Commit graph

18968 commits

Author SHA1 Message Date
jesus
8549189cbb test(proxy): use local_model_cost_map fixture for alias listing test
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:59:24 +00:00
ryan-crabbe-berri
6b54238083
Merge pull request #40807 from BerriAI/litellm_service_account_team_key_mgmt
feat(keys): let team service account keys use key management endpoints for their own team
2026-09-18 12:58:41 -07:00
Yassin Kortam
2edda5aec3
Merge pull request #41757 from BerriAI/litellm_typesafe_compaction_guardrail
feat(guardrails): add TypeSafe Jev relevance-based compaction guardrail
2026-09-18 12:56:52 -07:00
jesus
32e6a86319 fix(router): report null cost for unpriced deployments instead of 0
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 12:52:57 -07:00
yassin
cf22e4b171 fix(proxy): make SettingsStore.clear terminate when the config file owns a key
MutableMapping.clear pops items until the mapping is empty, but __delitem__
keeps config-owned keys, so clear spun forever on any store that had loaded a
config file. unittest.mock.patch.dict calls clear on exit, which is why the
proxy-infra and proxy-endpoints shards hung at 99 percent until the 20 minute
job timeout on every run since the store started refusing config-owned writes

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:51:15 +00:00
Mateo Wang
a7a3f6802e
Merge pull request #41203 from clonylu/fix/anthropic-thinking-binding-beta-header
fix(anthropic): register thinking-binding-controls-2026-08-01 in beta headers config
2026-09-18 12:50:55 -07:00
mateo-berri
58beea2275 fix(passthrough): read the stream flag by truthiness in the timeout resolver 2026-09-18 12:45:54 -07:00
mateo-berri
8533cb9d59 test(proxy): drop the pass-through routes the reload tests add to the shared app
The two pass-through reload tests register routes on the shared app and
never take them off, so a later allowlist test in the same xdist worker
finds routes that no component exposes. Restore the app's route list
when each of those tests finishes.
2026-09-18 12:44:42 -07:00
Devin AI
4f904f69b6 test(proxy): restore app routes after pass-through reload tests
The two pass-through reload tests register real FastAPI routes on the
shared proxy app and only clean up the internal registry, so any test
running after them in the same worker sees stray /v1/kept-* and
/v1/deleted-* routes. test_component_allowlists counts those as
uncovered and fails. Snapshot app.routes and the registry up front and
restore both in a finally block.

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:44:18 +00:00
mateo-berri
7cfec730d5 fix(proxy): refuse config-owned keys on POST /config/update
POST /config/update stored keys the config file owns and answered 200
while the file silently kept winning. Run the config-owned check for
general_settings, litellm_settings, and router_settings before the first
database read, the way /config/field/update already does, so a refused
request stores nothing.

The config reload re-loaded the merged settings as yaml settings, which
turned every saved router setting read-only after one tick. Read the
saved router settings row instead so database-owned values stay writable.

The integration harness seeds num_retries through /config/update instead
of the config file, which is what the effective-settings and observed
routing tests need to keep exercising a database-owned value.
2026-09-18 12:42:08 -07:00
mateo-berri
c0b0ba20b2 fix(azure-ai): keep FLUX.2 tolerant of OpenAI-only image params 2026-09-18 12:41:40 -07:00
yassin
fe8cf02823 feat(mcp): resolve the allowlisted client identity from the JWT claim or an opt-in header instead of clientInfo.name
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:27:43 +00:00
yassin
1e7c5400fd fix(proxy): canonicalize azure speech paths and bill uploaded short audio
Resolve dot segments in the /azure_speech endpoint path before the endpoint family and the admin-only batch guard are decided, so the guard and the forwarded upstream path agree. Bill short-audio requests for the longer of the uploaded audio duration and the recognized duration, so a NoMatch or silence response still charges for the audio Azure processed

