BerriAI/litellm v1.108.0

on GitHub

Verify Docker Image Signature

All LiteLLM Docker images are signed with cosign. This release is signed with the key introduced in commit e82a3c6.

Verify using the pinned commit hash (recommended):

A commit hash is cryptographically immutable, so this is the strongest way to ensure you are using the published signing key:

cosign verify \
  --key https://raw.githubusercontent.com/BerriAI/litellm/e82a3c646908ba0a6ee0beddc4310d7a6cac4065/cosign.pub \
  ghcr.io/berriai/litellm:v1.108.0

Verify using the release tag (convenience):

Tags are protected in this repository and resolve to the same key. This option is easier to read but relies on tag protection rules:

cosign verify \
  --key https://raw.githubusercontent.com/BerriAI/litellm/v1.108.0/cosign.pub \
  ghcr.io/berriai/litellm:v1.108.0

Expected output:

The following checks were performed on each of these signatures:
  - The cosign claims were validated
  - The signatures were verified against the specified public key

What's Changed

• fix(a2a): normalize agent card protocolBinding casing before transport match by @devin-ai-integration[bot] in #37917
• fix(interactions): track cost and spend for Google Interactions API requests by @mateo-berri in #33310
• fix(bedrock): stop emitting an empty assistant delta after the finish_reason chunk by @devin-ai-integration[bot] in #36806
• fix(anthropic): reconcile enum with declared type in output_format schema by @dkindlund in #37882
• feat(azure_ai): support entra id / oauth auth on every azure ai foundry route by @devin-ai-integration[bot] in #35415
• fix(ui): boot the UI image as an arbitrary uid by anchoring nginx writes under /tmp by @devin-ai-integration[bot] in #37982
• fix(proxy): parse form-encoded video edit/extension bodies after auth by @Souravrajvi0 in #36513
• fix(anthropic): keep legacy thinking budget_tokens on Claude 4.6 models on /v1/messages by @mateo-berri in #38108
• fix(utils): make prompt_token_calculator count claude models again by @ryan-crabbe-berri in #38130
• fix(proxy): keep every value of a repeated form key, and gate the tests that hid it by @ryan-crabbe-berri in #37908
• fix(health): apply model_info.health_check_params to health check probes by @mateo-berri in #38101
• fix(runwayml): route every generation endpoint and fix video cost tracking by @mateo-berri in #38115
• fix(passthrough): attribute spend and release budget reservation on router-model /vllm and /azure routes by @mateo-berri in #38111
• fix: match OpenAI SDK wire format on image/video routes by @mateo-berri in #38104
• fix(ci): give three unit shards a job deadline that outlasts their pytest budget by @tin-berri in #38139
• feat(ui): add Gemini Family auto-router preset by @tin-berri in #38138
• fix(logging_worker): carry queued tasks across event-loop change instead of dropping them by @mateo-berri in #38144
• feat(proxy): enforce vector-store upload security controls on /v1/rag/ingest by @mateo-berri in #38135
• test(e2e): pin require_managed_files enforcement behind a marker-gated stack phase by @mateo-berri in #38117
• refactor(ui): move the dashboard onto class-variance-authority by @yuneng-berri in #38125
• refactor(utils)!: delete prompt_token_calculator by @ryan-crabbe-berri in #38132
• fix(auto-router): list configured auto-routers in the usage picker before they have traffic by @tin-berri in #38129
• refactor(ui): install the shadcn field primitive by @yuneng-berri in #38126
• fix(complexity_router): keep both ends of a clipped classifier context turn by @tin-berri in #38141
• fix(ci): ignore-list recursive form-field flatteners in recursive_detector by @mateo-berri in #38149
• fix(passthrough): stop leaking the caller's virtual key on credential-less Vertex passthrough by @mateo-berri in #38114
• fix(router): stop copying forwarded credentials into retry breadcrumbs by @mateo-berri in #38133
• feat(e2e): record and replay streamed provider responses chunk-for-chunk by @mateo-berri in #38136
• fix: tolerate stream chunks without a choices key in stream_chunk_builder by @AkshaySasi in #34382
• fix(files): decode x-litellm-model encoded file_id in chat + responses by @hclsys in #29832
• fix(videos): forward uploaded source file on /v1/videos/edits to the provider by @mateo-berri in #38155
• fix(s3_v2): percent-encode object keys once so signed and sent URLs match by @devin-ai-integration[bot] in #38005
• feat(ui): add error-code drilldown for failed requests on caching page by @ryan-crabbe-berri in #38156
• feat(search): add Grounding with Bing Search (bing_grounding) as a search provider by @mateo-berri in #38119
• ci: ban row-rewriting DML from prisma migrations by @yucheng-berri in #37899
• fix(langsmith): keep root-run ids self-consistent so batch ingest stops rejecting header-tagged requests by @yucheng-berri in #38116
• ci(e2e): record the e2e suite weekly and replay it on weekdays with zero egress by @mateo-berri in #38163
• chore(codeowners): unown ui container plumbing and generated files by @ryan-crabbe-berri in #38124
• fix(logging): skip parsing redacted tool call arguments by @devin-ai-integration[bot] in #38169
• feat(complexity_router): bound the classifier context block, not each turn in it by @tin-berri in #38145
• fix(http_handler): dispose aiohttp session when AsyncHTTPHandler is finalized without a running loop by @anmolg1997 in #36670
• fix(proxy): reset a stuck team member's budget by @yassin-berriai in #37971
• fix(anthropic/bedrock): request summarized adaptive thinking for reasoning_effort and use provider thinking token counts by @devin-ai-integration[bot] in #37979
• fix(completion_extras): forward reasoning_effort=max through the Responses API bridge by @mateo-berri in #38222
• feat(vertex_ai): add native Vertex AI Interactions API support by @mateo-berri in #38229
• test(mcp): drain the logging worker after each test so queued callbacks cannot leak into the next test by @ryan-crabbe-berri in #38228
• fix(ui): forward OAuth issuer/authorization/token/registration URLs from the MCP server edit form by @yassin-berriai in #38154
• fix(together_ai): default endpoints to api.together.ai instead of api.together.xyz by @mateo-berri in #38233
• fix(bedrock_mantle): register a Bedrock runtime passthrough config so /bedrock/model//invoke works by @mateo-berri in #38231
• fix(router): resolve provider from api_base in deployment validation and acompletion by @mateo-berri in #38235
• fix(model_prices): raise bedrock_mantle gpt-5.6 max_input_tokens to Mantle's enforced 1050000 by @mateo-berri in #38225
• fix(bedrock_mantle): normalize Codex input item types Mantle rejects by @mateo-berri in #38227
• fix(proxy): store the actual selected model in spend logs for Azure Model Router by @devin-ai-integration[bot] in #37770
• feat(router): per-group supported reasoning efforts with the max level by @tin-berri in #37897
• fix(proxy): stop expected 4xx responses from saturating worker CPU on failure logging by @devin-ai-integration[bot] in #38102
• fix(caching): use upstream RedisCluster on redis-py with per-connection recovery by @devin-ai-integration[bot] in #38171
• perf(auth): drop guaranteed-miss internal-cache Redis read from team object lookup by @devin-ai-integration[bot] in #38073
• fix(together_ai): route chat completions through a dedicated TogetherAIChatConfig by @mateo-berri in #38248
• fix(ui): read reasoning tokens from Responses API output_tokens_details by @yassin-berriai in #37952
• fix(dashboard): don't show a stale provider prompt-cache chip on a response-cache hit by @yassin-berriai in #37951
• fix(ui): render tag-based guardrail mode instead of crashing the guardrails page by @devin-ai-integration[bot] in #37493
• fix(scim): return user_id as Group members[].value on transformed group responses by @devin-ai-integration[bot] in #38161
• fix(scim): preserve existing team memberships when POST /Users adoption carries no groups by @devin-ai-integration[bot] in #38166
• fix(router): support mid-stream fallback for anthropic_messages route type by @yassin-berriai in #38153
• fix(auth): support wildcard prefixes in jwt team_allowed_routes by @devin-ai-integration[bot] in #37756
• feat(models): add missing Together AI serverless models to the cost map by @mateo-berri in #38230
• fix(cerebras)!: add max_retries and extra_headers to get_supported_openai_params by @deepanshululla in #36601
• fix(anthropic): translate tool_result document blocks in the /v1/messages bridge by @mateo-berri in #38251
• fix(team): serialize member_add, member_delete, and delete under the team's advisory lock by @yassin-berriai in #37969
• docs(pr-template): split Caveats bullets into severity tiers and call for plain engineering language by @mateo-berri in #38252
• fix(anthropic): carry tool_result document blocks through the /v1/messages responses bridge by @mateo-berri in #38261
• fix(together_ai): pass tools through for models missing from the registry by @mateo-berri in #38265
• fix(anthropic): carry user-content document blocks through the /v1/messages responses bridge by @mateo-berri in #38267
• fix(rerank): emit latency and cost headers on /rerank by @devin-ai-integration[bot] in #35419
• perf(streaming): add shared JSONFragmentAccumulator for Vertex and Anthropic by @deepanshululla in #36610
• fix(together_ai): strip internal thinking fields from outbound messages, keep reasoning_content by @mateo-berri in #38275
• fix(router): persist attempted_fallbacks and original_model_group into spend logs metadata by @yucheng-berri in #38107
• fix(logging): redact tool call arguments to valid JSON and preserve null content by @yucheng-berri in #38182
• fix(ui): stack policy flow builder below the popup layer so guardrail options render by @devin-ai-integration[bot] in #38273
• test: gate the test tree on B003 so a test cannot swap os.environ for a plain dict by @ryan-crabbe-berri in #38274
• refactor(repositories): type prisma table access with one generic protocol by @mateo-berri in #38205
• fix(anthropic): buffer streamed responses carrying server-fulfilled tools so retrieval tool calls never reach the client by @mateo-berri in #36245
• test(together_ai): regression suite across chat, responses, and messages surfaces by @mateo-berri in #38283
• feat(logging): add async_post_call_failure_deployment_hook by @deepanshululla in #36657
• chore: bump litellm-enterprise 0.1.59 -> 0.1.60, litellm 1.99.0 -> 1.100.0 by @yuneng-berri in #38243
• test(e2e): cover Together AI reasoning, tool calls, template kwargs, and cost through a live proxy by @mateo-berri in #38286
• fix(logging): keep tracebacks for provider-originated 4xx errors by @mateo-berri in #38296
• chore(ci): promote internal staging to main by @yuneng-berri in #38293
• refactor(ui): install the shadcn alert primitive by @yuneng-berri in #38300
• refactor(ui): re-pull label, textarea, separator and skeleton from the registry by @yuneng-berri in #38302
• feat(prometheus): configure deployment caller identity by @mphilippnv in #38221
• test(e2e): let the Together replayed-reasoning case survive a single provider miss by @mateo-berri in #38314
• fix(otel): map /v1/messages provider errors before failure logging by @mateo-berri in #38310
• fix(exceptions): map upstream status codes for providers with no exception_type branch by @mateo-berri in #38318
• fix(passthrough): record ownership of streamed responses under managed ids by @mateo-berri in #38320
• fix(proxy): encrypt streamed responses ids on /openai/v1/responses and /responses aliases by @mateo-berri in #38325
• test(cost-calc): pin the rate fallbacks inside a tiered-pricing tier by @yuneng-berri in #38327
• test(e2e): cover the Bedrock provider-feature cells customers run by @mateo-berri in #38232
• fix(together_ai): fail open on response_format instead of dropping it for unregistered models by @mateo-berri in #38269
• fix(proxy): honor DATABASE_DISABLE_PREPARED_STATEMENTS in componentized entrypoints by @devin-ai-integration[bot] in #38363
• fix(anthropic-responses): preserve structured output strictness by @eugene-yao-zocdoc in #38211
• chore(typing): roll up the daily tech debt cleanups from Aug 20 to Aug 26 by @devin-ai-integration[bot] in #37922
• fix(router): resolve hidden aliases for explicit lookup by @daniel-meismer-zocdoc in #38272
• fix(ui): keep focus in the add model public name input while typing by @devin-ai-integration[bot] in #38366
• fix(model_prices): price 1-hour cache writes on claude-3-haiku and claude-3-opus at 2x input by @mateo-berri in #38371
• fix(proxy): keep the caller's Google token on credential-less Vertex passthrough under custom auth by @mateo-berri in #38299
• fix(mcp): preserve provider access token lifetime by @daniel-meismer-zocdoc in #38271
• chore(ui): remove stale "New" badges from the dashboard by @yuneng-berri in #38374
