisions): keep APIConnectionError for unreachable deployments on every provider
`_map_upstream_exception` now maps a status `_handle_error` synthesized for a
non-HTTP failure to `APIConnectionError` before `exception_type` runs, so a
Perplexity (openai-compatible) deployment that cannot be reached no longer
answers `InternalServerError` where main answers `APIConnectionError`.
Audit cells: the unreachable-deployment integration cell is parametrized over
the five providers, OpenRouter connection failures are pinned on chat, stream,
embeddings, decisions and the SDK, a deployment whose provider has no decisions
config gets the 400 naming every supported provider, and a unit cell pins the
mapper alone. The first docstring sentence of the five translation bases is
dropped.
---------
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
(cherry picked from commit 0e09e99146)
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(decisions): add the OpenAI Decisions spec types and the System One translation
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(decisions): share one DecisionsModel config and require model and usage on responses
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(decisions): rename the shared pydantic parent to DecisionsObjectBase
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(decisions): match the SDK on strict choice values and drop the invented min_length bounds
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(decisions): compose DecisionsRequest from model and body and read the translator top down
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(decisions): use the plural Decisions prefix only for the request and response envelopes
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(decisions): import assert_never from typing_extensions for Python 3.10
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>
(cherry picked from commit 79f62db620)
* feat(decisions): add unified /v1/decisions endpoint for Jev-compatible providers (#44236)
Backport of #44236 to rc/1.105.0.
Cherry-picked from 8b1990b4bc (main).
* feat(decisions): serve System One format at /v1/systemone and OpenAI format at /v1/decisions
Backport of #45184 to rc/1.105.0.
Cherry-picked from 3339469e73 (litellm_systemone_endpoint). The playground System One tab is not on this line, so its UI hunks are dropped.
* fix(decisions): import assert_never from typing_extensions for Python 3.10
Backport of the #45184 follow-up to rc/1.105.0.
Cherry-picked from 29e715473c (litellm_systemone_endpoint).
* fix(decisions): cap OpenAI-format questions at the System One limit
Backport of the #45184 follow-up to rc/1.105.0.
Cherry-picked from 7ef132947a (litellm_systemone_endpoint).
* fix(decisions): return 400 for bodies that are not JSON
Backport of the #45184 follow-up to rc/1.105.0.
Cherry-picked from b6bb885324 (litellm_systemone_endpoint).
* test(decisions): send System One bodies to /v1/systemone in integration tests
Backport of the #45184 follow-up to rc/1.105.0.
Cherry-picked from 77872f6cda (litellm_systemone_endpoint).
* feat(decisions): add OpenAI as a Decisions provider
System One requests to openai/ models are translated to OpenAI's Decisions shape on the way out and OpenAI's answers are translated back, so both /v1/systemone and /v1/decisions can route to gpt-6-luna.
(cherry picked from commit b94bbbc318)
* refactor(decisions): move the OpenAI wire translation into llms/openai
(cherry picked from commit d209a85076)
* feat(decisions): route both request formats through a shared decisions IR
System One and OpenAI-format bodies now convert to one internal representation, and each provider translates it to its own wire format. OpenAI deployments receive the caller's messages, images, names, typed choice values, level descriptions and safety_identifier unchanged. System One providers return a 400 for image input. Cached and cache-write tokens are priced on both response shapes, and provider response extras survive the round trip.
(cherry picked from commit 0dd207fde1)
* fix(decisions): keep provider fields inside System One answers
(cherry picked from commit 1348de864f)
* fix(decisions): load on Python 3.10 and 3.11 and bill cached tokens once under custom pricing
Decisions IR dataclasses used a mappingproxy default, which is only hashable from Python 3.12, so import litellm failed on 3.10 and 3.11. Decisions usage now reaches cost_per_token as a Usage with prompt_tokens_details, so custom pricing no longer adds cached and cache-write tokens on top of input_tokens that already include them
(cherry picked from commit 39012f1c22)
* fix(decisions): honor litellm OpenAI key and base settings for the openai provider
OpenAI decisions now resolve the key from litellm.api_key, litellm.openai_key, then OPENAI_API_KEY, and the base from litellm.api_base, OPENAI_BASE_URL, then OPENAI_API_BASE, matching other OpenAI calls. Providers own their configured key and base lookup through the endpoint config.
