Continue browsing this topic cluster with SEO-safe static pagination.
LiteLLM LiteLLM Updated May 21, 2026
Fix phantom budget blocking in LiteLLM proxy where Redis spend:end_user:* counters accumulate unreleased reservations, causing random HTTP 429 BudgetExceededError for end users who haven't reached their budget Includes evidence for LiteLLM troubleshooting demand.
BudgetExceededError: Budget has been exceeded! EndUser=<id> Current cost: $52.195456, Max budget: $50.0 — actual DB spend shows $0.001; Redis counter inflates without finalization, cycling every ~4 minutes LiteLLM LiteLLM Updated May 21, 2026
Fix mid-stream fallback not triggering in LiteLLM Router when primary model fails during streaming; users see errors instead of automatic model switching Includes evidence for LiteLLM troubleshooting demand.
MidStreamFallbackError not handled during Router.aresponses() streaming — mid-stream model fallback silently fails without retrying on secondary model LiteLLM LiteLLM Updated May 21, 2026
Fix GCP IAM authentication failure when running LiteLLM proxy with Redis Cluster — the sync auth path creates a standalone redis client via redis_connect_func that doesn't participate in Redis Cluster topology Includes evidence for LiteLLM troubleshooting demand.
GCP IAM auth fails with Redis Cluster — sync path uses redis_connect_func which RedisCluster bootstrap ignores LiteLLM LiteLLM Updated May 21, 2026
Fix premature connection drops (~60-70s) in LiteLLM proxy streaming — likely proxy timeout or keep-alive configuration issue Includes evidence for LiteLLM troubleshooting demand.
The litellm call will disconnect after 60-70 seconds LiteLLM LiteLLM Updated May 21, 2026
Fix persistent ghost model entries in LiteLLM dashboard when running multi-instance proxy — deletion not propagating through Redis-backed model registry Includes evidence for LiteLLM troubleshooting demand.
Deleted model reappears in /v2/model/info across multiple LiteLLM instances LiteLLM LiteLLM Updated May 21, 2026
Fix missing usage/spend tracking in LiteLLM proxy when upstream provider (e.g., Ollama Cloud) returns duplicate response IDs causing log collision Includes evidence for LiteLLM troubleshooting demand.
Spend logs silently dropped when provider returns non-unique response IDs (Ollama Cloud) LiteLLM LiteLLM Updated May 21, 2026
Improve LiteLLM proxy error propagation so timeout, auth, and rate-limit failures from upstream models are surfaced with correct error codes to callers Includes evidence for LiteLLM troubleshooting demand.
LiteLLM proxy routes requests to upstream models but fails to propagate error details when upstream timeout/auth failures occur, returning generic proxy errors LiteLLM LiteLLM Updated May 21, 2026
Fix LiteLLM Bedrock authentication failure caused by case-sensitive Content-Type header key collision causing double headers in SigV4 signing Includes evidence for LiteLLM troubleshooting demand.
SigV4 401 Unauthorized on bedrock/claude_platform/* route — signature never matches due to duplicate Content-Type header values LiteLLM LiteLLM Updated May 21, 2026
Fix LiteLLM proxy fallback chain broken by leaked internal rate-limiting parameters into provider API calls Includes evidence for LiteLLM troubleshooting demand.
BadRequestError / RateLimitError: Unrecognized request arguments supplied: _litellm_rate_limit_descriptors, _litellm_tpm_reserved_model (when using virtual keys with tpm/rpm limits + fallback chains) LiteLLM LiteLLM Updated May 21, 2026
Understand why LiteLLM masks rate limit 429 errors with misleading 'no deployments' message instead of surfacing the actual rate limit Includes evidence for LiteLLM troubleshooting demand.
RateLimitError: Error code: 429 - {'error': {'message': 'No deployments available for model…'}} LiteLLM LiteLLM Updated May 21, 2026
Developer runs LiteLLM gateway; model pricing and OpenRouter pricing fetches hang for ~3 minutes due to 60-second AbortSignal timeout on internal network requests that should succeed Includes evidence for LiteLLM troubleshooting demand.
model-pricing fetches fail with 60s AbortSignal timeout LiteLLM LiteLLM Updated May 21, 2026
开发者在使用 LiteLLM 代理后端时遇到 400 错误但 opencode 不显示日志或错误信息，需要调试和解决请求格式问题 Includes evidence for LiteLLM troubleshooting demand.
InvalidHTTPResponse fetching litellm proxy — 400 error not surfaced to user, retrying fails silently, must start new session LiteLLM LiteLLM Updated May 21, 2026
开发者配置 LiteLLM 代理后，Gemini 模型返回 NotFoundError 导致标题生成失败，需要正确的模型映射配置 Includes evidence for LiteLLM troubleshooting demand.
litellm.NotFoundError: GeminiException — Error generating conversation title with LLM: Invalid model name or configuration passed to LiteLLM proxy LiteLLM OpenRouter Updated May 21, 2026
Developers integrating OpenRouter OAuth (model provider auth via OpenRouter) get opaque errors without description or hint fields, making it impossible to programmatically handle auth failures. Includes evidence for OpenRouter troubleshooting demand.
OAuth authorization code flow and device/auth requests fail without standard OAuth error_description/hint error fields, leaving clients unable to parse failure reasons LiteLLM LiteLLM Updated May 21, 2026
Users deploying LiteLLM proxy with Vertex AI/Gemini streaming see silent failures where 429 errors are hidden — stream appears successful but response is truncated. Need to detect and handle mid-stream rate limits properly. Includes evidence for LiteLLM troubleshooting demand.
LiteLLM terminates stream with finish_reason="stop" instead of raising 429 RESOURCE_EXHAUSTED; client receives [DONE] with no error indication LiteLLM LiteLLM Updated May 21, 2026
Users of LiteLLM proxy with AWS Bedrock or Vertex AI hit timeouts but the timeout config is silently ignored, leaving requests hanging for 600s before surfacing to user Includes evidence for LiteLLM troubleshooting demand.
litellm.Timeout: Connection timed out. Timeout passed=600.0, time taken=600.006 seconds; error returned directly to user LiteLLM LiteLLM Updated May 21, 2026
Fix LiteLLM proxy failing to route requests due to missing or misconfigured model entries Includes evidence for LiteLLM troubleshooting demand.
litellm.NotFoundError: No available model in LiteLLM proxy configuration LiteLLM LiteLLM Updated May 21, 2026
Production LiteLLM proxy user expects incremental streaming for Anthropic model calls but receives all content in a single buffered batch — breaks real-time chat UX Includes evidence for LiteLLM troubleshooting demand.
Proxied Anthropic Response is Buffered: response is sent in one large batch despite stream=true setting; works correctly for OpenAI models under same config LiteLLM LiteLLM Updated May 21, 2026
After upgrading to LiteLLM v1.85.0, AWS Bedrock proxy requests return HTTP 200 but with empty response body — clients receive success status but no content; silent data loss affecting production inference pipelines Includes evidence for LiteLLM troubleshooting demand.
[Bug]: Bedrock pass-through /converse returns 200 with empty body on v1.85.0 LiteLLM LiteLLM Updated May 21, 2026
LiteLLM MCP (Model Context Protocol) gateway crashes when listing available tools; concurrent cancel-scope causes RuntimeError; schema migration drops columns on service restart breaking persisted tool configs Includes evidence for LiteLLM troubleshooting demand.
MCP Gateway: tools/list crashes with cancel-scope RuntimeError + schema columns drop on restart