Issues

• [Bug]: Streamed /v1/completions without stream_options puts "usage": {0, 0, 0} on every chunk

  llm translation
  BerriAI/litellm#47953 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: /v1/messages/count_tokens returns 500 for a search_result block, and for any content block type the local counter does not list

  llm translation
  BerriAI/litellm#47952 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: /v1/completions with response_format returns 500 "AsyncCompletions.create() got an unexpected keyword argument 'response_format'"

  llm translation
  BerriAI/litellm#47951 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: A Chat message with an unknown content part type, or a Completions prompt of the wrong type, gets HTTP 500 APIConnectionError and is retried

  llm translation
  BerriAI/litellm#47950 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: Errors relayed from an OpenAI-compatible backend lose param and arrive as a stringified JSON (or Python repr) inside message

  llm translation
  BerriAI/litellm#47949 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: /v1/messages still answers auth, model-access and malformed-JSON errors in the OpenAI error shape

  llm translation
  BerriAI/litellm#47948 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: Proxy rewrites JSON integers larger than 64 bits as floats before forwarding (e.g. a tool schema maximum of 12345678901234567890123 reaches the model as 1.2345678901234568e+22)

  BerriAI/litellm#47947 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: Non-streamed /v1/messages response has no stop_sequence key unless a stop sequence fired (hosted_vllm)

  llm translation
  BerriAI/litellm#47946 · Filter by author Vezio opened on Oct 22, 2026

• [Feature]: hosted_vllm Chat Completions: keep vLLM's stop_reason and token_ids on the choice, as /v1/completions and streamed Chat already do

  llm translation
  BerriAI/litellm#47945 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: Streamed /v1/responses through the proxy has no SSE event: lines and ends with data: [DONE]

  llm translation
  BerriAI/litellm#47944 · Filter by author Vezio opened on Oct 22, 2026

• [Bug]: Logging Request Metadata - Crashes

  bug
  BerriAI/litellm#47935 · Filter by author bd-kwoods opened on Oct 22, 2026

• [Bug]: Vertex AI rerank reloads Google credentials on every request

  llm translation
  BerriAI/litellm#47932 · Filter by author demetere opened on Oct 22, 2026

• [Bug]: Groq rejects the next turn after a streamed reasoning reply because reasoning_content is sent back

  llm translation
  BerriAI/litellm#47931 · Filter by author ishan-1010 opened on Oct 22, 2026
  1

• [Bug]: Responses WebSocket bypasses request-selected pre-call guardrails

  llm translation
  BerriAI/litellm#47927 · Filter by author shyamsunder1230 opened on Oct 22, 2026

• [Bug]: reasoning_effort="none" omits thinking instead of disabling it on default-on Claude models

  llm translation
  BerriAI/litellm#47924 · Filter by author snozwoz opened on Oct 22, 2026
  2

• [Feature]: Support WAV format for Elevenlabs

  enhancement
  BerriAI/litellm#47923 · Filter by author gouglhupf opened on Oct 22, 2026

• [Bug]: ModelNotMappedError in litellm/utils.py

  bugllm translation
  BerriAI/litellm#47922 · Filter by author kahidna opened on Oct 22, 2026
  1

• [Bug]: /v1/responses rejects spec-valid assistant history messages (output_text parts) with HTTP 500

  llm translation
  BerriAI/litellm#47918 · Filter by author mwielgosz-esky opened on Oct 22, 2026
  1

• [Bug]: spend is 0 when base_model names a priced model and custom_llm_provider differs

  llm translation
  BerriAI/litellm#47908 · Filter by author wk722 opened on Oct 22, 2026
  1

• [Bug]: Backport eMudhra root CA fix (#47670) to stable/1.106.x — 1.106.x images still trust emSign ECC Root CA - G3

  BerriAI/litellm#47904 · Filter by author JehandadK opened on Oct 22, 2026
  5

• [Bug]: OTel v2 LangfuseMapper drops langfuse.environment / release / version (parity regression from OTel v1)

  llm translation
  BerriAI/litellm#47898 · Filter by author Praveena-617 opened on Oct 22, 2026
  1

• [Feature]: Per-model-group (or per-key) opt-in for cancel_on_disconnect

  llm translation
  BerriAI/litellm#47896 · Filter by author deepanshululla opened on Oct 22, 2026

• [Bug]: Tool results with images fail on Gemini 2.5 since v1.89.0 ("Multimodal function responses are not supported for this model")

  claude codellm translation
  BerriAI/litellm#47895 · Filter by author Atharva-Kanherkar opened on Oct 22, 2026
  1

• [Bug]: provider: litellm_fallback create fails with "Root object was present, but now absent" with multiple proxy replicas

  BerriAI/litellm#47888 · Filter by author kris-runzer opened on Oct 22, 2026

• v1.83.14-stable arm64 image contains amd64 binaries (exec format error on ARM Linux)

  BerriAI/litellm#47887 · Filter by author stefano-edgible opened on Oct 21, 2026