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:27:25 +00:00
mateo-berri
b4dc081c27 fix(mistral): never round reasoning_effort none onto the strength ladder 2026-09-18 12:22:34 -07:00
Devin AI
a774e71175 test(bedrock_mantle): document mantle in-region pricing invariant
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:18:45 +00:00
Devin AI
2299846b48 fix(bedrock_mantle): align gpt-5.6-sol pricing with the AWS in-region model card
Co-authored-by: kusumakarb <kusumakarbvss@gmail.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:18:02 +00:00
yassin
b5ad17e848 test(docs): read only the first column of the router_settings reference table
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:16:26 +00:00
yassin
649d299c44 Merge remote-tracking branch 'origin/main' into litellm_fix_tpm_window_reset_sibling_counters 2026-09-18 19:13:05 +00:00
yassin
afa8b8a903 test(docs): read only the first column of the router_settings reference table
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:11:39 +00:00
yassin
84f7adec2e fix(passthrough): match deepgram listen routes served under a path prefix
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 19:06:24 +00:00
Mateo Wang
a2626726a2
Merge pull request #40500 from BerriAI/litellm_bedrock_aws_auth_params
fix(bedrock): send aws_session_tags on every STS call via one typed auth struct
2026-09-18 12:01:52 -07:00
yassin
c5919a3c0e Merge remote-tracking branch 'origin/main' into litellm_mcp_client_allowlist
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

# Conflicts:
#	litellm/proxy/_experimental/mcp_server/server.py
#	litellm/proxy/proxy_server.py
2026-09-18 19:01:08 +00:00
ryan-crabbe-berri
ed9e666b54
Merge pull request #41351 from BerriAI/litellm_routing_groups_atomic_validation
fix(router): validate routing_groups at save time and keep invalid DB groups from blocking SSO load
2026-09-18 11:59:21 -07:00
mateo-berri
c1e39810ec fix(mistral): send reasoning_effort as a level the model accepts
Declare the live-verified reasoning_effort_levels on the Mistral cost-map
entries and round an undeclared request to the nearest declared level
(up to the weakest level at least as strong, down to the strongest when
the request exceeds the ceiling). Codex's default medium no longer 400s
on mistral-medium-latest, mistral-small-latest, or the vibe-cli family;
an entry that declares nothing keeps forwarding the value verbatim
2026-09-18 11:58:04 -07:00
kerry
fb8816d826 Merge remote-tracking branch 'origin/main' into litellm_lit_8128_off_peak_pricing_schema 2026-09-18 18:55:24 +00:00
mateo-berri
73fddb999e fix(router): resolve stream_timeout before generic timeouts on the passthrough route 2026-09-18 11:53:26 -07:00
yassin
307df09792 test(mcp): assert the self-revoke response instead of echoing the delete mock
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 18:51:43 +00:00
ryan-crabbe-berri
839c040b53
Merge pull request #41695 from BerriAI/litellm_issue_classifier
ci: classify new issues into domain, provider, kind, priority and lift labels
2026-09-18 11:50:56 -07:00
Mateo Wang
adb5ee0881
Merge pull request #41275 from BerriAI/litellm_azure_strip_file_format
fix(azure): strip litellm format field from file and image content parts
2026-09-18 11:46:23 -07:00
yassin
3a5f286273 Merge remote-tracking branch 'origin/main' into litellm_fix_tpm_window_reset_sibling_counters 2026-09-18 18:44:33 +00:00
Mateo Wang
88c9dd1294
Merge pull request #41511 from BerriAI/litellm_foundry_a2a_entra_agents
feat(a2a): reach Microsoft Foundry agents with Entra auth and versioned card discovery
2026-09-18 11:44:14 -07:00
mateo-berri
cf08440557 fix(azure-ai): price FLUX.2 Flex by its resolved cost map key 2026-09-18 11:42:52 -07:00
mateo-berri
a4da989aa9 test(bedrock): type the STS recording helper 2026-09-18 11:42:52 -07:00
yassin
1e6b33ffab Merge remote-tracking branch 'origin/main' into litellm_azure_speech_passthrough
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

