scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-09 22:31:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 87
46316 commits 4181 branches 1403 tags 2.3 GiB
Commit graph

14483 commits

Author SHA1 Message Date
mateo-berri
41192ef085 feat(ui): toggle internal health check visibility in request logs 2026-08-26 12:16:24 -07:00
mateo-berri
e56c42862c fix(health): probe Azure GA realtime path for transcription-only models 
The realtime health check always built the Azure websocket URL with the
default beta protocol, so GA-only transcription models such as
azure/gpt-realtime-whisper got probed at /openai/realtime and were
rejected with HTTP 400 on every /health run, while real calls through
the proxy resolved the GA path via intent=transcription and worked.

The probe now resolves the protocol the way the real call path does:
an explicit realtime_protocol (argument, deployment litellm_params, or
LITELLM_AZURE_REALTIME_PROTOCOL) wins, transcription-only models fall
back to GA with intent=transcription, and everything else keeps beta.
Transcription-only detection reads both mode and supported_endpoints
from get_model_info because a live proxy overwrites the catalog mode
with the operator's deployment model_info (mode: realtime) during
router registration, while supported_endpoints survives it.
get_model_info now propagates supported_endpoints from the cost map;
it declared the field but never populated it.
 2026-08-26 12:14:52 -07:00
Yuneng Jiang
291484a5e2
Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_/search-tools-sync-issue-e522a2 2026-08-26 12:11:06 -07:00
KnyazSh
736411bd67 Merge branch 'litellm_internal_staging' into feature/improve-gigachat-provider 2026-08-26 19:10:57 +00:00
mateo-berri
ece187ea24 Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_anthropic_fast_mode_cache_and_response_speed 
# Conflicts:
#	tests/test_litellm/test_cost_calculator.py
 2026-08-26 12:10:33 -07:00
yuneng-jiang
da528e455c
Merge pull request #38307 from BerriAI/litellm_proxy_types_validator_contracts 
test(proxy): pin the request-validation contracts in proxy/_types.py
 2026-08-26 12:09:53 -07:00
Marty Sullivan
5d8769bbaf fix(gemini-realtime): keep the client's voice on Vertex AI native-audio Live 
Vertex AI Live accepts a speechConfig on setup for native-audio models, so
stripping it silently discarded the voice a client asked for. Confirmed against
a live BidiGenerateContent session on gemini-live-2.5-flash-native-audio and on
gemini-live-2.5-flash-preview-native-audio-09-2025: both return setupComplete
with speechConfig present.

The strip stays in place for Google AI Studio, which was never verified to
accept it, via an overridable predicate matching the existing
_include_function_response_id pattern. The responseModalities TEXT to AUDIO
coercion is unchanged, since Vertex does reject TEXT on these models.
 2026-08-26 12:09:07 -07:00
Mateo Wang
ace28fd97a
Merge pull request #37384 from daniel-meismer-zocdoc/litellm_mcp_dcr_bridge_complete_challenges 
fix(mcp): complete DCR bridge OAuth challenges
 2026-08-26 12:07:54 -07:00
Mateo Wang
8dc17e808a
Merge pull request #38240 from BerriAI/devin_ai_anthropic_messages_missing_key 
fix(anthropic): raise missing-credential error on /v1/messages passthrough
 2026-08-26 12:06:37 -07:00
Mateo Wang
c13b278d79
Merge pull request #38369 from BerriAI/litellm_anthropic_geo_uplift_missing_models 
fix(cost-map): add US data residency uplift to claude-sonnet-4-6 and mythos entries
 2026-08-26 12:05:25 -07:00
mateo-berri
cbb50bb37e fix(responses): flush streaming cache write cancelled at event loop shutdown 2026-08-26 12:05:11 -07:00
Mateo Wang
c7b9060fb1
Merge pull request #38291 from BerriAI/devin_ai_lit6160_health_check_image_edit_mode 
fix(health): support `mode: image_edit` in health checks
 2026-08-26 12:04:53 -07:00
Devin AI
0cc407a02d Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_registry_audit_bedrock_sol_anthropic_1hr 2026-08-26 19:02:56 +00:00
mateo-berri
50f69a3a81 fix(health): strip client_secret, azure_ad_token, and other credential fields from /health output 2026-08-26 12:00:13 -07:00
Mateo Wang
0ada822928
Merge pull request #38094 from eugene-yao-zocdoc/litellm_redis_credential_provider 
fix(redis): support credential providers across clients
 2026-08-26 11:59:56 -07:00
mateo-berri
8a2fc2cc9f Merge remote-tracking branch 'origin/litellm_internal_staging' into fix/gh-36898-health-leak-extra-headers 2026-08-26 11:57:55 -07:00
yucheng-berri
ede4f3e8ab
test(prometheus): cover caller-identity config failure cases (#38380) 
* test(prometheus): cover caller-identity config failure cases