• test(cost-estimate): pin the prices and period totals /cost/estimate returns by @yuneng-berri in #38315
• fix(ci): let the mutation workflow find covered lines so it generates mutants by @yuneng-berri in #38305
• fix(model_prices): raise bedrock_mantle gpt-5.5 and gpt-5.4 max_input_tokens to Mantle's enforced 1050000 by @mateo-berri in #38368
• fix(azure/realtime): authenticate realtime websocket with Azure AD token when no api-key by @devin-ai-integration[bot] in #34658
• fix(bedrock): map reasoning_effort to reasoning.effort for OpenAI GPT-5.x on Converse by @6matt in #38279
• test(prometheus): cover caller-identity config failure cases by @yucheng-berri in #38380
• fix(redis): support credential providers across clients by @eugene-yao-zocdoc in #38094
• fix(health): support mode: image_edit in health checks by @devin-ai-integration[bot] in #38291
• fix(cost-map): add US data residency uplift to claude-sonnet-4-6 and mythos entries by @mateo-berri in #38369
• fix(anthropic): raise missing-credential error on /v1/messages passthrough by @devin-ai-integration[bot] in #38240
• fix(mcp): complete DCR bridge OAuth challenges by @daniel-meismer-zocdoc in #37384
• test(proxy): pin the request-validation contracts in proxy/_types.py by @yuneng-berri in #38307
• docs(CLAUDE.md): add pull-before-work rule by @mateo-berri in #38386
• fix(anthropic): scale cache costs by fast mode and trust served speed by @mateo-berri in #38378
• fix(pricing): add azure gpt-5.6 cache write rates and correct data zone priority by @mateo-berri in #38370
• docs: tighten the pull-before-work rule in CLAUDE.md by @devin-ai-integration[bot] in #38389
• fix(health): strip credential fields from GET /health output by @Siraj637909 in #37090
• fix(minimax): attach MINIMAX_API_KEY on anthropic messages requests by @mateo-berri in #38393
• refactor(ui): replace hand-picked z-index values with one named scale and lint it by @ryan-crabbe-berri in #38282
• fix(health): probe Azure GA realtime path for transcription-only models by @mateo-berri in #38390
• fix(bedrock): parse cacheDetails for Converse 1h/5m cache write cost split by @danielva-monday in #36762
• fix(caching): flush async cache writes cancelled at event loop shutdown by @mateo-berri in #38385
• fix(router): resolve model_group_alias before pre-routing strategy dispatch by @tin-berri in #38382
• feat(proxy): enforce rpm/tpm on model add + fix validation error title in UI by @kunal2002 in #36518
• refactor(ui): move every page header onto the shared PageHeader by @yuneng-berri in #38306
• fix(proxy): stop cache eviction errors from failing /key/update by @yuneng-berri in #38308
• fix(aiohttp): honor global ssl_verify on the aiohttp_openai handler path by @mateo-berri in #38400
• fix(logging_worker): rescue dequeued logging tasks lost at event loop close by @mateo-berri in #38394
• fix(caching): require the namespace delimiter when checking already-namespaced redis keys by @mateo-berri in #38403
• fix(prompts): reject keyed prompt_data with prompt_id and populate prompt version by @mateo-berri in #38404
• fix(cost_calculator): resolve real cost key when model_name alias contains '/' by @ksk2023 in #38344
• fix(cost-map): correct prompt_cache_min_tokens for Claude Fable 5 and backfill Anthropic re-export entries by @mateo-berri in #38405
• test(azure-ai): pin the 422 retry that drops the field the provider rejected by @yuneng-berri in #38309
• refactor(ui): read the auto-router tier set through one row list by @tin-berri in #38408
• fix(proxy): stop empty DB router_settings lists from clobbering yaml fallbacks by @mateo-berri in #38406
• fix(team): allow no-reset default budgets for team members by @devin-ai-integration[bot] in #37708
• fix: forward image content lists to DeepSeek vision models by @yucheng-berri in #38397
• fix(fireworks_ai): stop using the trace id as the session affinity key by @Hamjaster in #35754
• fix(gemini-realtime): keep the client's voice on Vertex AI native-audio Live by @yucheng-berri in #38395
• fix(vertex_ai): bill Gemini grounding per unique web search query by @ousamabenyounes in #36397
• fix(health): make the image_edit health probe moderation-safe by @mateo-berri in #38417
• test: gate the test tree on fifteen assertion and handler rules it already satisfies by @ryan-crabbe-berri in #38361
• fix(proxy): key lazy openapi stubs off registered features, not sys.modules by @mateo-berri in #38416
• fix(proxy): derive auto-router health from its underlying models by @tin-berri in #38174
• fix(responses): let cache-control injection reach the system prompt from instructions by @tin-berri in #38120
• fix(gemini): bill Google Maps grounding as its own SKU by @mateo-berri in #38418
• fix(speech): keep proxy metadata and completion cost through the TTS completion bridge by @mateo-berri in #38414
• fix: map Gemini ON_DEMAND_FLEX traffic type to flex service tier by @bisma-nawaz in #37724
• feat(langfuse): support langfuse_environment as a per-key dynamic callback param by @devin-ai-integration[bot] in #38264
• feat(proxy): hide unhealthy models from model listings, opt-in by @yuneng-berri in #38313
• fix(mcp): honor admin-entered OAuth URLs on authorize after issuer yield by @mateo-berri in #38379
• fix(model_prices): correct gemini-3.5-flash-lite flex cache-read pricing by @mateo-berri in #38422
• fix(cost): price gemini-live-2.5-flash-native-audio realtime sessions by @mateo-berri in #38419
• fix(cost-map): correct Gemini TTS and native-audio rates by @mateo-berri in #38412
• fix(prompts): propagate PATCHed prompt templates to every worker and pod by @mateo-berri in #38411
• fix(model_prices): bill gemini -latest/preview alias cache reads at 10% of input by @mateo-berri in #38423
• fix(proxy): sync search tools into the router on management writes by @yuneng-berri in #38392
• feat(guardrails): track Azure Prompt Shield usage and cost with spend isolation by @yucheng-berri in #38387
• fix(prompts): apply prompt templates before routing on /v1/responses and honor ignore_prompt_manager_model by @mateo-berri in #38407
• fix(cost): make cost-breakdown headers respect service tier by @mateo-berri in #38424
• fix(mcp): add litellm[mcp] extra and actionable error when streamable_http_client is missing by @mateo-berri in #38399
• revert(proxy): remove router_model_name from auto-routed response bodies by @tin-berri in #38429
• fix(google_genai): price streamed generateContent with the provider that served it by @devin-ai-integration[bot] in #36055
• fix(logging): stop billing and logging response reads as LLM calls by @devin-ai-integration[bot] in #36890
• fix(budget): serialize model_max_budget before the /budget/update write by @yuneng-berri in #38430
• fix(ui): block the auto-router submit on a missing classifier model and an orphaned keyword rule by @tin-berri in #38427
• feat(complexity_router): heuristic-first classifier chaining by @tin-berri in #38428
• test(e2e): un-skip the per-model budget update case by @yuneng-berri in #38437
• fix(cost): stop double-billing cached tokens that overlap a modality by @Srivatsa03 in #37407
• feat(ui): add Teams list CSV export with budgets, model grants, and rate limits by @devin-ai-integration[bot] in #38436
• fix(mcp): accept raw x-litellm-api-key on streamable HTTP admission by @devin-ai-integration[bot] in #38364
• fix: bound row count on GET /spend/logs to stop unbounded LiteLLM_SpendLogs scans by @devin-ai-integration[bot] in #38420
• fix(scim): apply default_team_params (incl. models) to SCIM-created teams by @devin-ai-integration[bot] in #38433
• fix(prompts): propagate prompt deletes to every worker and pod by @mateo-berri in #38434
• fix(anthropic_adapter): carry web search cost into /v1/messages breakdown headers by @mateo-berri in #38439
• fix(ui): show custom technical keywords on every router whose scorer runs by @tin-berri in #38451
• fix(e2e): move the vertex realtime suite off the retired Live preview model by @yuneng-berri in #38454
• feat(newrelic): per-team cost and usage metrics via team callbacks by @yucheng-berri in #37610
• fix(ui): carry a preset's per-tier litellm_params through the prefill by @tin-berri in #38453
• fix(e2e): size the mid-conversation-system cache prefix above the minimum deterministically by @yuneng-berri in #38468
• fix(e2e): disable thinking on the gemini chat cost test instead of racing its budget by @yuneng-berri in #38469
• feat(ui): put the auto-router savings hero on a spend rail and a four-tile row by @tin-berri in #38470
• feat(ui): toggle internal health check visibility in request logs by @mateo-berri in #38391
• fix(mcp): canonicalize bearer scheme on bridge egress by @daniel-meismer-zocdoc in #38398
• refactor: clean up fresh tech debt from 2026-08-27 window by @devin-ai-integration[bot] in #38484
• fix(exception_mapping_utils): map unmapped exceptions when model and provider are unset by @mateo-berri in #38496
• fix(ui_sso): resolve highest privilege Entra app role, not first in claim by @imranismail in #36728
• fix(proxy): regenerate lazy OpenAPI snapshot and guard it in CI by @mateo-berri in #38410
• feat(ui): add cache hit/miss filter to Request Logs by @devin-ai-integration[bot] in #38432
• fix(bedrock): sign rerank requests with the shared header-filtered SigV4 helper (internal copy of #36462) by @mateo-berri in #38093
• fix(bedrock): sign rerank requests with the shared, header-filtered SigV4 helper by @noahnistler in #36462
• fix(ui): order the auto-routers table newest first so a new router lands on page one by @tin-berri in #38545
• feat(ui): run the Anthropic Family preset's reasoning tier on Opus 5 at high thinking by @tin-berri in #38490
• build(ui): bump nginx to 1.31-alpine by @devin-ai-integration[bot] in #38541
• feat(otel): support per-team/per-key service.name for OTel v2 destinations by @devin-ai-integration[bot] in #38532
• test(e2e): de-flake the cost-header cache read and the router fallback control by @yuneng-berri in #38435
• feat(gemini): day-0 support for gemini-3.5-transcribe and transcribe-live by @mateo-berri in #38540
• feat(health): opt-in model-group allowlist for background health checks and health-check routing by @mateo-berri in #38539
• fix(mcp): keep upstream OAuth Authorization when jwt signer hook injects one on tools/call by @devin-ai-integration[bot] in #38555
• fix: suppress misleading register_model unresolved-cost warnings for entries without custom pricing by @devin-ai-integration[bot] in #38542
• feat(proxy): opt-in budget rollover carrying overage into the next window by @devin-ai-integration[bot] in #38514
• fix(auth): skip guaranteed-miss team lookup for the litellm-dashboard sentinel by @devin-ai-integration[bot] in #38471
• fix(key_management): allow /key/update to keep or shrink MCP server grants the key already holds by @devin-ai-integration[bot] in #38463
• fix: keep schema reconciliation from fighting a partitioned LiteLLM_SpendLogs by @devin-ai-integration[bot] in #38452
• fix(ui): open select popups below the trigger instead of over it by @tin-berri in #38554
• fix(realtime): bill trailing audio when a Gemini transcribe Live session closes by @mateo-berri in #38563
• test(e2e): cover key generate and update on the Admin UI path by @yuneng-berri in #38448
• chore: bump litellm-enterprise 0.1.60 -> 0.1.61, litellm-proxy-extras 0.4.89 -> 0.4.90 by @yuneng-berri in #38566
• fix(ui): let the paginated search select keep what the user types by @yuneng-berri in #38475
• fix(otel): anchor MCP tool-call spans to the gateway's own trace, link the client's context by @yucheng-berri in #38317
• fix: roll up the open deflake fixes for the MCP logging queue, PTU rollup, license gate, and pricing test isolation by @devin-ai-integration[bot] in #37833
• fix(model_prices): rolling registry audit - verified models and rates for Novita, DeepInfra, W&B, Bedrock Sol, Gemini, Fireworks, Azure gpt-5.6, Mistral, Together by @devin-ai-integration[bot] in #38207
• test(together_ai): assert fail-open supported params for models missing from the registry by @mateo-berri in #38487
• test(e2e): let the together tool tests accept parallel calls by @yuneng-berri in #38567
• feat(mcp): let a resolved OAuth token target a custom upstream header by @tin-berri in #38456
• feat(together_ai): map reasoning_effort per model class by @mateo-berri in #38263
• feat(dashscope): support qwen-image-3.0 and qwen-image-3.0-pro image generation by @devin-ai-integration[bot] in #38449
• fix(cost): apply Together AI cache read pricing and per-model registry rates by @mateo-berri in #38280
• fix(guardrails): forward aws_external_id when the bedrock guardrail assumes a role by @devin-ai-integration[bot] in #38376
• fix(transcription): synthesize srt/vtt output for adapters without native subtitle formats by @mateo-berri in #38561
• fix(streaming): preserve provider service-tier metadata so Vertex flex streams bill at flex rates by @mateo-berri in #38458