# Conflicts:
#	helm/litellm/templates/ingress.yaml
#	litellm/proxy/_types.py
#	litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py
#	litellm/proxy/pass_through_endpoints/success_handler.py
#	terraform/litellm/aws/locals.tf
#	terraform/litellm/gcp/locals.tf
#	tests/test_litellm/proxy/middleware/test_billable_request_metrics_middleware.py
2026-09-18 18:42:22 +00:00
Yassin Kortam
1653d132c5
Merge pull request #41506 from BerriAI/litellm_vertex_gcs_file_content_streaming
feat(vertex_ai): stream GCS batch output files from /v1/files/{id}/content
2026-09-18 11:42:14 -07:00
mateo-berri
42541a9233 fix(proxy): drop echoed cost-map pricing on a row's next save and build /model/info pricing stamps without mutation 2026-09-18 11:42:07 -07:00
yassin
aeca6ed7ba chore: merge main into litellm_deepgram_listen_websocket_passthrough
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 18:39:39 +00:00
yassin
048aaad627 Merge remote-tracking branch 'origin/main' into litellm_mcp_admin_terminate_sessions_revoke_credentials 2026-09-18 18:39:16 +00:00
Yassin Kortam
0594dd7caf
Merge pull request #41539 from BerriAI/litellm_vault_login_secret_namespace
feat(vault): add separate login and secret namespaces for HashiCorp Vault
2026-09-18 11:38:49 -07:00
Yassin Kortam
84ae0805ba
Merge pull request #41620 from BerriAI/litellm_team_member_temp_budget_increase
feat(proxy): temporary budget increase for team members
2026-09-18 11:38:34 -07:00
ryan-crabbe-berri
294a9e15de ci: classify new issues into domain, provider, kind, priority and lift labels
Every issue opened from now on is gated on the template headings, sent once
through the LiteLLM proxy with a strict JSON schema, and labelled from the
manifest in .github/labels.json. Old-template issues are not touched. The bug
template shrinks to Description, Config, LiteLLM Version and Steps to Repro,
both templates gain a domain dropdown, and the labelers that keyed off the old
component dropdown go away.
2026-09-18 11:37:55 -07:00
Yassin Kortam
f6d9b2552f
Merge pull request #41692 from BerriAI/litellm_mcp_gateway_sessions_by_client_user
feat(mcp): show live gateway sessions by AI client and user
2026-09-18 11:37:16 -07:00
Yassin Kortam
ca79c393d5
Merge pull request #41636 from BerriAI/litellm_per_key_end_user_default_budget
feat(proxy): per-key default budget for dynamically created customers
2026-09-18 11:36:59 -07:00
Yassin Kortam
f129e7e2d0
Merge pull request #41515 from BerriAI/litellm_transcribe_passthrough
feat(proxy): add Amazon Transcribe pass-through with completion-time job pricing
2026-09-18 11:35:06 -07:00
yassin
3219a875d0 fix(proxy): serialize in-memory window rollover and reset siblings once per expired window
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 18:20:18 +00:00
Yassin Kortam
d42f448e41
Merge pull request #41585 from BerriAI/litellm_lazy_fastapi_bpe_imports
perf: defer fastapi and tiktoken BPE imports out of import litellm
2026-09-18 11:19:13 -07:00
Yassin Kortam
660f3dfdd3
Merge pull request #41555 from BerriAI/litellm_max_parallel_requests_queue_size
feat(router): reject with 429 when a deployment's max_parallel_requests slots are all in use
2026-09-18 11:18:08 -07:00
kerry
4b215e2a60 fix(schema): require a schedule and well-formed windows in off_peak_pricing
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 18:17:36 +00:00
mateo-berri
860b0203c0 test(bedrock): expect canonical session tags in the dynamic auth params propagation test 2026-09-18 11:17:34 -07:00
kerry
cbcb55af89 fix(schema): classify off_peak_pricing as a structured object in the model prices schema generator
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
2026-09-18 18:11:17 +00:00