(cherry picked from commit e465919004)
* test(decisions): clear every OpenAI base setting in the proxy OpenAI deployment test
(cherry picked from commit a2951b318d)
* fix(decisions): send OpenAI instructions for System One questions without them and reject single-option questions
(cherry picked from commit e4081f77c6)
* fix(decisions): return guardrail_information on OpenAI-format decisions when requested
Backport of the #45184 and #45214 guardrail fix.
Cherry-picked from 855f0bb765 (litellm_systemone_endpoint) as merged into #45214 at c19ada66ed (litellm_openai_decisions_provider).
---------
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Cherry-pick of the squash commit of #44307, which serves gpt-5.6+ chat
completions on the native Bedrock /openai/v1/chat/completions route by
default and adds the chat_completions/ opt-in for gpt-oss and grok.
Conflict resolution: tests/unit/llms/bedrock/test_bedrock_common_utils.py
kept the line's file plus the one test the pick adds, since rc/1.105.0
does not carry #44112's neighbouring tests. The pick's
bedrock_route_for_request builds a MappingProxyType, whose import #44112
brought to main, so the import is added to litellm/llms/bedrock/common_utils.py
here.
(cherry picked from commit 233db9f285)
* fix(bedrock): accept Converse messages with no content key
A user or tool message whose content key is missing (or null, which the
message cleanup strips) made every Bedrock Converse request fail with
APIConnectionError 'content' before reaching Bedrock. The Converse
transform now reads content with .get for those messages, as it already
did for assistant messages: a content-less user message adds no block
and a content-less tool message becomes a toolResult with empty content.
The str branch also sends the continue message text instead of the
original whitespace-only text.
* fix(bedrock): send the continue message for a content-less user turn
* test(bedrock): type the content-less Converse message test parameters
* fix(bedrock): accept a Converse system message with no content key
* refactor(bedrock): read the system message content with get
* test(integration): audit Bedrock Converse messages without content
Adds the /audit cells for a chat message whose content key is missing or
null on a Converse-routed Bedrock deployment: happy, sad, edge, and chaos
rows through the OpenAI SDK, the Anthropic SDK, and raw httpx against the
scripted upstream, asserting the caller's response, the body the peer
received, and the spend row. The owned-proxy readiness deadline in the
integration harness is now INTEGRATION_PROXY_READY_SECONDS (default 70).
* test(integration): bound stray spend rows in the mid-burst restart cell
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* chore(lint): remove the LIT002 mutable-construction rule
Drop LIT002 from scripts/check_type_discipline.py along with its helpers,
its budget entry, its unit tests, and the AGENTS.md and gate docstring
mentions. `# mutable-ok` now only suppresses LIT001, so the markers that
only existed to silence LIT002 became LIT013 stale suppressions and are
removed. The files whose layout depended on those trailing comments are
reformatted with ruff format.
Every other LIT rule count is unchanged and the ASTs of all touched
litellm/ files match main apart from one docstring.
* chore(lint): keep the leftover mutable-ok markers for a follow-up
Restore the ~1.4k `# mutable-ok` markers stripped in the previous commit
so this PR only touches the checker, its tests, the budget, and docs.
Those markers no longer suppress anything, so `# mutable-ok` is exempt
from LIT013 until a follow-up strips them.
* Revert "chore(lint): keep the leftover mutable-ok markers for a follow-up"
This reverts commit c35bc0b84e.
---------
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
* test(ci): add used_client_oauth_token to the GCS pub/sub spend-log golden
#43063 stamps used_client_oauth_token into spend-log metadata, so
test_async_gcs_pub_sub_v1 failed on main with an extra metadata key
* test(ui): give the auto-router threshold save wait room for the availability debounce
#42625 keeps Save disabled while a 300ms-debounced availability check runs.