• fix(realtime): bill Gemini Live native-audio output tokens at the audio rate by @mateo-berri in #38457
• fix(anthropic): carry tool_reference tool results through the guardrail translation round trip by @mateo-berri in #38465
• fix(anthropic-adapter): pass provider-native and OpenAI-format tools through on /v1/messages by @mateo-berri in #38431
• test(e2e): serve the vision image from our own fixture by @yuneng-berri in #38575
• feat(together_ai): add zai-org/GLM-5.3-Flash to the model registry by @mateo-berri in #38486
• fix(ui): stop server-searched comboboxes from clobbering picks and queries by @ryan-crabbe-berri in #38574
• feat(model_prices): let a map entry declare its exact reasoning_effort levels by @tin-berri in #38481
• fix(anthropic): carry the adaptive effort tier to every bridged Claude target by @tin-berri in #38533
• feat(alerting): add native Microsoft Teams alerting destination by @devin-ai-integration[bot] in #38367
• chore(proxy): resync the generated API artifacts with the current models by @tin-berri in #38587
• fix(router): reject complexity-router settings written outside complexity_router_config by @tin-berri in #38570
• feat(ui): session-level cache observability in request logs by @devin-ai-integration[bot] in #38442
• fix(ui): link Virtual Keys hint through the migrated /ui route by @devin-ai-integration[bot] in #38596
• fix(anthropic): carry the effort tier only where the target declares reasoning_effort by @tin-berri in #38592
• fix(presidio): chunk oversized text before /analyze so large content blocks do not fail by @yucheng-berri in #38483
• fix(logging): stop stream-based log collectors classifying INFO logs as errors by @yucheng-berri in #38476
• feat(ui): dry-run an auto-router config against the backend before saving it by @tin-berri in #38595
• fix(guardrails): add fail-open mode to CrowdStrike AIDR guardrail by @yucheng-berri in #38568
• fix(router): copy instead of mutating caller metadata when scrubbing fallback stamp keys by @yucheng-berri in #38586
• feat(proxy): opt-in enforce_fallback_model_access authorizes router fallbacks against the calling key by @ryan-crabbe-berri in #38572
• fix(langfuse): warn and drop invalid LANGFUSE_TRACING_ENVIRONMENT instead of failing requests by @yucheng-berri in #38582
• fix(tencent): route thinking through extra_body in chat completions by @FelipeRodriguesGare in #38100
• test-check-commits by @nickhac in #36344
• feat(proxy): dry-run a real request body on /auto_router/test_routing by @tin-berri in #38590
• fix(shadow_eval): refuse a judge model that also serves one of the arms it grades by @tin-berri in #38589
• fix(anthropic): resolve /v1/messages effort tiers through the capability owner by @tin-berri in #38492
• fix(router): fall over on raised mid-stream errors in /v1/messages streams by @mateo-berri in #38606
• feat(models): add daily Together AI model registry sync script and workflow by @mateo-berri in #38257
• feat(ui): the model and wire layer for operator-defined auto-router tier sets by @tin-berri in #38602
• fix(moonshot, together_ai): send the reasoning effort Kimi K3 accepts by @tin-berri in #38611
• fix(ui): one-click theme toggle and matching Docs/Blog styling in the top bar by @ryan-crabbe-berri in #38601
• feat(proxy): opt-in flags to require rpm/tpm on model and project create by @ansh-agrawal in #36514
• fix(exceptions): keep a refused connection an APIConnectionError by @yuneng-berri in #38624
• chore(ci): promote internal staging to main by @yuneng-berri in #38616
• fix(anthropic): drop and self-heal empty thinking blocks on /v1/messages by @tin-berri in #38625
• fix(anthropic): handle per-level reasoning_effort flags without supports_reasoning by @tin-berri in #38618
• fix(complexity_router): route client housekeeping calls to the cheapest tier by @tin-berri in #38598
• test: fix staging CI regressions from #38182, #38144, #38265, #37962, and #37969 by @mateo-berri in #38304
• feat(spend): report prompt caching savings as total and gateway-attributed by @tin-berri in #38134
• fix(proxy): let llm_api virtual keys read /model_group/info by @devin-ai-integration[bot] in #38662
• feat(ui): edit the auto-router tier set with custom classifier-defined tiers by @tin-berri in #38603
• fix(proxy): count tools, system, and Anthropic image and document blocks in the count_tokens fallback (internal copy of #36671) by @mateo-berri in #38657
• test: refresh the suites that drifted from langfuse and OpenAI's retired Assistants API by @yuneng-berri in #38637
• test(e2e): unskip four tests whose blockers no longer hold by @yuneng-berri in #38640
• feat(proxy): add paginated GET /public/v1/model_hub by @yuneng-berri in #38636
• refactor(ui): type search tool params from the generated schema by @yuneng-berri in #38633
• feat(a2a): semantic search over the agent registry via GET /v1/agents?query and an agent_search MCP tool by @mateo-berri in #38609
• fix(model_prices): add bedrock_mantle gpt-5.5/5.4 272K tiers, align sol with AWS invoice by @mateo-berri in #38615
• fix(ui): keep the usage filter visible when the caller's scope is empty by @yuneng-berri in #38581
• fix(registry): add Gemini Omni 1.1 Flash, xAI grok-imagine image models, Mistral cache-read pricing, GLM 5.3 Flash + Kimi K2.7 Code entries by @devin-ai-integration[bot] in #38560
• fix(logging): preserve null end user in callbacks by @yucheng-berri in #38642
• test: close mutation-testing gaps in container, skills and openai-like config factories by @yuneng-berri in #38677
• fix: enforce MCP toolsets attached to a team, org, or internal user by @yucheng-berri in #38488
• fix(tests): drain the global logging worker in RAG aquery billing tests by @devin-ai-integration[bot] in #38653
• chore(techdebt): type new signatures and drop slop comments from the last 24h by @devin-ai-integration[bot] in #38644
• feat(auto_router): write and preview the classifier prompt an edited tier set sends by @tin-berri in #38605
• fix(proxy): price the /v1/messages cost header by the deployment model, not the client alias by @mateo-berri in #38691
• fix(streaming): preserve provider model for cost calculation by @aaaaaandrew in #38656
• feat(models): sync together_ai model registry by @github-actions[bot] in #38694
• test(e2e): cover Together reasoning_effort=none, json_schema, and cache-read pricing by @mateo-berri in #38697
• fix(model_info): map supports_parallel_function_calling into ModelInfo by @mateo-berri in #38692
• test(litellm_utils_tests): give the aiohttp transport tests teeth by @devin-ai-integration[bot] in #38066
• fix(proxy): use default_factory for _BudgetCascade.rollover_caps by @yassin-berriai in #38687
• feat(guardrails): add Lakera v2 skip-message honoring and advisory (inject_system_message) mode by @deepanshululla in #34940
• fix(router): drop a tier param the routed target cannot take by @tin-berri in #38622
• fix(router): scrub fallback stamp keys in place and strip them at the proxy boundary by @yucheng-berri in #38690
• fix(shadow-eval): validate Anthropic SDK judge credentials by @tin-berri in #38701
• feat(ui): run the Lite preset's medium and complex tiers at their documented efforts by @tin-berri in #38482
• chore: drop stray learnings.md workflow note from CLAUDE.md by @yucheng-berri in #38704
• fix(spend): remove the proxy-wide autorouter savings baseline override by @tin-berri in #38700
• fix(router): tier-pinned reasoning_effort supersedes client effort carriers by @tin-berri in #38698
• fix(proxy): reset a key's budget-window counters on spend reset by @yassin-berriai in #38686
• feat(shadow_eval): measure both arms' cost so a job reports what the router would have saved by @tin-berri in #38631
• fix(gpt-5): stop forwarding temperature and top_p to reasoning models that reject them by @tin-berri in #38593
• fix(streaming): report response_cost and Anthropic citations from stream_chunk_builder by @mateo-berri in #38696
• feat(ui): make provider logos readable in dark mode by @yuneng-berri in #38588
• feat(terraform): resource and data source parity with the community provider by @shivamrawat1 in #38158
• test(shadow_eval): configure the anthropic sdk judge in the funnel-seed test by @yucheng-berri in #38717
• fix(ci): fall back to github.token when the GH_TOKEN secret is unset in the Together sync workflow by @mateo-berri in #38695
• feat(terraform): add litellm_jwt_key_mapping resource by @devin-ai-integration[bot] in #38714
• feat(terraform): coverage-enforcing CI gate against the latest OpenAPI spec by @devin-ai-integration[bot] in #38710
• fix(terraform): refresh jwt key mapping allowlist entries now that the resource is merged by @devin-ai-integration[bot] in #38720
• fix(guardrails): record post_call scans on native /v1/messages streams by @mateo-berri in #38713
• chore(deps): raise RestrictedPython floor to 8.5 by @yuneng-berri in #38715
• fix(proxy): trigger async_pre_call_hook on POST /v1/files uploads by @devin-ai-integration[bot] in #38607
• test(responses): adapt temperature tests to the gpt-5 reasoning validation by @mateo-berri in #38748
• fix(router): pin batch, file, and fine-tuning job ids to their owning model group on fallback by @mateo-berri in #38742
• fix(bedrock): map real batch record counts and guard zero-count retire by @mateo-berri in #38744
• fix(aws): forward aws_external_id in Bedrock embeddings and SageMaker credential loading by @mateo-berri in #38727
• fix(aws): build every AWS endpoint and ARN from the region's partition (aws-cn, aws-us-gov) by @mateo-berri in #38747
• refactor(types): replace Any with real types across 178 backend files by @mateo-berri in #38501
• feat(hosted_vllm): add vLLM-Omni videos API by @mubashir1osmani in #38148
• fix(databricks): derive OAuth token URL from workspace origin by @ump45nose in #35940
• test(e2e): add logging e2e coverage (s3_v2, gcs_bucket, team langfuse callback, datadog failure) by @yucheng-berri in #38552
• fix(batches): fill a managed batch page past rows that will not parse by @mateo-berri in #38738
• fix(proxy): tag routing misses proxy-merged tags when chat requests carry litellm_metadata by @mateo-berri in #38739
• fix(health): honor allow_requests_on_db_unavailable in readiness probe by @devin-ai-integration[bot] in #37640
• feat(mcp): support asymmetric (RS256) signing for MCP gateway session tokens by @devin-ai-integration[bot] in #38728
• fix(ui): make code blocks follow the theme in dark mode by @yuneng-berri in #38771
• perf(rust): use pythonize for bridge serialization by @yujonglee-berri in #38764
• fix(policy): let the AI policy suggester drop sampling params its model refuses by @tin-berri in #38594
• feat(ui): link team and key model chips to the models page filtered to that group by @ryan-crabbe-berri in #38626
• fix(ui): make the logs JSON viewer follow the theme in dark mode by @yuneng-berri in #38778
• chore: bump litellm-enterprise 0.1.61 -> 0.1.62, litellm-proxy-extras 0.4.90 -> 0.4.91 by @yuneng-berri in #38779
• refactor: clean up tech debt that landed on 2026-08-29 by @devin-ai-integration[bot] in #38743
• fix(managed resources): let keys with no user_id or team_id read their own batches and files by @devin-ai-integration[bot] in #34849
• fix(proxy): pin model reconcile read to the writer DB so /model/new does not 500 under read replica lag by @devin-ai-integration[bot] in #38580
• fix(ui): restore the reopen control for the log drawer's trace sidebar by @yuneng-berri in #38782
• fix(azure): use /openai/v1 image routes for v1, preview and latest api versions by @devin-ai-integration[bot] in #38285
• fix(soniox): align synthesized SRT/VTT cues to real speech timing by @dan2k3k4 in #34440
• fix(bedrock): route all cohere.embed models to the cohere embedding config by @devin-ai-integration[bot] in #38670
• feat(mcp_gateway): add RFC 7662 introspection for gateway session tokens by @devin-ai-integration[bot] in #38726
• fix(headroom): resolve CCR retrieval on streaming /chat/completions by @devin-ai-integration[bot] in #35017
• fix: bound Hugging Face config fetch and keep embedding tests off the network by @devin-ai-integration[bot] in #38752
• feat(vertex-ai): add veo 3.1 lite model metadata by @emerzon in #30782
• feat(proxy): cyberark conjur secret manager configuration via Admin UI by @devin-ai-integration[bot] in #38445
• ci: build the benchmark environment outside the CodSpeed runner by @devin-ai-integration[bot] in #38426
• fix(proxy): give every requests call a timeout so a silent server cannot hang the caller by @ryan-crabbe-berri in #38234
• fix(batches): aggregate reasoning tokens and per-line pass/fail counts by @mubashir1osmani in #37208
• fix(ui): allow in-place editing of classifier numeric inputs by @tin-berri in #38803
• fix(anthropic): emit signature-only thinking blocks on the /v1/messages bridge by @tin-berri in #38809
