Issues

• [Feature]: Let a session-affinity pin yield when its deployment is saturated and another deployment in the group is idle

  BerriAI/litellm#45868 · Filter by author lobo235 opened on Oct 10, 2026

• [Bug]: Anthropic batch accounting undercounts 1h cache writes and omits web-search fees

  llm translation
  BerriAI/litellm#45830 · Filter by author mavericksea-ai opened on Oct 10, 2026

• [Bug]: Getting 500 error

  bug
  BerriAI/litellm#45825 · Filter by author adarsha193 opened on Oct 10, 2026

• [Feature]: Model Additon: Microsoft Decison 1

  enhancementllm translation
  BerriAI/litellm#45807 · Filter by author wagnerjt opened on Oct 10, 2026
  1

• [Bug]: OpenTelemetry callback records tool-call responses as an empty message (gen_ai.output.messages has "parts": []), and "stop" as the finish reason for single-chunk streamed tool calls

  bugllm translation
  BerriAI/litellm#45796 · Filter by author armanunix opened on Oct 10, 2026
  1

• [Bug]: Pass-through endpoint URL containing 'predict' substring is misidentified as Vertex AI route — KeyError('predictions') kills logging task, SpendLogs never written

  llm translation
  BerriAI/litellm#45787 · Filter by author yelnyboy opened on Oct 10, 2026
  1

• [Bug]: Vertex AI serves a STALE access token up to the second it expires, Google answers 401 ACCESS_TOKEN_EXPIRED

  llm translation
  BerriAI/litellm#45785 · Filter by author mindw opened on Oct 10, 2026

• [Bug]: OpenTelemetry callback creates no span for a streaming call the app stops reading early

  bugllm translation
  BerriAI/litellm#45736 · Filter by author armanunix opened on Oct 10, 2026
  2

• [Bug]: routing_strategy "cost-based-routing" breaks all sync Router methods with RouterRateLimitError (single healthy deployment) — permanently disables mcp_semantic_tool_filter

  llm translation
  BerriAI/litellm#45718 · Filter by author automatCZ opened on Oct 10, 2026
  2

• [Bug]: /v1/messages does not accumulate deployment-level tpm for native anthropic/... deployments

  llm translation
  BerriAI/litellm#45702 · Filter by author xinliu0727-sudo opened on Oct 10, 2026
  2

• [Bug]: Counting tokens for a base64 PDF has no cost bound, so a few uploads stall every token count on the worker

  bugllm translation
  BerriAI/litellm#45676 · Filter by author RachelHuangZW opened on Oct 9, 2026

• [Bug]: Claude models on Vertex AI fail at batch creation (404 batchPredictionJobs) — OpenAI batch translation doesn't cover Anthropic's native schema

  llm translation
  BerriAI/litellm#45671 · Filter by author manarayanareddy-clgx opened on Oct 9, 2026
  2

• [Bug]: Responses API mid-stream 429 can be recorded against the fallback deployment and cool it down

  llm translation
  BerriAI/litellm#45645 · Filter by author AlexanderKazakov opened on Oct 9, 2026

• [Bug]: chat->Responses->chat bridge converts string user content to text blocks, which breaks Databricks json_object ("must contain the word json" even when present)

  llm translation
  BerriAI/litellm#45618 · Filter by author jimmychen-p72 opened on Oct 9, 2026
  2

• [Bug]: custom_auth_run_common_checks never runs the key max_budget check (virtual_key_max_budget_check) on custom-auth requests

  llm translation
  BerriAI/litellm#45605 · Filter by author bxm156 opened on Oct 9, 2026
  2

• [Bug]: langfuse_otel never sends the completion start time, so Langfuse shows no time to first token

  llm translation
  BerriAI/litellm#45596 · Filter by author HardMax71 opened on Oct 9, 2026

• [Bug]: OTel v2 omits first-chunk timing on streaming pass-through requests

  llm translation
  BerriAI/litellm#45588 · Filter by author hiteshPrabhu opened on Oct 9, 2026

• [Bug]: OTel v2 metrics omit resolved provider labels on pass-through calls

  llm translation
  BerriAI/litellm#45587 · Filter by author hiteshPrabhu opened on Oct 9, 2026

• [Bug]: Missing usage in a provider response is reported as 0 tokens, which can't be told apart from a real 0

  bugllm translation
  BerriAI/litellm#45586 · Filter by author MichaelSoegaard opened on Oct 9, 2026
  2

• [Bug]: Langfuse callback drops the trace of every successful /v1/systemone and /v1/decisions call ('DecisionsUsage' object has no attribute 'get')

  llm translation
  BerriAI/litellm#45584 · Filter by author jeffreylasut opened on Oct 9, 2026
  1

• [Feature]: Prometheus metrics for team-member budgets (max, remaining, time to reset)

  BerriAI/litellm#45578 · Filter by author markyates opened on Oct 9, 2026

• [Bug]: stream_chunk_builder drops Bedrock thinking_blocks that carry a signature but no text (Claude Sonnet 5, display omitted)

  llm translation
  BerriAI/litellm#45575 · Filter by author vikram-typeset opened on Oct 9, 2026
  2

• [Bug]: DB-stored models encrypted before LITELLM_SALT_KEY was set are silently dropped (no log, health stays OK)

  BerriAI/litellm#45574 · Filter by author stanvx opened on Oct 9, 2026
  1

• [Bug]: Old agents/models/mcps not deregistered from Access Group

  bug
  BerriAI/litellm#45573 · Filter by author IgnacioHeredia opened on Oct 9, 2026
  2

• [Bug]: Guardrail creation wizard clips content horizontally

  bug
  BerriAI/litellm#45564 · Filter by author zanda8893 opened on Oct 9, 2026
  1