Issues

• [Bug]: minimax provider silently drops top-level reasoning_effort (M3.1-Flash supports low|medium|high|max)

  llm translation
  BerriAI/litellm#49574 · Filter by author pkuqiuning opened on Nov 1, 2026

• [Feature]: Bump a2a-sdk from 1.1.0 to 1.3.0 so the A2A gateway accepts agents that send newer fields

  BerriAI/litellm#49570 · Filter by author gulldan opened on Nov 1, 2026

• [Bug]: /a2a/{agent_id} answers an unparsable body outside JSON-RPC: -32603 / HTTP 500 for an empty body, an OpenAI-style 400 for malformed JSON

  llm translation
  BerriAI/litellm#49569 · Filter by author gulldan opened on Nov 1, 2026

• [Bug]: Gemini image requests return 200 OK but write no SpendLogs row (silent skip; state-dependent)

  bugllm translation
  BerriAI/litellm#49567 · Filter by author RippedPuma opened on Nov 1, 2026
  1

• [Feature]: Add Webz.io News Search as a /search provider

  BerriAI/litellm#49565 · Filter by author ShakedDegani opened on Nov 1, 2026

• [Bug]: Response cache ignores api_base — backend A's cached response served to backend B

  bug
  BerriAI/litellm#49551 · Filter by author shadowmodder opened on Nov 1, 2026

• MCP Registry entry io.github.BerriAI/litellm-mcp points to a repository that does not exist

  BerriAI/litellm#49489 · Filter by author xhorizont opened on Oct 31, 2026

• pass_through_endpoints strips id from JSON-RPC request bodies (breaks JSON-RPC 2.0, name collision with internal param list)

  BerriAI/litellm#49479 · Filter by author sunilvalluru opened on Oct 31, 2026
  3

• [Feature]: Bridge A2A contextId to Vertex AI Agent Engine sessions (like Bedrock AgentCore / LangFlow)

  llm translation
  BerriAI/litellm#49451 · Filter by author GordonWei opened on Oct 31, 2026
  1

• [Bug]: Anthropic streaming guardrails parse transport chunks as complete SSE frames — warning spam and silently unscanned output text

  llm translation
  BerriAI/litellm#49449 · Filter by author shaxiaozz opened on Oct 31, 2026
  3

• [Feature]: Scope the A2A contextId to the calling key for all A2A agents, as already done for Bedrock AgentCore

  llm translation
  BerriAI/litellm#49431 · Filter by author gulldan opened on Oct 31, 2026

• [Bug]: GET /v1/agents returns static_headers unmasked to proxy admins, unlike litellm_params secrets

  BerriAI/litellm#49430 · Filter by author gulldan opened on Oct 31, 2026

• [Bug]: X-A2A-{agent}-{header} forwarding lets any caller set Authorization and Cookie on the agent request

  BerriAI/litellm#49427 · Filter by author gulldan opened on Oct 31, 2026

• [Bug]: A2A calls are logged under the client's JSON-RPC id, so SpendLogs drops calls that reuse an id

  BerriAI/litellm#49424 · Filter by author gulldan opened on Oct 31, 2026
  2

• [Bug]: TransportError in litellm/llms/vertex_ai/vertex_llm_base.py

  bugllm translation
  BerriAI/litellm#49419 · Filter by author Eshwar-Naidus opened on Oct 31, 2026
  3

• [Bug]: voyage/ embeddings ignore VOYAGE_API_BASE

  bug
  BerriAI/litellm#49411 · Filter by author toto-dev opened on Oct 31, 2026
  2

• [Bug]: SDK stream wrapper response_cost stays at 0.0 after exhaustion despite priceable usage

  llm translation
  BerriAI/litellm#49369 · Filter by author OMGBrewmaster opened on Oct 31, 2026
  1

• [Bug]: Bedrock Converse sends temperature to Grok 4.6/4.7 and Kimi K3, which reject it

  llm translation
  BerriAI/litellm#49366 · Filter by author IsmaelMartinez opened on Oct 31, 2026

• [Bug]: MCP tool execution failures drop is_error flag and do not propagate to Anthropic tool_result message

  llm translation
  BerriAI/litellm#49359 · Filter by author chauvuusvn opened on Oct 30, 2026
  3

• [Bug]: handle_nlp_cloud_chunk silently edits and truncates completions whose text contains "[DONE]"

  llm translation
  BerriAI/litellm#49332 · Filter by author lobo235 opened on Oct 30, 2026

• [Bug]: Exact-match cache is shared across virtual keys and teams

  claude code
  BerriAI/litellm#49313 · Filter by author MaxRink opened on Oct 30, 2026
  3

• [Bug]: Exact-match response cache never hits when OTel is enabled and enable_caching_on_provider_specific_optional_params is true

  claude code
  BerriAI/litellm#49311 · Filter by author MaxRink opened on Oct 30, 2026
  1

• Anthropic messages guardrail translation drops tool fields

  llm translation
  BerriAI/litellm#49253 · Filter by author sathvikchebrolu45 opened on Oct 30, 2026
  1

• [Bug]: MCP true_passthrough: expired upstream OAuth token returns 200 with empty tools/list instead of 401, so clients never refresh

  bug
  BerriAI/litellm#49223 · Filter by author gregoryjflaherty opened on Oct 30, 2026
  1

• [Feature]: Retrieve model alias health or model alias mappings

  enhancement
  BerriAI/litellm#49219 · Filter by author IgnacioHeredia opened on Oct 30, 2026