• fix(ui): keep a deleted-from search query instead of blanking the box by @yuneng-berri in #38830
• fix(proxy): run SMTP send_email off the event loop with a connection timeout by @devin-ai-integration[bot] in #38473
• feat(budgets): enforce shared budgets on model access groups by @ryan-crabbe-berri in #38784
• feat(vertex_ai): support gemini-3.5-transcribe on /v1/audio/transcriptions by @mateo-berri in #38740
• fix(anthropic_endpoints): serialize dict-detail HTTPExceptions on /v1/messages like sibling surfaces by @mateo-berri in #38741
• fix(proxy): count auto-router classifier cost in savings and benchmarks by @tin-berri in #38835
• feat(proxy): add LiteLLM_BudgetWindowSpend table for per-window budget spend by @ryan-crabbe-berri in #35854
• feat(policy_engine): execute post_call guardrail pipelines on streaming responses by @mateo-berri in #38788
• fix(models): registry audit: new Together/Fireworks/Gemini/Mistral/xAI models, xai retirement repricing, bedrock grok-4.6 caching, deprecation dates by @mateo-berri in #38804
• feat(proxy): maintain per-window budget spend rows in the spend writer by @ryan-crabbe-berri in #35886
• perf(proxy): read budget-window spend from the maintained window table by @ryan-crabbe-berri in #35887
• fix(together_ai): stop writing context_length as max_output_tokens in the serverless sync by @mateo-berri in #38820
• fix(openai): flatten top-level anyOf/oneOf/allOf in Responses API tool schemas by @mateo-berri in #38792
• fix(anthropic): cap reasoning_effort thinking budget below max_tokens on /v1/messages by @mateo-berri in #38836
• fix(azure): flatten top-level tool schema combinators for Azure Responses GPT-4-family deployments by @mateo-berri in #38837
• chore: rebuild Admin UI bundle for the next release by @yuneng-berri in #38846
• fix(proxy): list all non-team models for users with an empty model list by @devin-ai-integration[bot] in #38249
• fix(guardrails): stop Lakera monitor mode forwarding unmasked PII on Responses-API bodies by @yucheng-berri in #38841
• feat(mcp): bulk-import Anthropic MCP connectors via API and admin UI by @devin-ai-integration[bot] in #38444
• feat(ui): set a model access group's shared budget from the dashboard by @ryan-crabbe-berri in #38843
• fix(ui): drop stray text next to Close in the model connection test dialog by @yuneng-berri in #38852
• feat(helm): make extra ingress paths configurable in the componentized chart by @yassin-berriai in #35700
• fix(otel/v2): detach credential-routed tenant spans into their own trace by @yucheng-berri in #38847
• test(e2e): stop the reliability fallback tests flaking on gpt-5.5's reasoning budget by @yuneng-berri in #38833
• test(e2e): retry upstream-saturation failures in the claude CLI driver by @yuneng-berri in #38850
• chore(ci): promote internal staging to main by @yuneng-berri in #38805
• fix(proxy): emit timing headers and overhead for /v1/messages and /v1/responses by @yucheng-berri in #38840
• fix(tests): derive the no-cache-read-rate savings baseline from the model map by @tin-berri in #38863
• chore(typing): clear Any seams across 47 files, ratchet basedpyright ceilings -3,302 by @mateo-berri in #37778
• chore(typing): clear 1.2k basedpyright Any errors across 16 hotspot files by @mateo-berri in #36722
• feat(bedrock): honor streaming buffer/sampling config for unbuffered post_call scans by @mateo-berri in #38722
• feat(cli): set ENABLE_TOOL_SEARCH=true for lite claude by @mateo-berri in #38942
• fix(proxy): deliver budget alerts on webhook-only alerting and accept ALERTING_WEBHOOK_URL by @devin-ai-integration[bot] in #38441
• docs(claude.md): require tests to check behavior, not code structure by @ryan-crabbe-berri in #38772
• chore(newrelic): cover static default_team_settings per-team routing by @yucheng-berri in #38857
• fix: update stale source URLs and deprecation dates in model cost map by @devin-ai-integration[bot] in #38801
• feat(ci): close duplicate issues after a 3-day grace period by @mubashir1osmani in #38381
• docs(proxy): clarify spend semantics on /v2/user/info and /user/daily/activity by @devin-ai-integration[bot] in #38883
• fix(guardrails): configure Prompt Security file timeout policy by @davida-ps in #38083
• fix(bedrock): stop duplicating Converse config blocks inside inferenceConfig by @mateo-berri in #38993
• fix(guardrails): exclude images from HiddenLayer v1 scans by @Ashton-Sidhu in #29210
• feat(spend_tracking): persist router metadata in spend logs for internal router models by @mateo-berri in #39001
• fix(vertex_ai): graft default vertex path when api_base has a version-only path by @devin-ai-integration[bot] in #38986
• fix(proxy): allow unblocking customers via /customer/update by @cat0825 in #34696
• feat(openai): support workload identity federation (OIDC token exchange) by @mateo-berri in #38995
• fix(otel): emit cache token counts on OTel v2 LLM spans by @devin-ai-integration[bot] in #38716
• feat(proxy): add /v1/responses/input_tokens token counting endpoint by @mateo-berri in #38997
• fix(docker): bump wolfi-base for glibc 2.44 and pin apk python to 3.13 by @mateo-berri in #38917
• fix(docker): bump wolfi-base for glibc 2.44 and pin apk python to 3.13 in migrations image by @mateo-berri in #38973
• feat(friendli): add zai-org/GLM-5.3-Flash model pricing by @Lee-Si-Yoon in #38880
• chore(techdebt): clear fresh debt from the 2026-08-29 and 2026-08-30 windows by @devin-ai-integration[bot] in #38884
• fix(bedrock): surface Nova Sonic user transcripts, speech events, and usage in realtime API by @devin-ai-integration[bot] in #38597
• fix(guardrails): carry Anthropic url image sources through to guardrails by @samtsai15 in #38940
• feat(friendli): add zai-org/GLM-5.3 model pricing by @Lee-Si-Yoon in #38881
• fix(router): apply model renames to the in-memory deployment list by @yatishgoel in #38479
• test(e2e): cover SCIM token creation and SCIM API auth in the Admin UI suite by @yuneng-berri in #39027
• feat(gigachat): add native API passthrough routes with spend logging by @mateo-berri in #38913
• feat(gigachat): add passthrough gigachat route by @KnyazSh in #25886
• feat(complexity-router): add classification_mode to skip classifier on continuation turns by @tin-berri in #38861
• fix(proxy): preserve model table columns on master key rotation by @mateo-berri in #38878
• fix(speech): stop forwarding response_format as a chat param for Gemini TTS by @mateo-berri in #38819
• fix(proxy): return 200 from /model/block and /model/unblock instead of 500 by @mateo-berri in #38873
• feat(complexity_router): escalate oversized prompts to a tier that fits before dispatch by @tin-berri in #38844
• feat(shadow_eval): target teams and users so JWT-auth traffic can be evaluated by @tin-berri in #39015
• fix(anthropic_messages): drain upstream in a detached pump so client … by @nuernber in #36008
• refactor(proxy): bound the budget window seed by time instead of request ids by @ryan-crabbe-berri in #38851
• fix(proxy): ship psycopg so partitioned SpendLogs detection actually runs by @yassin-berriai in #38994
• test(e2e): assert user-observable behavior instead of DOM structure by @yuneng-berri in #39016
• build(rust): configure native extension profiles by @yujonglee-berri in #39020
• fix(ui): keep litellm_credential_name from LiteLLM Params JSON when no credential is selected by @devin-ai-integration[bot] in #39005
• Revert "fix(ui): keep litellm_credential_name from LiteLLM Params JSON when no credential is selected" by @yassin-berriai in #39046
• fix(auth): quiet malformed virtual key rejections to stdout by @yucheng-berri in #38838
• fix(proxy): wire team-level logging callbacks into passthrough endpoints by @yucheng-berri in #38979
• feat(complexity_router): opt-in modality-based capability routing for image requests by @tin-berri in #39032
• fix(ui): let the auto-router scoring tier list follow the theme by @tin-berri in #39040
• feat(ui): auto-router controls for context-window escalation by @tin-berri in #39054
• fix(redis): coerce env var string types and fix param discovery through decorator wrappers by @koladefaj in #30644
• feat(key management): show budget window usage on /key/info by @Thijmen in #37044
• fix(websearch): reject invalid explicit search tool selections by @georgeatparallel in #38113
• feat(shadow_eval): compare several auto-routers on one job's sampled traffic by @tin-berri in #39028
• fix(speech): honor pcm/wav response_format for Gemini TTS and reject unsupported containers by @mateo-berri in #38868
• fix(proxy): match /v1/audio/speech content-type to the returned audio format by @mateo-berri in #38798
• test(e2e): drop the two mgmt registry cells no shared-proxy test can cover by @ryan-crabbe-berri in #39055
• feat(ui): one classification frequency picker for complexity auto-routers by @tin-berri in #39042
• test(e2e/ui): automate 8 manual QA checklist flows by @yuneng-berri in #39025
• fix(key_management): allow non-admin key_type preset transitions on /key/update by @yuneng-berri in #39051
• chore(typing): clear 1.1k basedpyright Any errors across 53 backend files by @mateo-berri in #38796
• fix(openai): forward reasoning_effort for unknown model aliases instead of failing closed by @mateo-berri in #39065
• test(e2e-ui): poll credential availability before Test Connect to deflake multi-instance runs by @yuneng-berri in #39073
• feat(ui): modality routing toggle on the auto-router create and edit forms by @tin-berri in #39059
• fix(proxy): include litellm_model_table in GET /v2/team/list by @yassin-berriai in #39045
• fix(bedrock): mask signed request headers in guardrail debug log by @mateo-berri in #39044
• fix(bedrock): forward aws_external_id in files and batches credential loading by @mateo-berri in #39066
• fix(mcp): persist alias MCP grants verbatim instead of rewriting to local server ids by @ryan-crabbe-berri in #39119
• fix(responses): json-encode object tool call arguments in the chat completions bridge by @devin-ai-integration[bot] in #35417
• fix(cost): bill OCR annotation pages via annotation_cost_per_page by @mateo-berri in #38985
• fix(policy_engine): restore request guardrails list after pipeline allow by @mateo-berri in #39038
• fix(embeddings): omit encoding_format when the client omits it on OpenAI-compatible calls by @mateo-berri in #38774
• test: deflake MCP registry state, savings cost map, and MCP identity env reload tests by @devin-ai-integration[bot] in #38891
• feat(helm): add Argo CD PreSync hook and rollout strategy knobs to the componentized chart by @devin-ai-integration[bot] in #39112
• fix(registry): veo 3.1 pricing tiers + roll up open registry PRs (glm-5.2, Qwen3.8-Flash, gemma-4-31b, scribe_v2, fireworks/databricks deepseek v4) + deprecation dates by @devin-ai-integration[bot] in #38990
• test(ui): budget DOM-structure assertions in dashboard tests by @yuneng-berri in #39082
• test(ui): assert DataTable behavior instead of DOM structure by @yuneng-berri in #39084
• test(ui): query the screen instead of the render result by @yuneng-berri in #39085
• fix(ui): stop checkboxes stretching to the full width of a form field by @yatishgoel in #39108
• chore: bump litellm-enterprise 0.1.62 -> 0.1.63, litellm-proxy-extras 0.4.91 -> 0.4.92, litellm 1.100.0 -> 1.101.0 by @yuneng-berri in #39140
• revert: restore search tool fallback when no router is configured by @yuneng-berri in #39146
• test(websearch): register configured search tool in pre-request hook test by @mateo-berri in #39074
• feat(proxy): default to the v2 migration resolver, keep v1 as an opt-out by @yuneng-berri in #31125
• build(deps): bump browserslist to 4.28.8 to clear osv-scan by @devin-ai-integration[bot] in #39142
• fix(ui): render the skill detail page with theme tokens by @yuneng-berri in #39130
• feat: add Azure AI DeepSeek V4 Flash 0731 pricing by @yujonglee-berri in #39023
• fix(streaming): keep response id stable across streamed chunks by @Timik232 in #38106
• test(e2e/ui): cover the Budgets page create, edit and delete flows by @yuneng-berri in #39052
• feat(dashscope): add QwenCloud and Qwen AI Platform provider aliases by @mateo-berri in #39149
• fix(bedrock): forward native structured outputs on Invoke instead of silently inlining the schema by @tin-berri in #39070
• test(e2e/ui): cover creating, testing and deleting a guardrail by @yuneng-berri in #39053
• refactor(types): replace Any with precise types across 73 modules by @mateo-berri in #39104
• feat(models): add Claude Fable 5.1 across Anthropic, Bedrock, Vertex AI, and Azure AI by @devin-ai-integration[bot] in #39148
• feat(guardrails): add Alice guardrail by @seanyasno-af in #38898
• test(e2e/ui): cover the Logs page filter drawer by @yuneng-berri in #39056
• test(e2e/ui): stop the suite failing on things that are not regressions by @yuneng-berri in #39063
• test(e2e/ui): cover the team Settings tab by @yuneng-berri in #39058
