scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-04 02:31:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 69
52992 commits 3112 branches 1436 tags 4.4 GiB
Commit graph

52992 commits

This branch
This branch
All branches
Author SHA1 Message Date
Mateo Wang
7283293d83
Merge pull request #41138 from BerriAI/litellm_bedrock_files_s3_endpoint_url 
fix(bedrock): carry s3_endpoint_url and s3_region_name into file content downloads
 2026-09-18 15:04:01 -07:00
Moe Khalil
b9e5bb3abb test(proxy): allow JEV dependency in budget fixtures 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 22:03:37 +00:00
joshua
19ef8e47a6 feat(ui): link MCP Servers page to the user's connected MCP servers 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 22:01:56 +00:00
yucheng-berri
84df4c0d1b
Merge pull request #41783 from BerriAI/litellm_rate_limit_fallback_guardrails 
fix(proxy): keep requested model guardrails and key disable_fallbacks on rate-limit fallback
 2026-09-18 14:59:16 -07:00
mateo-berri
55249c7128 fix: set vertex gemma-4-26b-a4b-it-maas context window to 262144 2026-09-18 14:58:58 -07:00
mateo-berri
3edbf60e9c fix(proxy): requeue the daily tag rollup on commit failure without the Redis buffer 2026-09-18 14:58:34 -07:00
mateo-berri
76d1abba72 refactor(responses): map status codes to error codes with a lookup 
Ends _response_error_code in an unconditional return so CodeQL stops flagging mixed explicit and implicit returns. No behavior change: every status maps as before.
 2026-09-18 14:57:15 -07:00
Yucheng He
ca287c1b15 fix(mcp): preserve restricted admin submission fields
Some checks failed
LiteLLM Rust / rust-lint (push) Has been cancelled
Details
LiteLLM Rust / rust-test (push) Has been cancelled
Details
LiteLLM Rust / rust-wheel (push) Has been cancelled
Details
Terraform Modules / fmt, validate, test (aws) (push) Has been cancelled
Details
Terraform Modules / fmt, validate, test (gcp) (push) Has been cancelled
Details
Terraform Provider / gofmt, vet, build, test (push) Has been cancelled
Details
Terraform Provider / Provider endpoints vs proxy OpenAPI schema (push) Has been cancelled
Details
2026-09-18 14:55:31 -07:00
mateo-berri
cf05466a27 fix(gemini): map every documented finishReason and reset per-candidate state 
A content-less candidate is now kept as a choice whenever it carries a
finishReason, with the raw value on the choice's provider_specific_fields.
NO_IMAGE, IMAGE_RECITATION, IMAGE_OTHER and ESCALATION map to content_filter;
UNEXPECTED_TOOL_CALL and MISSING_THOUGHT_SIGNATURE map to stop. The
/v1/responses bridge reports content_filter and refusal as incomplete with
incomplete_details, and tool calls and reasoning no longer leak from one
candidate into the next.
 2026-09-18 14:54:18 -07:00
Yassin Kortam
87694c26ef
Merge pull request #41324 from BerriAI/litellm_daily_global_spend_table 
feat(proxy): add LiteLLM_DailyGlobalSpend key-free rollup for the usage dashboard
 2026-09-18 14:53:08 -07:00
Moe Khalil
7c493ff3b9 test(auto-router): reconcile JEV integration checks 
Co-authored-by: Moe Khalil <moe@berri.ai>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:52:24 +00:00
mateo-berri
e0a74dabd1 Merge remote-tracking branch 'origin/main' into litellm_config_update_rejects_config_owned_keys 2026-09-18 14:51:11 -07:00
Yassin Kortam
47209d37f2
Merge pull request #41882 from BerriAI/litellm_azure_speech_api_base_prefix 
fix(proxy): classify Azure Speech short audio behind a prefixed api base
 2026-09-18 14:50:28 -07:00
yassin
cca7ab8b1b test(mcp): type the REST allowlist test stubs 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:50:16 +00:00
yujonglee
59604b2b19
Merge pull request #41884 from BerriAI/litellm_ocr_test_matrix 
test(ocr): declarative provider x auth x input matrix for tests/ocr_tests
 2026-09-18 14:50:05 -07:00
Yassin Kortam
52d6aab421
Merge pull request #41554 from BerriAI/litellm_deepgram_listen_websocket_passthrough 
feat(passthrough): deepgram streaming /v1/listen WebSocket passthrough with duration-based cost tracking
 2026-09-18 14:48:37 -07:00
yucheng
2a7dcc77b2 test(team): mock the membership upsert the member add now issues on team create 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:47:32 +00:00
yucheng
0e74dd2811 test(team): drop the docstrings from the roster audit event tests 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:46:33 +00:00
yucheng
5e8247a1c0 fix(team): emit audit events for member_delete and role changes and carry the final roster on team create 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:46:33 +00:00
Mateo Wang
d45e04a9fd
Merge pull request #41062 from BerriAI/litellm_mistral_codex_reasoning_effort_client_metadata 
fix(mistral): accept reasoning_effort on all models and drop client_metadata for Codex compatibility
 2026-09-18 14:43:58 -07:00
Yujong Lee
f72b7155ac fix(ocr): map Rust upstream 401/403 to the public auth exceptions 
The httpx.Response built for a Rust upstream failure had no request attached,
so constructing openai.AuthenticationError raised RuntimeError inside the
exception mapper and every bad-key OCR call surfaced as APIConnectionError 500
instead of AuthenticationError 401 (the Python path already returned 401)