This test waits for Save right after the change, so the whole debounce lands
inside waitFor's 1s default and it times out under CI load. It is the
recurring UI Unit Tests failure on main since #42625 landed
* test(e2e): expect no pricing tier on bills for streamed calls OpenAI served at default
#42870 added both the rule that a served default or standard tier bills at
base pricing and records no service_tier, and streamed tests expecting the
row to record 'default'. They have failed on every scheduled litellm-e2e run
since. The tests now map the served tier to the pricing basis the bill must
record and check input is billed at that basis's rate; the messages case
registers custom rates so the rate check has something to compare against
* test(e2e-ui): wait for the call-id search before hovering the logs row
The row the spec hovers is already on the unfiltered first page, so it was
found before the search request returned. The search response then
re-rendered the table under the mouse, and the Base UI tooltip never opened.
Reproduced with Playwright against a local proxy: hovering right after the
fill never shows the tooltip, hovering after the search response shows the
call id every time
* test(e2e): run the Together structured-output case on the hybrid Qwen with reasoning off
The case picked the cheapest Together row flagged supports_response_schema.
DeepSeek-V4-Flash-0731 hit its cost-map deprecation date on 2026-09-29, so the
pick moved to GLM-5.3-Flash, a reasoning-only model that spends the 1024-token
budget thinking and returns content=None. Qwen3.5-9B is the pinned hybrid model
the reasoning_effort=none case already exercises, and Together lists it with
structured output support
* test(integration): read the agent 365 guardrail status by its own name in spend logs
The MCP shard runs under xdist against one database, and a sibling file creates a
default_on pre_mcp_call content filter there. The owned proxy reloads DB guardrails, so
that filter's 'success' entry could land first in guardrail_information and the test
read it instead of the agent 365 verdict
* test(unit): ignore asyncio's leaked-task records in the budget limiter push-failure log check
gc.collect() inside the caplog window can collect a pending task an earlier test left
on a closed loop, and asyncio logs 'Task was destroyed but it is pending' into this
test's records. The check still counts every LiteLLM logger, and unretrieved task
exceptions on this loop still go through the asserted exception handler
* test(e2e-ui): fill the create-tag fields inside the dialog
#42949 added 'Filter by tag name' and 'Filter by description' inputs to the Tag
Management page, so page-wide getByLabel('Tag Name') and getByLabel('Description')
match two elements and Playwright's strict mode fails the create step
* test(integration): run integration proxies with the CI license
Multi-worker proxies start each uvicorn worker in a fresh process, so every
worker reads the license from its environment. Forward LITELLM_LICENSE into the
proxy and test runner environments
* ci: save GitHub Actions caches only from main and bump codecov-action to 5.5.5
Every pull request saved its own uv, maturin, Rust and Prisma caches, about
4.5 GB per PR, so the repository's 10 GB cache budget evicted main's entries
within minutes. Pull request jobs then missed every cache, downloaded all
dependencies from PyPI and hit the install step timeouts. Pull requests now
restore only, and main keeps the caches warm for them. test-linting and
check-ui-api-types run only on pull requests and keep saving
codecov-action 5.5.4 imports its signing key from the deleted codecovsecurity
keybase account, so every upload failed signature verification. 5.5.5 reads it
from codecovsecops; the key ID matches the one signing the current CLI
* test(unit): join the session-minting thread before collecting the handler
asyncio.to_thread resumes the test as soon as the worker sets its result,
while the pool thread can still hold the work item and through it the
handler. gc.collect() then cannot finalize the handler and the session stays
open. A pool that shuts down before the test continues drops that reference
* test(integration): relaunch owned proxies that lose their port, expire idle gateway connections early
owned_proxy_process released its reserved port and the proxy bound it only
after full startup, so another xdist worker or an outgoing connection could
take it first and the proxy exited with 'address already in use'. The launch
now retries on a fresh port when that happens and stops every failed attempt.
