Error category
Fix LiteLLM proxy, gateway, virtual key, provider routing, budget, schema validation, and OpenAI-compatible API errors.
This LiteLLM troubleshooting hub collects real error signatures, quick fixes, common causes, and step-by-step debugging paths for developers who need a practical answer instead of a broad overview.
150 articles in this category.
Start by matching the exact error message, then check the technology, environment, credentials, network path, and deployment context. The pages below are grouped so you can move from broad LiteLLM symptoms to specific root-cause families without relying on client-side search.
Fix LiteLLM proxy errors when calling Anthropic adaptive-thinking models, where LiteLLM incorrectly forces tool_choice parameter and causes API requests to be rejected by Anthropic. Includes evidence for LiteLLM troubleshooting demand.
LiteLLM forces tool_choice parameter on adaptive-thinking Anthropic models (Claude Opus 4.7 etc.) causing request rejection — fix applied in pull #28114 Diagnose and fix OpenRouter model routing errors when models return unclear failure messages, affecting production deployments using OpenRouter as a paid API proxy layer. Includes evidence for OpenRouter troubleshooting demand.
OpenRouter model routing returns ambiguous error when requested model is unavailable or disabled, without distinguishing between temporary unavailability and permanent deprecation Implement fair queuing in LiteLLM proxy to prevent single client from monopolizing model capacity and causing starvation for other tenants Includes evidence for LiteLLM troubleshooting demand.
No fair queuing — traffic burst from one client starves other clients' requests Developer using LiteLLM proxy finds that restricted users can see full model list via GET /v1/models endpoint despite access group restrictions, creating a security/authorization inconsistency Includes evidence for LiteLLM troubleshooting demand.
GET /v1/models ignores user.models restriction — shows all proxy models regardless of user access groups Fix LiteLLM API connection timeout errors by adjusting request_timeout or retry settings Includes evidence for LiteLLM troubleshooting demand.
litellm.APIConnectionError: Request timed out. Please increase the max_retries parameter. LiteLLM proxy admin sees usage data from end-of-month incorrectly carried over into next month's total, causing 'Month-to-Date' counter not to reset, leading to premature budget exhaustion Includes evidence for LiteLLM troubleshooting demand.
[Bug]: Duplicate Usage Aggregation Across Billing Cycles — April 30th tokens double-counted in May, pre-exhausting balances LiteLLM proxy admin running on Kubernetes finds that querying Usage page with large datasets triggers OOMKill on Prisma Query Engine, leaving proxy alive but unable to communicate with database Includes evidence for LiteLLM troubleshooting demand.
[Bug]: Prisma Query Engine unable to recover from OOMKill in K8s deployment — query-engine-de stuck in defunct state after large dataset access Developer using OpenRouter free tier encounters sudden 404 errors as models get removed, rate-limited, or replaced. Needs programmatic handling and fallback model rotation Includes evidence for OpenRouter troubleshooting demand.
Model claude-3-sonnet is not available / free LLMs on OpenRouter keep going 404 Fix rate limit handling in LiteLLM proxy where upstream 429s are incorrectly re-wrapped as connection errors and leak security-sensitive token hashes Includes evidence for LiteLLM troubleshooting demand.
Rate limit error message body leaks full SHA-256 token hash on 429 responses; wrapped as APIConnectionError instead of RateLimitError Debug LiteLLM proxy rate limiting and timeout issues when aggregating multiple LLM providers Includes evidence for LiteLLM troubleshooting demand.
RateLimitError: Request failed due to rate limiting. Retry after 30 seconds Fix LiteLLM returning 401 authentication error when using Azure AD token provider with DefaultAzureCredential Includes evidence for LiteLLM troubleshooting demand.
Authentication Error 401 when connecting Azure OpenAI via LiteLLM Fix LiteLLM proxy returning 400 BadRequest when Claude Code attempts thinking mode with DeepSeek models Includes evidence for LiteLLM troubleshooting demand.
API Error: 400 litellm.BadRequest Fix LiteLLM stream memory leak — prevent unbounded tool_call arguments accumulation during streaming Includes evidence for LiteLLM troubleshooting demand.
Unbounded memory growth in LiteLLM stream — current_tool_calls[i]['arguments'] += args_delta accumulates without limit Fix LiteLLM proxy budget bypass vulnerability that allows users to exceed spending limits Includes evidence for LiteLLM troubleshooting demand.
Budget bypass vulnerability allowing unlimited spending in LiteLLM proxy (infinite money glitch) Fix LiteLLM streaming fallback error: mid-stream fallback adds unsupported assistant prefill block, causing HTTP 400 on Claude Sonnet 4.6 / Opus 4.7 Includes evidence for LiteLLM troubleshooting demand.
litellm.BadRequestError: AnthropicException - {"type":"error","error":{"type":"invalid_request_error","message":"This model does not support assistant message prefill. The conversation must end with a user message."}} Fix inability to remove budget limits from LiteLLM API keys once they have been configured, blocking billing flexibility Includes evidence for LiteLLM troubleshooting demand.
budget_limits on LiteLLM keys cannot be removed/overridden after initial set via UI or API Fix LiteLLM v1.84.0 Docker container failing to start Includes evidence for LiteLLM troubleshooting demand.
Docker image ghcr.io/berriai/litellm:1.84.0 start failure in container logs Fix LiteLLM fallback configuration not triggering when primary model fails Includes evidence for LiteLLM troubleshooting demand.
No fallback model group found for original model_group=primary Fix LiteLLM Azure OpenAI authentication error when using DefaultAzureCredential — works with direct OpenAI client but fails through LiteLLM proxy Includes evidence for LiteLLM troubleshooting demand.
litellm.AuthenticationError: AzureException AuthenticationError - Access denied due to invalid subscription key or wrong API endpoint Fix LiteLLM OpenRouter image generation failure with ImageUsage Pydantic validation error Includes evidence for LiteLLM troubleshooting demand.
Error transforming OpenRouter image generation response: 2 validation errors for ImageUsage output_tokens: Input should be a valid integer, got a number with a fractional part [type=int_from_float, input_value=14417.92] Continue through the full static archive for this hub. Every listed page is crawlable and keeps its existing canonical URL.