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:40:13 +00:00
Devin AI
16d63eaf88 Merge remote-tracking branch 'origin/main' into litellm_fix_tpm_window_reset_sibling_counters 2026-09-18 21:38:26 +00:00
Moe Khalil
969cde4f0c feat(ui): complete JEV auto router configuration and connection probes 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:38:15 +00:00
Moe Khalil
86e079d7a8 feat(auto-router): integrate JEV context and usage accounting 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:38:15 +00:00
ryan-crabbe-berri
595768b54b fix(proxy): narrow project_id without a cast and fold the unbudgeted cases into the budget matrix test
Some checks failed
LiteLLM Rust / rust-lint (push) Has been cancelled
Details
LiteLLM Rust / rust-test (push) Has been cancelled
Details
LiteLLM Rust / rust-wheel (push) Has been cancelled
Details
Terraform Modules / fmt, validate, test (aws) (push) Has been cancelled
Details
Terraform Modules / fmt, validate, test (gcp) (push) Has been cancelled
Details 
The lint job failed on one new typing.cast (LIT006) in the cost callback, and the test-quality gate behind it would have failed next on a test whose only assertion inspected a mock (TQ002). project_id is now narrowed with isinstance, and the zero and negative max_budget cases run through the existing parametrized budget test, which asserts the raised error or a clean admit with no alert
 2026-09-18 14:33:41 -07:00
jesus-berri
ee7d2b5094
Update litellm/proxy/management_endpoints/key_management_endpoints.py 
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
 2026-09-18 14:31:34 -07:00
jesus
8983eefea5 fix(auth): drop redundant cast on team_object in centralized checks 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:30:32 +00:00
Mateo Wang
a6bd779bd1
Merge pull request #39424 from emerzon/litellm_azure_ai_flux_2_flex 
feat(azure_ai): support FLUX.2 flex images
 2026-09-18 14:30:25 -07:00
yassin
0536fb3062 fix(mcp): fail closed on empty JWT claims and gate the REST tool routes on mcp_allowed_clients 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:29:59 +00:00
yassin
0b5b69ea3a fix(deepgram): forward only the first model and language values to /listen 
Authorization and pricing read the first model and language query value, but the raw query was forwarded, so Deepgram (which honours the last repeated value) could be sent a model the key was never allowed. Later duplicates of those two keys are now dropped before the upstream URL is built