uvicorn closes idle keep-alive connections after 5 seconds and httpx expired
them at the same 5 seconds, so a request sent right at that mark could reuse a
socket the server was closing and get 'Connection reset by peer'. Gateway
clients now drop idle connections after 2 seconds
* ci(circleci): give the base SDK wheel build the same 30 minute no-output window as the Windows build
The release profile builds with fat LTO and one codegen unit, so the final
link of litellm-cache-s3 runs silently for minutes. Successful builds take
711 to 749 seconds, right at the default 10 minute no-output limit, and about
30% of recent runs were killed there
* test(integration): model the budget-reset database outage as 10 seconds instead of 5 refused connections
The proxy retries the database about every 30 seconds and each retry opens
roughly one connection, so a 5-connection outage took 3 to 4 retries to clear
and recovery landed between 60 and 90 seconds, straddling the test's 80 second
reset window. A fixed 10 second outage still refuses the immediate reconnect
and recovers on the next retry
* ci: move the unit-test uv cache split into a composite action
check_workflow_startup_safety sums every setup step's timeout, so the save and
restore variants each counted 5 minutes although only one runs. One composite
step keeps the setup ceiling at 35 minutes
* test(unit): point tiktoken at the bundled cache for every unit test
The rust_bridge tokenizer tests loaded o200k_base before any test in their
xdist worker had imported default_encoding, so tiktoken fell back to the
temp cache and tried to download under pytest-socket. Move the session
fixture from litellm_core_utils/conftest.py to the root unit conftest.
* test(integration): answer model discovery probes in the hosted_vllm wire tests
The router's periodic upstream model info refresh sends GET /v1/models to
hosted_vllm deployments, so a wire server that is live during a refresh
sees an extra request. Answer the probe with an empty model list and leave
it out of the provider-call assertions, matching the responses bridge
tests.
* feat(guardrails): honor litellm_params.timeout in every HTTP guardrail
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(guardrails): accept timeout kwarg in presidio and responses-handler post stubs
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): bound hiddenlayer startup jwt call by configured timeout, drop akto from timeout coverage
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(guardrails): narrow hiddenlayer startup auth timeout without cast
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): bound hiddenlayer jwt refresh by configured timeout
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(guardrails): keep provider timeout defaults when unset and bound only rubrik moderation calls
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(guardrails): cover model_armor and run timeout probes concurrently
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(guardrails): match sink calls to the exact guardrail name
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(transcription): honor base_url alias for Groq Whisper and report it as the api base
transcription() and speech() only accepted api_base, so a deployment configured with base_url leaked the alias into the provider params, which Groq rejected as an unknown param, and the request never reached the internal gateway. get_api_base() now reads the same alias so response headers and logs show the configured endpoint instead of the provider default
Resolves LIT-9071
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(speech): keep base_url after existing audio params, route Vertex speech to it, skip empty alias
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(hosted_vllm): keep reasoning_content on assistant messages in _transform_messages
vLLM accepts reasoning_content (200 on the wire) and qwen/deepseek/glm
chat templates consume it, so popping it made reasoning models lose
earlier reasoning across tool loops. thinking_blocks is still removed
for vLLM compatibility.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(hosted_vllm): forward replayed reasoning_content only when it is a string
* test(integration): cover hosted_vllm reasoning_content replay across endpoints
* test(integration): require the surviving worker to serve its held requests in the sigkill chaos cell
---------
Co-authored-by: mateo <mateo@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Co-authored-by: yassin <yassin@berri.ai>
* fix(responses): scan and mask top-level instructions with guardrails
The Responses guardrail translation handler put a non-empty top-level instructions field into structured_messages as a system row but never into the flat texts list, so guardrails that scan texts skipped it, flat-text masking could not rewrite it, and PANW latest-only selection failed its alignment guard whenever instructions were present.
Seed texts with the instructions row, carry that offset into the flat-text write-back so a rewritten row lands on data["instructions"], and account for the leading row in the PANW Responses alignment.