• test(e2e/ui): cover the Usage page activity tabs by @yuneng-berri in #39061
• fix(ui): render the guardrail garden detail page with theme tokens by @yuneng-berri in #39131
• fix(responses): tool call id shape breaks gpt-5 -> claude fallback conversations by @mateo-berri in #39144
• fix(openai): drop tool_choice when request has no tools on chat completions by @mateo-berri in #39147
• chore(ci): promote internal staging to main by @yuneng-berri in #39141
• test(ui): pick select options by role instead of by text by @yuneng-berri in #39175
• feat(cost): support time-based off-peak pricing in cost calculation by @Srivatsa03 in #31725
• fix(openai): flatten top-level tool schema combinators on chat completions by @mateo-berri in #38839
• fix(s3): bound s3 object keys and download filenames for long Responses API ids by @yucheng-berri in #39164
• revert: default the proxy back to the v1 migration resolver by @yuneng-berri in #39178
• fix(prometheus): bound requested_model label cardinality on client failure paths by @mateo-berri in #39136
• feat(ui): add search to the Agent Hub tab and admin agents table by @ryan-crabbe-berri in #39155
• fix(anthropic): fix response_format for claude-fable-5-1 on Vertex AI and Bedrock by @devin-ai-integration[bot] in #39184
• fix: keep litellm_credential_name from LiteLLM Params JSON and gate stored credential attach to proxy admins by @yassin-berriai in #39047
• chore(ci): promote internal staging to main by @yuneng-berri in #39186
• test: exempt MockTransport request-shape embedding tests from VCR replay by @mateo-berri in #39185
• fix(ui): render the logs Tools panel with theme tokens by @yuneng-berri in #39129
• fix(proxy): default max_idle_connection_lifetime to 60s on DB URLs by @devin-ai-integration[bot] in #39134
• fix(mcp): follow tools/list pagination from upstream servers by @yucheng-berri in #39172
• fix(proxy): resolve router model aliases in /utils/supported_openai_params by @mateo-berri in #39000
• fix(azure): flatten top-level tool schema combinators on Azure chat completions by @mateo-berri in https://github.com/BerriAI/litellm/pull/38870
• fix(bedrock): route streamed responses-API output through the unified guardrail by @mateo-berri in https://github.com/BerriAI/litellm/pull/38734
• fix(ui): hide model write affordances from view-only admin sessions by @mateo-berri in https://github.com/BerriAI/litellm/pull/38872
• fix(cli): quote the Claude Code apiKeyHelper for cmd.exe on Windows by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39174
• fix(logging): guarantee max_parallel_requests slot release when streaming logging fails by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39093
• feat(alerting): slack alerts for per-user daily/monthly spend thresholds and spend anomaly detection by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/38438
• fix(docker): add public Wolfi apk repo to runtime image by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39033
• fix(router): keep order fallback on the requested order level by @emerzon in https://github.com/BerriAI/litellm/pull/38969
• test(e2e): cover retry-on-timeout and the context-window fallback by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39197
• fix(budget): reject known estimates over remaining budget under fail_closed_budget_enforcement by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39214
• fix: stop a cleared Team field from blocking personal key creation by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39206
• test: record each e2e test's source location in the JUnit report by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39209
• feat(router): fall back on anthropic safeguard refusals on /v1/messages by @tin-berri in https://github.com/BerriAI/litellm/pull/39157
• fix(proxy): report requested model on Anthropic streaming message_start by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/35816
• fix(helm): reuse the generated master key Secret on helm upgrade by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39219
• fix(mcp): report per-server outcomes in aggregate REST tools/list by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39232
• fix(cost-map): retry transient boot fetch failures and recover config deployments dropped by a stale cost map by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39230
• perf(scim): resolve group members with one user table read per member by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39228
• fix(docker): install bedrock-realtime extra in monolith proxy images by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39223
• fix(aiohttp_transport): map transport-internal CancelledError to a retryable ConnectError by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39240
• fix(bedrock): gate Converse cachePoint emission on model prompt caching support by @tin-berri in https://github.com/BerriAI/litellm/pull/39210
• fix(datadog_llm_obs): send tool calls, tool results and cache tokens in DD's own fields by @tin-berri in https://github.com/BerriAI/litellm/pull/39222
• feat(prometheus): expose per-key and per-team rate limit allowed and used gauges by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39236
• feat(scim): add placeholder listing and merge so a shadowed account can be healed by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39231
• fix: normalize provider-specific cache token fields in OTel v2 usage by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39202
• fix: stop deployment default API key limits leaking into provider requests by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39211
• fix(proxy): keep passthrough logging metadata and model_info dicts when team callbacks are wired by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39216
• fix(guardrails): deliver modify_response block as valid SSE on streaming chat and Responses by @mateo-berri in https://github.com/BerriAI/litellm/pull/39036
• fix(search): forward search-tool params through the router, complete Parallel AI v1 param mapping by @jliounis in https://github.com/BerriAI/litellm/pull/37883
• fix(bedrock): stop Converse crashing on bearer-token auth without SigV4 credentials by @mateo-berri in https://github.com/BerriAI/litellm/pull/39166
• fix(policy_engine): apply post_call pipeline text rewrites on streams by @mateo-berri in https://github.com/BerriAI/litellm/pull/39233
• fix(docker): install saml extra in litellm-backend image by @ojensen-berri in https://github.com/BerriAI/litellm/pull/39291
• fix(guardrails): run apply_guardrail-only providers in logging_only mode by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39297
• feat(gemini): day-0 pricing for gemini-3.8-flash by @mateo-berri in https://github.com/BerriAI/litellm/pull/39340
• fix(vertex): avoid duplicate DeepSeek OCR model namespace by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39194
• feat(streaming): carry final response cost on streamed usage by default by @mateo-berri in https://github.com/BerriAI/litellm/pull/39069
• fix(rerank): map provider errors with the resolved provider on sync and async paths by @mateo-berri in https://github.com/BerriAI/litellm/pull/39176
• test(e2e): read JUnit properties off the real collected pytest Item by @mateo-berri in https://github.com/BerriAI/litellm/pull/39246
• feat(proxy): configurable display_name for the Anthropic-shaped /v1/models listing by @mateo-berri in https://github.com/BerriAI/litellm/pull/39238
• fix(helm): scale the classic chart's HPA out at the documented 60 percent CPU by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35975
• fix(gemini): return enabled thinking content by default by @mateo-berri in https://github.com/BerriAI/litellm/pull/39160
• fix: run access group key sync UPDATEs on the writer, not the read replica by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39128
• fix(models): registry audit 2026-09-01: openai realtime and long-context tiers, mistral aliases, voyage, xai, fireworks, together, scaleway, azure ai, govcloud, azure gov, cloudflare whisper, deprecation dates by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39170
• fix: apply optional_pre_call_checks and reject unsupported router settings on /config/update by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39249
• fix(vector_stores): s3 vectors search router bypass + rag query config drop + ui error swallow by @michelligabriele in https://github.com/BerriAI/litellm/pull/34788
• fix(models): key Azure DeepSeek V4 Flash 0731 by its Foundry catalog id by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39341
• fix(deps): raise the tornado and pypdf floors for six new advisories by @mateo-berri in https://github.com/BerriAI/litellm/pull/39188
• fix(headroom): stop re-compressing retrieved CCR content in client tool loops by @QuantumBreakz in https://github.com/BerriAI/litellm/pull/38591
• feat(agentcore-a2a): derive runtime session id from A2A message.contextId by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39371
• fix(proxy): share per-model budget counters across replicas through the spend counter cache by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39375
• fix(proxy-extras): give prisma migrate deploy its own timeout budget by @mateo-berri in https://github.com/BerriAI/litellm/pull/39365
• fix(proxy): route container create and list through model_list deployments by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39220
• test(build): validate release wheel contracts by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39021
• refactor(rust): extract domain-neutral Python interop by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39026
• refactor(rust): standardize the core Error type by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39331
• fix(ui): preserve full AgentCore runtime ARN in agent edit form by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39382
• feat(ui): update OpenAI preset model tiers by @tin-berri in https://github.com/BerriAI/litellm/pull/39396
• fix(router): resolve realtime session model to routed deployment by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/36811
• fix(security): restrict and validate file uploads at /v1/files and /upload/logo by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39379
• feat(auth): enforce configurable password policy and SSO-only login by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39381
• fix(agents): redact secret litellm_params fields from all /v1/agents responses by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39389
• fix(otel): stamp Langfuse root observation input and output from the request task by @mateo-berri in https://github.com/BerriAI/litellm/pull/39369
• fix(guardrails): track and tear down presidio sibling callbacks on delete and update by @mateo-berri in https://github.com/BerriAI/litellm/pull/39271
• fix(spend): keep every-deployment scope on gateway cache-injection marks by @mateo-berri in https://github.com/BerriAI/litellm/pull/39241
• fix(proxy/db): keep prisma predicates from raising TypeError under a mocked prisma module by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39253
• fix(proxy): word database 503s by whether the fault is transient by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39256
• refactor(utils): remove the dead get_api_key provider-key resolver by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39260
• feat(mcp): semantic tool search for the native MCP Gateway by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39404
• fix(logging): redact credential query params from the uvicorn access log by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39293
• feat(model_prices): add meta/muse-spark-1.3 and its contributor tier by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39417
• refactor(core): move audio transcription into core by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39126
• fix(proxy): build coordination Redis from REDIS_* env vars unconditionally by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39410
• test: add interactive Rust Python parity harness in https://github.com/BerriAI/litellm/pull/39419
• test(proxy): verify NO_DOCS/NO_REDOC/NO_OPENAPI restrict every doc surface by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39378
• test(bedrock): accept the router kwarg in the knowledge base search fake by @mateo-berri in https://github.com/BerriAI/litellm/pull/39420
• refactor(python-bridge): split routes and add shared function tracing by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39031
• fix(python-bridge): harden sync and async execution boundaries by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39332
• refactor(python-bridge): declare sync and async routes once by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39333
