Issues

• [Docs/Bug]: health check routing does not exclude a dead deployment until the first background health check finishes, so the first requests after startup pay the full timeout

  llm translation
  BerriAI/litellm#50084 · Filter by author Atsu1979 opened on Nov 4, 2026

• [Bug]: /v1/messages to Vertex Claude sends messages before system/tools, and the prompt cache misses on tool-loop rounds

  bugllm translation
  BerriAI/litellm#50070 · Filter by author brandonnovick opened on Nov 4, 2026
  2

• [Bug]: v1.108.0 MCP tools/list blocks 18-50 s on legacy subscription listener after notifications/initialized

  BerriAI/litellm#50065 · Filter by author gruuna-bannert opened on Nov 4, 2026

• [Bug]: nvidia_riva sends no authorization header on plaintext gRPC and duplicates it over TLS (riva-client version dependent)

  bug
  BerriAI/litellm#50059 · Filter by author AlexisGitHu opened on Nov 4, 2026

• Can the provider logo in Model Management be customized for models routed through an OpenAI/Anthropic-compatible gateway?

  llm translation
  BerriAI/litellm#50056 · Filter by author jweisman11 opened on Nov 4, 2026

• [Bug]: /v1/messages guardrail tool write-back drops web_search server tools and turns mcp_toolset into litellm_unnamed_tool_0

  llm translation
  BerriAI/litellm#50055 · Filter by author afonsofigs opened on Nov 4, 2026

• [Feature]: Queue requests at capacity by priority, with fair turns between tenants (priority has no effect under load today)

  llm translation
  BerriAI/litellm#50054 · Filter by author parjanyamodi opened on Nov 4, 2026

• [Feature]: Opt-in upstream error details for OpenAPI-backed MCP tools

  BerriAI/litellm#50033 · Filter by author stg609 opened on Nov 4, 2026
  1

• [Bug]: False BudgetExceededError on /vertex_ai passthrough due to duplicate user_api_key_auth call leaking budget reservations and 1M-token fallback for Gemini "contents"

  bugllm translation
  BerriAI/litellm#50030 · Filter by author ahkui opened on Nov 4, 2026
  1

• [Bug]: Sticky Session not working - Router always sends Request to a different Endpoint

  bugllm translation
  BerriAI/litellm#50029 · Filter by author luckylinux opened on Nov 4, 2026
  2

• [Bug]: Azure Messages-to-Responses detour drops encrypted reasoning; follow-up to #40451 / LIT-7419

  llm translation
  BerriAI/litellm#50024 · Filter by author roni-frantchi opened on Nov 4, 2026
  2

• [Bug]: Messages-to-Responses bridge defaults JSON schema strictness to false while Messages-to-Chat uses true

  llm translation
  BerriAI/litellm#50023 · Filter by author roni-frantchi opened on Nov 4, 2026
  1

• [Bug]: /v1/messages answers 503 "authentication database is temporarily unreachable" when the upstream call fails

  llm translation
  BerriAI/litellm#49998 · Filter by author fancyboi999-bot opened on Nov 4, 2026
  2

• [Bug]: Bedrock Converse replays reasoningText.signature to non-Anthropic models (GLM 5.3) causing Bad Request

  llm translation
  BerriAI/litellm#49969 · Filter by author JetFault opened on Nov 3, 2026
  2

• [Bug]: Respawned worker reuses the CLI-minted DB token with NUM_WORKERS > 1 (AZURE_POSTGRESQL_AUTH / IAM_TOKEN_DB_AUTH) → P1010 after expiry

  llm translation
  BerriAI/litellm#49963 · Filter by author mihidumh opened on Nov 3, 2026

• [Bug]: A stream that fails after it opens logs the raw provider exception class (e.g. AnthropicError) in error_information

  llm translation
  BerriAI/litellm#49940 · Filter by author alpaylan opened on Nov 3, 2026

• [Bug]: Sync Router.responses() ignores the Router's num_retries and fallbacks, Router.aresponses() honours them

  llm translation
  BerriAI/litellm#49938 · Filter by author alpaylan opened on Nov 3, 2026

• [Bug]: OpenAI HTTP 403 raises APIError or BadRequestError instead of PermissionDeniedError

  llm translation
  BerriAI/litellm#49936 · Filter by author alpaylan opened on Nov 3, 2026
  1

• [Bug]: anthropic.messages.acreate(stream=True) on Bedrock raises the raw BedrockError for an exception event in the stream

  llm translation
  BerriAI/litellm#49935 · Filter by author alpaylan opened on Nov 3, 2026

• [Bug]: Router stream=True with a Gemini deployment raises the raw VertexAIError (async) or a TypeError (sync) when the stream fails to open

  llm translation
  BerriAI/litellm#49933 · Filter by author alpaylan opened on Nov 3, 2026

• [Bug]: /engines/{model}/chat/completions and /openai/deployments/{model}/chat/completions return the provider model name without streaming

  llm translation
  BerriAI/litellm#49932 · Filter by author alpaylan opened on Nov 3, 2026
  1

• [Bug]: anthropic.messages.acreate(stream=True) caches an empty reply from a non-Anthropic model and replays it on every repeat

  llm translation
  BerriAI/litellm#49931 · Filter by author alpaylan opened on Nov 3, 2026
  1

• [Bug]: Health probes cause context to be evicted

  bug
  BerriAI/litellm#49930 · Filter by author mhewitt-dev opened on Nov 3, 2026
  2

• [Bug]: A cache hit returns finish_reason "tool_calls" for a stored Anthropic response that had "stop"

  llm translation
  BerriAI/litellm#49929 · Filter by author alpaylan opened on Nov 3, 2026

• [Bug]: max_messages is counted on the translated chat messages for responses() and anthropic.messages on bridged providers

  llm translation
  BerriAI/litellm#49928 · Filter by author alpaylan opened on Nov 3, 2026