* test(prometheus): narrow pytest.raises with match to satisfy PT011
 2026-08-26 11:53:49 -07:00
mateo-berri
c5e3b21019 test(utils): add priority cache write tier key to intended map schema 2026-08-26 11:51:57 -07:00
Mateo Wang
80843ae7cb
Merge pull request #38279 from 6matt/litellm_bedrock_converse_gpt5_reasoning_effort 
fix(bedrock): map reasoning_effort to reasoning.effort for OpenAI GPT-5.x on Converse
 2026-08-26 11:48:18 -07:00
mateo-berri
1a696de40c fix(caching): flush async cache writes cancelled at event loop shutdown 2026-08-26 11:47:10 -07:00
Yuneng Jiang
91e7eb115d
fix(proxy): sync search tools into the router on management writes 
Creating a search tool through the UI only wrote the row; the router was updated
solely by the add_deployment job, so the tool was unusable for up to
PROXY_CONFIG_RELOAD_INTERVAL_SECONDS (30s by default) even on the worker that
served the write. Tools declared in config.yaml load straight into the router at
startup, which is why they never showed the delay.

The create, update and delete endpoints now refresh the router inline, matching
what the MCP server endpoints already do. The refresh is best-effort: the row is
already committed, so a failure must not surface as a 500 and push the caller
into a retry that creates duplicates.