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:29:29 +00:00
Yujong Lee
9767878425 test(ocr): replace per-provider OCR test classes with a declarative provider x auth x input matrix 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:27:03 +00:00
ryan-crabbe-berri
a5f6ce7bb1 Merge remote-tracking branch 'origin/main' into litellm_lit_3269_project_spend_tracking 
# Conflicts:
#	tests/test_litellm/proxy/db/test_db_spend_update_writer.py
 2026-09-18 14:22:07 -07:00
ryan-crabbe-berri
1e8b8f7c33 ci(duplicate-check): describe the workspace-write sandbox accurately 2026-09-18 14:21:22 -07:00
ryan-crabbe-berri
a43a4924a6
Merge pull request #40878 from BerriAI/litellm_null_cost_unpriced_deployments 
fix(router): report null cost for unpriced deployments instead of 0
 2026-09-18 14:20:52 -07:00
ryan-crabbe-berri
2332d3f183
Merge pull request #40737 from BerriAI/litellm_logs_user_email_display 
fix(ui): show internal user email in logs table and log detail drawer
 2026-09-18 14:20:16 -07:00
yujonglee
b0b2f13548
Merge pull request #41873 from BerriAI/litellm_rust_exception_type_port 
feat(rust): port exception_type to litellm-core-utils
 2026-09-18 14:17:17 -07:00
ryan-crabbe-berri
1f27c442b4 ci(duplicate-check): let Codex reach GitHub from its sandbox 
Every gh search in the first real runs failed with "error connecting to
api.github.com", so the verdict was always null. The legacy
sandbox_permissions key no longer grants network in read-only mode; the
workspace-write sandbox has a network_access switch that does. Pin the CLI
to the version the prompt was proven on
 2026-09-18 14:17:00 -07:00
mateo-berri
eb96d885ce fix(proxy): end failed responses streams with [DONE] 
Emit data: [DONE] after event: response.failed, and after a late failure
when a terminal event already went out, so OpenAI SDK clients see the
same stream end as a completed response. Restore the lazy OpenAPI
snapshot to its Python 3.12 rendering, which is what CI regenerates.
 2026-09-18 14:16:47 -07:00
mateo-berri
4a951847bb fix(responses): merge deployment litellm_params into native websocket response.create frames 2026-09-18 14:15:39 -07:00
yassin
3449ae9d0d fix(proxy): advance the daily global spend marker in one conditional upsert so overlapping runs cannot rewind it 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:15:32 +00:00
yassin
f1b9642c41 fix(proxy): classify Azure Speech short audio behind a prefixed api base 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:15:01 +00:00
Yucheng He
bbbd03089d fix(mcp): retain viewer submission sanitization precedence 2026-09-18 14:14:28 -07:00
Yucheng He
0b2dd9ba86 fix(mcp): sanitize submissions for restricted admin keys 2026-09-18 14:14:28 -07:00
Yucheng He
f59cd303c6 fix(mcp): preserve scopes through admin server edits 2026-09-18 14:14:28 -07:00
jesus
9cff026bae test(mcp): satisfy patch quality checks 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 14:14:28 -07:00
jesus
3a97dc4d4a fix(mcp): satisfy type-discipline lint 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 14:14:28 -07:00
jesus
25c8926c48 fix(mcp): keep oauth scopes in admin api credential redaction 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 14:14:28 -07:00
Mateo Wang
59c24abcbe
Merge pull request #33101 from BerriAI/litellm_fix_responses_ws_litellm_params_leak 
fix(responses): stop managed Responses WebSocket from leaking litellm_params into provider request body
 2026-09-18 14:11:19 -07:00
yassin
0897b663c5 Merge remote-tracking branch 'origin/litellm_deepgram_listen_websocket_passthrough' into litellm_deepgram_listen_websocket_passthrough 2026-09-18 21:08:45 +00:00
yassin
93d61abfa5 fix(deepgram): refuse /listen sessions that have no streaming price 
A caller could pick a model with only a pre-recorded registry row, or no row at all, and the session would be billed at the pre-recorded rate or logged at zero cost, so budgets did not apply. The route now closes the WebSocket with 1008 before dialing Deepgram unless deepgram/streaming/<model> (or the -multilingual row for language=multi) is an exact registry hit, and the logging handler applies the same check so a registry change under a live session records the duration with no cost instead of a substitute rate

Regression tests cover the route refusal, an operator-supplied streaming row for another model being accepted, and the handler never substituting the pre-recorded rate

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-18 21:08:14 +00:00