• feat(python): unify Rust opt-in and bridge policy by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39334
• feat(router): add heuristic v2 complexity routing by @tin-berri in https://github.com/BerriAI/litellm/pull/39276
• fix(anthropic): upgrade legacy thinking to adaptive on adaptive-only Claude models for chat, Bedrock Converse, Invoke, Vertex AI, and Databricks by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39159
• fix(proxy): mark session/SSO/SAML cookies Secure behind a TLS-terminating reverse proxy by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39391
• fix(bedrock): honor BEDROCK_MANTLE_API_BASE on bedrock/mantle messages and chat URLs by @mateo-berri in https://github.com/BerriAI/litellm/pull/39364
• fix(bedrock): strip client_metadata from converse additionalModelRequestFields by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/35967
• chore(techdebt): clear fresh debt from the 2026-08-31 and 2026-09-01 windows by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39091
• fix(mcp): cap tools preview and test-connection at the listing timeout and name the unreachable upstream by @mateo-berri in https://github.com/BerriAI/litellm/pull/38791
• fix(hosted_vllm): forward truncate_prompt_tokens on rerank requests by @mateo-berri in https://github.com/BerriAI/litellm/pull/39363
• fix(messages): drop cache_control ttl on non-Anthropic /v1/messages passthrough by @mateo-berri in https://github.com/BerriAI/litellm/pull/39355
• fix(bedrock_mantle): carry per-request AWS credentials into chat completions SigV4 signing by @mateo-berri in https://github.com/BerriAI/litellm/pull/39362
• feat(router): add a hybrid classifier that defers near tier boundaries by @tin-berri in https://github.com/BerriAI/litellm/pull/39403
• fix: recover the v2 migration resolver from concurrent migrate deploy deadlocks by @mateo-berri in https://github.com/BerriAI/litellm/pull/39187
• fix(ollama_chat): stamp finish_reason tool_calls when tool calls streamed before the done chunk by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39010
• fix(router): route Claude Code subagents through session router by @moe-berri in https://github.com/BerriAI/litellm/pull/39239
• fix(responses): keep namespace tools intact when a guardrail returns them unchanged by @mateo-berri in https://github.com/BerriAI/litellm/pull/39366
• fix(vector-store): resolve embedding credentials per request by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/38936
• test(e2e/ui): give the seeded users passwords that pass the default password policy by @mateo-berri in https://github.com/BerriAI/litellm/pull/39442
• fix(http_handler): honor HTTP(S)_PROXY / NO_PROXY when force_ipv4 uses the httpx transport by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39443
• fix(proxy): stop leaking internal exception details to clients by @yassin-berriai in https://github.com/BerriAI/litellm/pull/39380
• fix(guardrails): forward mode and streaming params to crowdstrike_aidr handler by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39317
• fix(mcp): gate the connect-time OBO pre-flight on the key's allowed servers by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39447
• fix(responses): keep provider response headers in streaming logging callbacks by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/38131
• fix(mcp): fence an outbound-token write against an overlapping invalidation by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35398
• feat(cli): pre-fill the SSO verification code in the browser when the proxy allows it by @mateo-berri in https://github.com/BerriAI/litellm/pull/39428
• fix(ui): paginate request logs by session groups server-side by @tin-berri in https://github.com/BerriAI/litellm/pull/39257
• feat(proxy): serve the auto-router preset catalog at runtime by @tin-berri in https://github.com/BerriAI/litellm/pull/39412
• docs: define Rust Python harness structure by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39456
• fix(guardrails): apply PUT /guardrails/{id} to the serving worker immediately and reject invalid configs with 422 by @mateo-berri in https://github.com/BerriAI/litellm/pull/38877
• test(responses): expect the 404 OpenAI now returns for an unknown model by @mateo-berri in https://github.com/BerriAI/litellm/pull/39457
• fix(guardrails): skip streaming guardrail rounds that re-scan cleared output by @mateo-berri in https://github.com/BerriAI/litellm/pull/39386
• fix: keep litellm importable on Python 3.10 and guard 3.11-only typing imports in CI by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39448
• fix(proxy): keep SpendLogs and callback session ids in sync when the request has none by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39450
• feat(router): arm safeguard-refusal fallback on generic chains when no content-policy list exists by @tin-berri in https://github.com/BerriAI/litellm/pull/39274
• feat(azure): support credential chain for storage by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39229
• chore(crowdstrike): expect the deduped end-of-stream scan in crowdstrike cadence test by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39467
• fix(model_armor): handle Anthropic Messages and Responses streams in post_call by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39181
• test: add OCR python-to-rust test parity ledger (WIP) in https://github.com/BerriAI/litellm/pull/39434
• feat(complexity_router): opt-in modality override of a kept session-affinity pin by @tin-berri in https://github.com/BerriAI/litellm/pull/39454
• feat(datadog_llm_obs): cost tag dimensions, router decision fields, reasoning token metric, redaction gating by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39402
• test(rust-python-harness): wire existing e2e SDK tests into the matrix in https://github.com/BerriAI/litellm/pull/39463
• fix(mcp): never exchange the LiteLLM virtual key as the upstream subject token by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39446
• test: add mistral ocr transformation parity coverage in https://github.com/BerriAI/litellm/pull/39482
• test(vector-store): accept embedding_executor in the Bedrock KB hook fake handler by @mateo-berri in https://github.com/BerriAI/litellm/pull/39472
• refactor(s3_vectors): embed search queries through the shared vector store executor by @mateo-berri in https://github.com/BerriAI/litellm/pull/39474
• fix(xai): bill from the cost xAI reports instead of recomputing it (internal copy of #36281) by @mateo-berri in https://github.com/BerriAI/litellm/pull/39441
• feat(ui): add 1M context auto-router preset by @tin-berri in https://github.com/BerriAI/litellm/pull/39490
• fix(ui): stop the create team form resetting organization and models by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39476
• fix(ui): read the preset catalog at runtime in the dashboard tests by @tin-berri in https://github.com/BerriAI/litellm/pull/39478
• fix(sso): resolve multi-valued role claims to the highest privilege role by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39480
• fix(guardrail): hide-secrets playground redaction and guardrail telemetry by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39398
• fix(test): drop the duplicate embedding_executor arg in the Bedrock KB fake handler by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39502
• fix(ui): keep Virtual Keys list state in the URL so it survives leaving the page by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39481
• fix(proxy): 404 a credential delete that matched nothing, and raise instead of return by @eeshsaxena in https://github.com/BerriAI/litellm/pull/36260
• fix(proxy-extras): only spend a migrate-deploy attempt when a pass made no progress by @mateo-berri in https://github.com/BerriAI/litellm/pull/39506
• feat(cli): enable Claude Code gateway model discovery by default in lite claude by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39445
• fix(docker): bump nginx runtime to 1.31.5-alpine3.24 and pin digest by @rakeshrepository in https://github.com/BerriAI/litellm/pull/39561
• fix: 1.99.0-rc2 UI bug batch (empty org on key create, session pagination, access group rename/delete) by @mateo-berri in https://github.com/BerriAI/litellm/pull/39436
• feat(auto-router): support classifier reasoning effort by @moe-berri in https://github.com/BerriAI/litellm/pull/39372
• fix(ui): replace the key detail URL entry when a virtual key is rotated by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39471
• test(timeout): time out against the local fake endpoint instead of api.openai.com by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39583
• test(harness): add OCR parity with migration strategy runners by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/38765
• fix(databricks): strip thinking_blocks and reasoning_content from outbound messages by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39409
• test(ocr): record provider fixtures in the migration harness by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39425
• feat(ui): keyset-paginate request logs by session trace by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/38794
• fix(proxy/db): translate libpq sslrootcert and verify-* into Prisma's strict TLS params by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39563
• fix(agents): keep the published agent in public_agent_groups by @mateo-berri in https://github.com/BerriAI/litellm/pull/39554
• fix(mcp): scope allow-all servers to virtual keys by @tin-berri in https://github.com/BerriAI/litellm/pull/39531
• fix(team): generate team IDs for blank input by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39571
• fix(bedrock_mantle): stop dropping the web_search tool on /v1/responses by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/35987
• chore: bump litellm-enterprise 0.1.63 -> 0.1.64, litellm-proxy-extras 0.4.92 -> 0.4.93 by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39595
• fix(images): forward gpt-image supported params like background to OpenAI and Azure by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39525
• fix(proxy): return persisted team memberships from /user/new so first CLI login gets the default team by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39545
• fix(spend_tracking): add missing_session_id: omit to leave SpendLogs.session_id null without a client session by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39458
• fix: stop a cleared Organization field from failing key creation by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39316
• fix(ui): show MCP servers and agents inherited from access groups on team overview by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39215
• fix(proxy): expose configured mode for auto-router models by @moe-berri in https://github.com/BerriAI/litellm/pull/39619
• fix(ui): aggregate session token usage in the logs table by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39598
• fix(cost): apply off_peak_pricing in the dashscope cost calculator by @mateo-berri in https://github.com/BerriAI/litellm/pull/39592
• test(bedrock): drop EOL cohere.command-r-plus-v1:0 from local_testing by @mateo-berri in https://github.com/BerriAI/litellm/pull/39608
• fix(openai): default stream usage on PrivateLink and regional api.openai.com hosts by @mateo-berri in https://github.com/BerriAI/litellm/pull/39614
• fix(proxy): drop anthropic-beta on the Vertex passthrough count-tokens route by @mateo-berri in https://github.com/BerriAI/litellm/pull/39597
• fix(headroom): resolve CCR retrieval on streaming /v1/responses by @mateo-berri in https://github.com/BerriAI/litellm/pull/38808
• fix(openai): bridge gpt-5.4+ tool calls to /v1/responses on every api.openai.com host by @mateo-berri in https://github.com/BerriAI/litellm/pull/39587
• fix(router): pin JWT-authenticated callers by user id in deployment_affinity by @mateo-berri in https://github.com/BerriAI/litellm/pull/39594
• fix(cost): bill bedrock_mantle web search at $12 per 1k queries using Bedrock's reported count by @mateo-berri in https://github.com/BerriAI/litellm/pull/39610
• fix(azure_ai): don't reclassify Foundry deployments as azure provider by @mateo-berri in https://github.com/BerriAI/litellm/pull/38975
• fix(vector_stores): only list vector stores the caller was granted by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39612
• feat(models): add gpt-6-astra pricing and metadata by @mateo-berri in https://github.com/BerriAI/litellm/pull/39622
• chore(ci): promote internal staging to main by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39593
• feat(router): limit heuristic_v2 auto-routers to one without the auto_router license feature by @tin-berri in https://github.com/BerriAI/litellm/pull/39468
• fix(ui): clear agents when updating team permissions by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39600