Two related gaps go with it. _init_search_tools_in_db skipped the router update
whenever the merged list came back empty, so deleting the last search tool left
it live in memory forever. And in store_model_in_db-off deployments the
add_deployment job is never scheduled, so DB-backed search tools never reached
the router at all; that branch now loads them at startup and keeps them fresh on
its own interval, the same way MCP servers already do.
 2026-08-26 11:46:54 -07:00
mateo-berri
c330d466f3 fix(anthropic): normalize oauth authorization header to one canonical casing 2026-08-26 11:45:51 -07:00
Mateo Wang
767e6015af
Merge pull request #34658 from BerriAI/litellm_azure_realtime_entra_id_auth 
fix(azure/realtime): authenticate realtime websocket with Azure AD token when no api-key
 2026-08-26 11:44:38 -07:00
Mateo Wang
2d3144c6c8
Merge pull request #38368 from BerriAI/litellm_fix_bedrock_mantle_gpt55_gpt54_context_window 
fix(model_prices): raise bedrock_mantle gpt-5.5 and gpt-5.4 max_input_tokens to Mantle's enforced 1050000
 2026-08-26 11:42:47 -07:00
mateo-berri
dbe52a80c1 test(cost): wrap overlong us data residency test declaration 2026-08-26 11:42:28 -07:00
mateo-berri
e97a84afcf fix(model_prices): add above-272k priority rates to azure us/eu gpt-5.6 entries 2026-08-26 11:40:19 -07:00
yuneng-jiang
7bc80994b7
Merge pull request #38305 from BerriAI/litellm_/testing-strategy-audit-c39e33 
fix(ci): let the mutation workflow find covered lines so it generates mutants
 2026-08-26 11:34:08 -07:00
mateo-berri
9c38f6f125 test(tencent): drive thinking tests off the real cost map instead of patched internals 2026-08-26 11:34:06 -07:00
yuneng-jiang
b03e913ccf
Merge pull request #38315 from BerriAI/litellm_cost_estimate_pricing_edges 
test(cost-estimate): pin the prices and period totals /cost/estimate returns
 2026-08-26 11:32:22 -07:00
mateo-berri
8307be68c9 Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_pr38100_tencent_thinking 2026-08-26 11:26:58 -07:00
mateo-berri
229e136783 fix(mcp): honor admin-entered OAuth URLs on authorize after issuer yield 
Co-authored-by: Cursor <cursoragent@cursor.com>
 2026-08-26 11:26:54 -07:00
Mateo Wang
ca6387fae7
Merge pull request #38271 from daniel-meismer-zocdoc/litellm_mcp_bridge_provider_token_lifetime 
fix(mcp): preserve provider access token lifetime
 2026-08-26 11:25:55 -07:00
Mateo Wang
def5ca6d68
Merge pull request #38299 from BerriAI/litellm_fix_vertex_pt_custom_auth_google_token 
fix(proxy): keep the caller's Google token on credential-less Vertex passthrough under custom auth
 2026-08-26 11:24:56 -07:00
yuneng-jiang
309da5e70e
Merge branch 'litellm_internal_staging' into litellm_/testing-strategy-audit-c39e33 2026-08-26 11:24:55 -07:00
mateo-berri
a0d1fef89d fix(anthropic): scale cache costs by fast mode and trust served speed 2026-08-26 11:24:12 -07:00
mateo-berri
8ff832901b fix(health): accept image_edit mode on /health/test_connection 2026-08-26 11:18:50 -07:00
mateo-berri
2068066d69 fix(anthropic): detect oauth Authorization header case-insensitively 2026-08-26 11:16:17 -07:00
Devin AI
f3c1e2e2a7 fix(guardrails): forward aws_external_id when the bedrock guardrail assumes a role 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-08-26 18:12:15 +00:00
Mateo Wang
47cfd3f5ca
Merge pull request #38371 from BerriAI/litellm_fix_claude3_1hr_cache_pricing 
fix(model_prices): price 1-hour cache writes on claude-3-haiku and claude-3-opus at 2x input
 2026-08-26 11:11:46 -07:00
Mateo Wang
a6754fa619
Merge pull request #38272 from daniel-meismer-zocdoc/litellm_hidden_alias_explicit_lookup 
fix(router): resolve hidden aliases for explicit lookup
 2026-08-26 10:56:26 -07:00
Mateo Wang
d46bcde2b5
Merge pull request #37922 from BerriAI/litellm_techdebt_20260822 
chore(typing): roll up the daily tech debt cleanups from Aug 20 to Aug 26
 2026-08-26 10:54:50 -07:00
mateo-berri
764048750e fix(mcp): name invalid_token challenges by the caller's requested spelling 2026-08-26 10:52:28 -07:00
mateo-berri
b349b9bf50 fix(pricing): add azure gpt-5.6 cache write rates and correct data zone priority 2026-08-26 10:50:44 -07:00
mateo-berri
6416a97a4d fix(model_prices): price 1-hour cache writes on claude-3-haiku and claude-3-opus at 2x input 2026-08-26 10:48:49 -07:00
mateo-berri
dcffd1da52 refactor(redis): drop docstrings restating the code 2026-08-26 10:45:56 -07:00
mateo-berri
a2cd2d8a4b fix(cost-map): add US data residency uplift to claude-sonnet-4-6 and mythos entries 2026-08-26 10:43:43 -07:00
mateo-berri
a3f654719e fix(proxy): strip the JWT that authenticated on credential-less Vertex passthrough 2026-08-26 10:43:04 -07:00
mateo-berri
b97d5e77eb fix(model_prices): raise bedrock_mantle gpt-5.5 and gpt-5.4 max_input_tokens to Mantle's enforced 1050000 2026-08-26 10:42:00 -07:00
Mateo Wang
95285c3433
Merge pull request #38211 from eugene-yao-zocdoc/litellm_anthropic_responses_strictness_pr 
fix(anthropic-responses): preserve structured output strictness
 2026-08-26 10:39:45 -07:00
devin-ai-integration[bot]
c11c654b8e
fix(proxy): honor DATABASE_DISABLE_PREPARED_STATEMENTS in componentized entrypoints (#38363) 
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-08-26 10:39:04 -07:00