Resolves LIT-8931
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(responses): reject empty guardrail rewrites instead of forwarding raw input
An explicit texts=[] answer from a guardrail now fails the count check and
raises UnappliableRequestRewrite like any other misaligned rewrite; only a
missing texts key means no rewrite. Types the out-param as dict[str, object]
and adds integration coverage for instructions blocking, masking, empty
instructions, tool loops, latest-only and concurrent workers
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(responses): type the texts-replacing guardrail helper explicitly
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(responses): honor skip_system_message_in_guardrail for instructions and system input items
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(responses): cover skip_system_message_in_guardrail on the live proxy
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(responses): keep skipped rows through full-coverage rewrites and align latest-only with skip_system
Trust a guardrail's structured_messages_cover_full_request claim only when it
returns as many rows as the full normalized request, otherwise merge the scoped
rows back so skipped instructions and system items survive the write-back.
Make PANW's Responses reasoning alignment skip-aware so latest-only still picks
the latest user turn when system content is excluded from texts.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(responses): annotate new guardrail tests with return types
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(responses): treat an empty guardrail texts answer as no rewrite like chat completions
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(responses): type the guardrail test doubles explicitly
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(streaming): keep the provider's served service_tier on streamed chunks and spend rows
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(streaming): satisfy type-discipline and strict ruff budgets
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(streaming): stamp the served service_tier on every Responses bridge chunk
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(anthropic-adapter): expose streamed chunks so disconnects bill partial spend
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(service-tier): cover anthropic and responses served-tier billing paths
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(anthropic-adapter): return a chunks-exposing stream so disconnects bill partial spend
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(service-tier): bill disconnects through the router's anthropic stream wrapper
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style: apply ruff format to the anthropic stream changes
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(coverage): ignore delegating properties the ast scan cannot see
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style: keep the cast-ok reasons on the cast call line
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(integration): cover served service_tier billing for streamed chat and messages, complete and disconnected
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(anthropic-cache): delegate chunks/messages/model through the messages stream cache writer
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(streaming): keep service_tier on OpenAI-compatible parsed chunks
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(streaming): parameterize delegated chunks and messages types
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(tests): follow the anthropic pass_through rename after merging main
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(anthropic): drain the logging worker between response cache tests
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(spend): cover azure, databricks, responses bridge and gemini served tiers in the stream billing integration test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(databricks): keep the served service_tier on streamed chunks and bill it
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(databricks): type the served service_tier chunk without a loose kwargs dict
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(cost): bill the served service_tier over the requested one
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(cost): drop explanatory comment from the tier resolution
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: kerry <kerry@berri.ai>
* fix(router): stream anthropic messages lifecycle frames live when no fallback can take over
The /v1/messages streaming wrapper buffered message_start and
content_block_start until the first content_block_delta and dropped
pings behind buffered frames unconditionally, even for requests no
fallback could ever recover. With adaptive thinking on Bedrock or
Vertex the client saw no bytes for the whole thinking pass and hit
read timeouts.
Buffering now applies only while a fallback can still take over
(generic or refusal chain resolving), and a ping is always forwarded
live since it carries no lifecycle and keeps the connection alive.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(router): mirror every dispatcher fallback path in the anthropic stream gate
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(router): skip already-tried order levels in the anthropic stream gate
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(router): keep a transport-split ping behind buffered lifecycle frames instead of forwarding its head live
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(router): credit the #39566 branch this fix supersedes
Co-authored-by: Radu Swigler <radu.porumba@gmail.com>
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>
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Radu Swigler <radu.porumba@gmail.com>
Claude Code attaches output_config to mid-conversation system messages and
sends the per-turn-control-2026-07-01 beta with it. The Vertex beta map
dropped that beta, so Vertex rejected the body with
'messages.N.output_config: Extra inputs are not permitted'.
Forward the beta for vertex_ai, the way azure_ai already does, and add it on
the Vertex Messages path whenever a message carries output_config.
