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
Note truncated.