• fix(auto_router): bill the routing embedding to the caller's key and team by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39532
• test(router): cover get_configured_mode so router_code_coverage passes by @mubashir1osmani in https://github.com/BerriAI/litellm/pull/39630
• fix: treat gpt-6 names as the gpt-5 request family in OpenAI and Azure configs by @mateo-berri in https://github.com/BerriAI/litellm/pull/39631
• fix(prompts): key the in-memory prompt registry by environment by @mateo-berri in https://github.com/BerriAI/litellm/pull/38440
• fix(ui): let the Internal Users search box match user_id as well as email by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39604
• test(responses): bound the background stream cancel e2e so an upstream stall skips fast by @mateo-berri in https://github.com/BerriAI/litellm/pull/39617
• fix(vertex): add the API version to versionless project routes on the Vertex passthrough by @mateo-berri in https://github.com/BerriAI/litellm/pull/39625
• chore(ci): promote internal staging to main by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39648
• fix(spend_tracking): key /v1/messages spend rows on the msg_ id the client received by @mateo-berri in https://github.com/BerriAI/litellm/pull/39511
• ci(rust): build and test the ai-gateway server feature by @mateo-berri in https://github.com/BerriAI/litellm/pull/39493
• ci(ui): run the UI build check through the image's ui-builder stage by @mateo-berri in https://github.com/BerriAI/litellm/pull/39496
• fix(proxy): parse numeric multipart fields on /v1/images/edits back into numbers by @mateo-berri in https://github.com/BerriAI/litellm/pull/39510
• fix(guardrails): remove the module-global translation mapping that leaked between tests by @mateo-berri in https://github.com/BerriAI/litellm/pull/39543
• feat(azure_ai): add grok-4.6 to the model cost map by @mateo-berri in https://github.com/BerriAI/litellm/pull/39426
• fix: attach vector store search_results when a guardrail is registered by @mateo-berri in https://github.com/BerriAI/litellm/pull/38984
• fix(proxy): stop putting the literal string "None" in error payloads by @mateo-berri in https://github.com/BerriAI/litellm/pull/39521
• fix(router): keep retry breadcrumbs per request and out of the request snapshot by @mateo-berri in https://github.com/BerriAI/litellm/pull/39491
• fix(vector-stores): survive a failing vector store search in the chat completions hook by @mateo-berri in https://github.com/BerriAI/litellm/pull/39495
• fix(utils): redact credential kwargs from the set_verbose request line by @mateo-berri in https://github.com/BerriAI/litellm/pull/39526
• fix(bedrock): skip the SigV4 credential chain when a bearer token is configured by @mateo-berri in https://github.com/BerriAI/litellm/pull/39411
• fix(proxy-extras): kill the whole Prisma process group when a command times out by @mateo-berri in https://github.com/BerriAI/litellm/pull/39466
• fix(rag): forward the managed vector store's params to the search call by @mateo-berri in https://github.com/BerriAI/litellm/pull/39452
• fix(utils): redact credentials nested in extra_body on the verbose optional-params line by @mateo-berri in https://github.com/BerriAI/litellm/pull/39538
• fix(containers): pass upstream error status through and forward list pagination params by @mateo-berri in https://github.com/BerriAI/litellm/pull/39464
• fix(anthropic_messages): key bridged streaming spend rows on the streamed msg_ id by @mateo-berri in https://github.com/BerriAI/litellm/pull/39541
• fix(helm): render ingress-nginx compatible path types via ingress.controller by @mateo-berri in https://github.com/BerriAI/litellm/pull/39465
• fix(cache): use sync Redis batch reads by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39358
• fix(guardrails): rebuild the serving worker guardrail on PUT instead of patching it in place by @mateo-berri in https://github.com/BerriAI/litellm/pull/39243
• test(team-race): wait on pg_locks instead of a fixed sleep by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39611
• fix(agents): hide agents from non-admins who were never granted them by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39636
• fix(team_endpoints): stop partial /team/update from wiping team metadata by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/36328
• test(router): cover configured mode lookup by @moe-berri in https://github.com/BerriAI/litellm/pull/39634
• fix(responses): encrypt the response id on every streamed event by @mateo-berri in https://github.com/BerriAI/litellm/pull/39534
• fix(mcp): resolve OAuth broker endpoints by server_id with IP access checks by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39432
• feat(proxy): enforce team isolation for provider-format batch ids and output files by @yuneng-berri in https://github.com/BerriAI/litellm/pull/33536
• fix(mcp): strip inbound auth scheme case-insensitively before token exchange by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39346
• fix(mcp): normalize a schemed authentication_token on the v2 and OpenAPI static paths by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39345
• fix(docker): match USE_DDTRACE case-insensitively and route build_from_pip through prod_entrypoint.sh by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39344
• perf(auth): skip object permission DB lookup when no vector stores requested by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39347
• fix(caching): don't trip redis circuit breaker on short timeout bursts by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/38999
• fix(access_groups): derive attached teams from the team table and reject unknown team ids by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39218
• test(router): drop duplicate get_configured_mode test failing ruff F811 by @mateo-berri in https://github.com/BerriAI/litellm/pull/39659
• feat(cost): honor off_peak_pricing reasoning and cache-creation rates by @mateo-berri in https://github.com/BerriAI/litellm/pull/39635
• fix(openai): mint workload identity tokens for PrivateLink and regional api.openai.com hosts by @mateo-berri in https://github.com/BerriAI/litellm/pull/39652
• feat(ui): find rows by a pasted ID on keys, agents, memory, audit, and request logs by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39661
• test(proxy-extras): fake run_prisma instead of subprocess.run in the migrate deploy harness by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39669
• fix(scim): default-team fallback on create and keep memberships when PUT /Users has no groups by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39623
• fix(router): count tools and Anthropic system prompt in context-window pre-call check by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39663
• test(proxy-extras): repoint the migrate-deploy harness at the run_prisma seam by @mubashir1osmani in https://github.com/BerriAI/litellm/pull/39673
• perf(spend): group /spend/logs summary by day in Postgres instead of per-row Prisma group_by by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39351
• fix(proxy): apply default_vertex_config location before building the Vertex passthrough base URL by @mateo-berri in https://github.com/BerriAI/litellm/pull/39662
• fix(anthropic_endpoints): return Anthropic type:error envelope for /v1/messages errors by @mateo-berri in https://github.com/BerriAI/litellm/pull/39037
• feat(ui): configure auto-router session affinity TTL by @tin-berri in https://github.com/BerriAI/litellm/pull/39679
• perf(mcp): cache SSO identity assertion reads on the ID-JAG path by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39348
• test(e2e): match the Internal Users search placeholder shipped by #39604 by @mateo-berri in https://github.com/BerriAI/litellm/pull/39678
• fix(ui): scroll admin table rows inside the table instead of the page by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39684
• fix(router): evict stale global pattern_router entries on upsert/delete by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39664
• fix(caching): keep a node timeout from forcing a cluster-wide topology reinit on redis-py 8.x by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39349
• fix(logging): blocked requests no longer report guardrail_status=success in multi-guardrail configs by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39596
• fix(snowflake): normalize Cortex Claude request shapes by @tin-berri in https://github.com/BerriAI/litellm/pull/39453
• feat(proxy): per-worker admission control that rejects excess requests with 503 by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39352
• fix(proxy): emit SSE keepalives on queue, rag, azure passthrough, usage chat and policy enrich streams by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39273
• fix(azure): restrict the storage credential chain to deployment identities by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39637
• fix(model_checks): drop wildcard routes like bedrock/* from /v1/models by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/31731
• fix(mcp): pre-flight the ID-JAG credential at the transport edge by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35392
• feat(caching): add semantic_cache_scope to isolate semantic cache hits per end user by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39590
• feat(ui): add one-click Auto Router setup by @moe-berri in https://github.com/BerriAI/litellm/pull/39693
• fix(ci): pin setup-uv to v10.0.1 by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39111
• refactor(tests): restructure rust python harness around strategy definitions by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39628
• test(ocr): complete Rust unit test parity by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39689
• feat: page the public model hub table off /public/v1/model_hub, keeping every filter by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39691
• refactor(rust): extract config crate by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39706
• feat(python): rename Rust rollout API by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39704
• fix: restore Python compatibility and test 3.10 through 3.14 by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/39399
• fix(spend-tracking): keep batch spend keys joinable after v1.99 provenance gate by @mateo-berri in https://github.com/BerriAI/litellm/pull/39568
• test: repair four chronically failing CI tests by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39770
• feat(team): report per-user spend within a team for JWT traffic by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39771
• feat(router): add classifier circuit breaker by @moe-berri in https://github.com/BerriAI/litellm/pull/39701
• ci: report every failing test in a job instead of stopping at the first by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39772
• test(caching): drive the redis stall burst off the clock, not asyncio.wait_for by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39773
• fix(ui): clamp server-paginated DataTable page index when rowCount shrinks by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39776
• feat(mcp): use x-mcp-<access_group>-* headers as default upstream credentials for group members by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39717
• fix(ui): paginate per-user usage with the shared server-side DataTable footer by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39682
• fix(proxy): log mid-stream /v1/messages failures as failures with partial usage by @mateo-berri in https://github.com/BerriAI/litellm/pull/39589
• fix(cost): honor off_peak_pricing in the fireworks_ai and perplexity cost calculators by @mateo-berri in https://github.com/BerriAI/litellm/pull/39632
• fix(organization): clear org budget limits when PATCH /organization/update sends null by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39670
• fix(organization): reject negative limits and unparseable budget_duration on PATCH /v2/organization by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39793
• feat(organization): expose PATCH /v2/organization/{organization_id} in the OpenAPI schema by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39794
• fix(auto-router): route 1M complex tier to GPT Sol by @tin-berri in https://github.com/BerriAI/litellm/pull/39797
• fix(bedrock_mantle): anchor MANTLE_HOST_RE so custom Mantle hosts are honored by @mateo-berri in https://github.com/BerriAI/litellm/pull/39361
• fix(health): probe test_connection with the credential the request names by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39801
• fix(router): bound auto-router classifier latency by @moe-berri in https://github.com/BerriAI/litellm/pull/39696
• fix(ui): accept any routing group name the backend accepts by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39807