* feat(fireworks_ai): route and list the auto, auto-instant and firerouter routers
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(fireworks_ai): drive the router request test through an httpx MockTransport
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(fireworks_ai): let custom firerouter/<models> IDs inherit the firerouter row's capabilities
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(fireworks_ai): integration coverage for router short names forwarding tool_choice and reasoning_effort
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(fireworks_ai): assert tool definitions reach the router upstream
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(providers): add Prism provider
* fix(providers): complete Prism registration
* feat(providers): expose Prism responses and messages
* feat(providers): add DeepSeek V4.1 Flash to Prism
* test(providers): exercise Prism endpoint requests
* fix(providers): align Prism pricing and limits with the live catalog
deepseek-v4.1-flash bills 0.17/0.63 USD per 1M input/output tokens and takes image input;
deepseek-v4-flash bills 0.17/0.21 and caps output at 384000 tokens, per GET /v1/models
* test(prism): assert cost-map invariants instead of pinning catalog facts
* test(prism): derive the asserted model list from the cost map instead of pinning it
* test(prism): capture requests through respx instead of appending to a list and swapping the client transport
---------
Co-authored-by: rajitkhanna <rajitskhanna@gmail.com>
Co-authored-by: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Co-authored-by: ryan <ryan@berri.ai>
test_ssl_verify_unit.py inserted tests/unit at the front of sys.path, so any later import of litellm_proxy_extras resolved to the tests/unit/litellm_proxy_extras test package. Whenever the CircleCI shard split collected that file before test_litellm_proxy_extras_logging.py, collection failed with ModuleNotFoundError. test_gemini_session_leak.py had the same insert for its own directory
completion() imported vertexai only to check that the package exists. Partner
models are reached with an authenticated httpx client and never use that SDK,
the same reasoning count_tokens in this file already follows (#28084). The
import loads all of google-cloud-aiplatform on the first request of every
process and made a google-auth-only install fail with a 400
* fix(responses): emit the reasoning item on streaming /v1/responses for signature-only thinking
Anthropic models return thinking blocks with empty text and the reasoning carried in the
signature: Claude Fable 5.1 and Claude Opus 5.5 by default, and Bedrock adaptive thinking
with or without an effort. On streaming /v1/responses the chat->Responses bridge opened a
reasoning output item only on reasoning_content text
(LiteLLMCompletionStreamingIterator._ensure_output_item_for_chunk), and
ChunkProcessor.get_combined_thinking_content kept an assembled thinking block only when it
had thinking text. Such a response emitted no reasoning item mid-stream and none in
response.completed, so a streaming Responses client could not replay the reasoning even
though the reasoning tokens were billed. Non-streaming /v1/responses was unaffected.
Open the reasoning item when the delta carries a signed or redacted thinking block, and
keep a signed block through stream assembly even when its thinking text is empty.
Unsigned text-only fragments are still dropped. The reasoning-text path is unchanged.
(cherry picked from commit bc9b6f8a5c)
* test(vertex_ai): move orphaned gemma streaming tests into the llm-vertex-ai shard
PR #43147 left a copy of the Gemma streaming tests under
tests/test_litellm/llms, a tree no CI shard claims, which broke
assert-ci-coverage and assert-shard-coverage on main. Fold the two
streaming tests into the existing tests/unit/llms/vertex_ai file so the
llm-vertex-ai shard runs them
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Chloe Lu <chloe.lxd@gmail.com>
Co-authored-by: Krrish Dholakia <krrishdholakia@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(vertex_ai): reproduce traced Gemma Responses stream failure
* fix(vertex_ai): wrap Gemma fake streams for Responses tracing
* test(vertex_ai): cover Gemma traced streams and usage options
* test(vertex_ai): inject gemma test deps and assert hidden usage accounting
Replace class-level patches in the Vertex AI shard test with the
provider's documented dependency-injection seams (httpx.MockTransport
client + credential cache), and pin the default/omit-usage trace
behavior: LiteLLM still accounts all tokens; ddtrace's metric is
absent by design, asserted rather than silent.
Mutation-checked: commenting out CustomStreamWrapper chunk accumulation
turns the new assertions red; restoring them turns green.
* test(vertex_ai): drop explanatory comment from usage-option assertions