• fix(model_prices): verified registry audit, Databricks Sep-2026 catalog, realtime image pricing, deprecation dates by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39388
• fix(jwt): invalidate JWT key mapping cache on /key/regenerate by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39808
• fix(complexity_router): fall back to a live peer when the decided tier model is fully cooled down by @tin-berri in https://github.com/BerriAI/litellm/pull/39675
• feat(guardrails): roll up Bedrock guardrail cost per usage counter by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39196
• fix(auto_router): derive tier definitions in prompt editor by @tin-berri in https://github.com/BerriAI/litellm/pull/39688
• fix(proxy): recognize opencode's bare x-session-id header for session affinity by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39802
• feat(cli): sync OpenCode models from /v1/models in lite opencode by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39789
• fix(spend-tracking): keep internal service-account key names readable in spend logs by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39572
• fix(anthropic): never carry cache_control on translated thinking blocks by @moe-berri in https://github.com/BerriAI/litellm/pull/39815
• fix(responses/mcp): keep follow-up calls stateless when store=false by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/36575
• fix(router): resolve retry_policy by exception hierarchy, add ServiceUnavailableErrorRetries and DefaultRetries by @shivamrawat1 in https://github.com/BerriAI/litellm/pull/35853
• fix(headroom): bound the /v1/compress and /v1/retrieve calls with a timeout by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39527
• feat(vector_stores): add a MongoDB vector store provider for Atlas and self-managed deployments by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39811
• fix(fireworks_ai): resolve tool_choice and reasoning support for short model names by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39763
• fix(datadog_llm_obs): keep the guardrail audit record under message redaction by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39702
• fix(proxy): invalidate end-user spend counter and cache on budget reset (#39726) by @amasen02 in https://github.com/BerriAI/litellm/pull/39729
• fix(proxy): strip every TypedDict qualifier before numeric form-field detection by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39780
• fix(bedrock): stop sending toolConfig tool definitions to guardrails on passthrough converse by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39281
• feat(shadow_eval): judge tool-call turns instead of dropping or erroring on them by @moe-berri in https://github.com/BerriAI/litellm/pull/39818
• feat(router): auto-escalate stalled complexity-router tasks by @moe-berri in https://github.com/BerriAI/litellm/pull/39809
• fix(team_endpoints): let member_delete clear a team left on the user row by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/38703
• feat(cost-map): add azure/gpt-6-astra and azure/us/gpt-6-astra Foundry pricing by @mateo-berri in https://github.com/BerriAI/litellm/pull/39827
• feat(complexity_router): let the LLM classifier see request images by @tin-berri in https://github.com/BerriAI/litellm/pull/39825
• feat(access-groups): resolve resource names on access group responses by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39822
• fix(datadog_llm_obs): keep guardrail_cost_by_unit on redacted spans by @mateo-berri in https://github.com/BerriAI/litellm/pull/39848
• test(store_model_in_db): assert the 400 contract in the unknown-model spend log test by @mateo-berri in https://github.com/BerriAI/litellm/pull/39842
• feat(cli): add lite debug claude session report and /debug-lite slash command by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39435
• test: deflake JWT tamper, fuzzy picker, tag routing, liveliness, redis stall burst, and pre-commit interrupt tests by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39306
• feat(otel): stamp litellm.request.route on the LLM call span by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39698
• feat(shadow_eval): scope a job to model groups, ANDed with its key, team, and user targets by @tin-berri in https://github.com/BerriAI/litellm/pull/39828
• test(e2e/batches): assert Bedrock batch cancel and list in the lifecycle by @mateo-berri in https://github.com/BerriAI/litellm/pull/39847
• refactor: clear fresh tech debt from the last 24 hours (2026-09-03, 2026-09-04) by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39518
• refactor(typing): cut 1,397 Any errors across 183 backend files by @mateo-berri in https://github.com/BerriAI/litellm/pull/39461
• fix(responses): decode JSON-string tool schemas before sending to the provider by @tin-berri in https://github.com/BerriAI/litellm/pull/39844
• feat(helm): render nodeSelector, tolerations, and affinity on the componentized chart migrations Job by @mateo-berri in https://github.com/BerriAI/litellm/pull/39843
• fix(mcp): let config.yaml MCP servers pin server_id by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39286
• feat(dashboard): configure classifier vision input by @tin-berri in https://github.com/BerriAI/litellm/pull/39840
• fix(batches): register ownership for every batch create path by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39810
• fix(proxy): gate the OpenAI websocket passthrough behind an explicit opt-in by @mateo-berri in https://github.com/BerriAI/litellm/pull/39841
• fix(ui): bring the inline-object lint budget back under its ceiling by @mateo-berri in https://github.com/BerriAI/litellm/pull/39856
• fix(realtime): relay the upstream websocket close to the client instead of hanging by @mateo-berri in https://github.com/BerriAI/litellm/pull/39851
• feat(router): meter auto-router tier and prompt customization against the auto_router license feature by @tin-berri in https://github.com/BerriAI/litellm/pull/39674
• fix(shadow_eval): size the judge output cap for a judge that reasons by @moe-berri in https://github.com/BerriAI/litellm/pull/39817
• feat(organization): expose PATCH /v2/organization/{organization_id} in the OpenAPI spec by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39672
• fix(ui): make Admin UI table pagination honor the selected page size by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39680
• chore: bump litellm-enterprise 0.1.64 -> 0.1.65, litellm-proxy-extras 0.4.93 -> 0.4.94 by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39912
• test(e2e): repair the wildcard readiness probe and the semantic auto-router spend assertion by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39804
• fix(guardrails): record guardrail information for undecorated custom apply_guardrail overrides by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39727
• feat(responses): honor supported_endpoints /v1/responses opt-in for OpenAI-compatible deployments by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39725
• fix(router): hold max_parallel_requests slot until streaming response is exhausted or closed by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39859
• fix(proxy): retry deadlocks and requeue spend logs on any DB write error by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39883
• fix(mcp): reject URL credentials for none auth by @tin-berri in https://github.com/BerriAI/litellm/pull/39926
• fix(hide-secrets): stop redacting benign identifiers by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39879
• perf(logging): scan large base64 payloads for log truncation off the event loop by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39890
• fix(proxy): make the invalid-model 403 path cheap under a burst of rejections by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39892
• test(e2e): cover Anthropic /chat/completions streaming and tool calls by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39916
• fix(router): coordinate async and sync failure handlers at remaining router call sites by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39887
• fix(cloudzero): infer daily batch schema from every row by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39871
• fix(cloudzero): preserve late resource tags by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39873
• feat(ui): deep link guardrail detail with ?guardrail= on guardrails pages by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39930
• test: repair two CI tests broken by intentional changes by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39932
• feat(terraform/gcp): dependencies-only mode and bring-your-own-network for GKE by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39695
• test(e2e): cover Anthropic and OpenAI prompt caching, Cohere embeddings, and costed /openai chat passthrough by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39920
• test(e2e): cover key spend reset, regenerate grace period, and the llm_api_routes grant by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39917
• test(e2e/ui): select 50 rows per page before asserting the Tags and Model Hub tables overflow by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39934
• feat(auto-router): decouple compression between the routing decision and the model call by @moe-berri in https://github.com/BerriAI/litellm/pull/39823
• feat(mcp): renew the stored SSO identity assertion behind ID-JAG by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35401
• feat(mcp): warn when an oauth2_id_jag server outruns the SSO provider's assertion capture by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35394
• perf: lazy-load SDK symbols so import litellm stays under 60 MB RSS by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39121
• test(e2e): cover presidio post_call, tool_permission, and weave logging cells by @yucheng-berri in https://github.com/BerriAI/litellm/pull/39279
• fix(ci): grant pull_requests write for release wheel reporter by @cursor[bot] in https://github.com/BerriAI/litellm/pull/39922
• feat(guardrails): add non-blocking flag() verdict to custom code guardrails by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39728
• feat(proxy): serve Prometheus /metrics from a separate process via --prometheus_metrics_port by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39889
• fix(mcp): scan and mask MCP tool call arguments in unified guardrails by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35142
• fix(proxy): reject ambiguous name or alias keys in mcp_tool_permissions on write by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39947
• test(e2e): stream a longer /v1/messages reply so the delta-count pin has margin by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39946
• feat(ui): show guardrail usage units and cost on the Guardrails Monitor by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39853
• fix(ui): show indirectly granted and name-keyed MCP servers in the tool matrix by @yassin-berriai in https://github.com/BerriAI/litellm/pull/35154
• test(e2e): prove Vertex context caching on the first cold call and on the spend row by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39938
• feat(ocr): add Cohere Parse support for cohere and azure_ai by @mateo-berri in https://github.com/BerriAI/litellm/pull/39862
• fix(proxy): keep guardrail cost in spend on cache hits by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39960
• test(e2e): judge /v1/messages streaming on the clock, not on the provider's delta count by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39953
• fix(ui): remove unreachable AI Hub dialog that put the session key in a URL by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/39968
• fix(ui): read Usage Total Requests tile from gateway request counts by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/39963
• revert: perf: lazy-load SDK symbols so import litellm stays under 60 MB RSS (#39121) by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39969
• chore: rebuild Admin UI bundle for the next release by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39959
• chore(ci): promote internal staging to main by @yuneng-berri in https://github.com/BerriAI/litellm/pull/39849
• fix(proxy): redact provider keys from pass-through failure tracebacks by @mateo-berri in https://github.com/BerriAI/litellm/pull/39964
• fix(file_search): scope emulated file_search to the request's vector stores by @mateo-berri in https://github.com/BerriAI/litellm/pull/39972
• feat(cli): add lite pi to run the pi coding agent through the proxy by @ryan-crabbe-berri in https://github.com/BerriAI/litellm/pull/36841
• feat(pricing): add GovCloud pricing for every live but unpriced Bedrock model by @mateo-berri in https://github.com/BerriAI/litellm/pull/39764

Full Changelog: https://github.com/BerriAI/litellm/releases/tag/v1.108.0