Error listings
Fix permanently broken Anthropic conversation when extended thinking block signature is rejected by API after N turns Includes evidence for Anthropic API troubleshooting demand.
messages.X.content.Y: Invalid `signature` in `thinking` block Fix LangChain PII Middleware stripping or corrupting tool call arguments in structured output, causing pipeline data loss Includes evidence for LangChain troubleshooting demand.
PIIMiddleware state hooks miss tool-call args and corrupt structured content Fix Ollama tool calls freezing and timing out on Apple Silicon Macs with MLX backend, specifically when generating large code outputs via write tool Includes evidence for Ollama troubleshooting demand.
Tool calling is not streaming on macOS with MLX, causing timeout when write tool outputs large code Fix codex CLI crashing on startup when configured with OpenRouter model provider, preventing all agent operations Includes evidence for OpenAI API troubleshooting demand.
@openai/codex codex-cli v0.133.0 fails during startup when using an OpenRouter model provider profile; codex_models_manager refreshes the available model catalog Fix Claude Code MCP server repeatedly failing with revoked refresh token when sharing config between MCP server and Claude Code plugin Includes evidence for Claude Code troubleshooting demand.
invalid_grant — Refresh token revoked Resolve Claude Code CLI inability to connect to HTTP-based MCP servers caused by Bun runtime TCP socket failure Includes evidence for Claude Code troubleshooting demand.
FailedToOpenSocket error when connecting to HTTP MCP servers via Claude Code CLI; Bun runtime bundled in CLI binary fails at TCP socket level Fix Claude Code failing to connect to custom OAuth MCP servers that don't support DCR, with no error message shown Includes evidence for Claude Code troubleshooting demand.
HTTP OAuth MCP servers silently fail without error when provider doesn't support Dynamic Client Registration (DCR) Fix Next.js module resolution errors during Vercel production deployments related to turbopack and CommonJS package imports Includes evidence for Vercel troubleshooting demand.
Module not found error in Next.js turbopack builds on Vercel when importing CommonJS packages; works locally but fails in Vercel deployment Fix Ollama model download failures due to manifest pulling errors including missing files, SSH key issues, DNS resolution failures, and version compatibility Includes evidence for Ollama troubleshooting demand.
Error: pull model manifest: file does not exist; also 'ssh: no key found', 'no such host', and 'pulling manifest: 412: requires newer version' Fix OpenRouter rate limit failover behavior where model switch breaks completely after hitting $5 free tier spending cap Includes evidence for OpenRouter troubleshooting demand.
After hitting OpenRouter API rate limit ($5 spending cap), switching models fails with 'model not allowed' error for ALL models even after limit resets Resolve 400 error caused by invalid thinking block signatures when resuming conversations, affecting streaming chat and function calling workflows Includes evidence for Anthropic API troubleshooting demand.
API returns 400 'Invalid signature in thinking block' when replaying conversation history with no recovery path Fix rate limit / quota errors when delegating OpenAI API calls through third-party tools authenticated via ChatGPT Plus OAuth Includes evidence for OpenAI API troubleshooting demand.
429 quota exceeded when calling third-party apps using ChatGPT Plus OAuth tokens Fix OAuth authentication failure where transient server errors cause permanent credential corruption preventing Claude Code login Includes evidence for Claude Code troubleshooting demand.
OAuth refresh corrupts credentials state during transient upstream 5xx, causing persistent 401 authentication loop across sessions Fix deployment failures where Cloudflare Workers SDK returns 502 during asset upload phase Includes evidence for Cloudflare troubleshooting demand.
502 Bad Gateway during Differential Uploads POST /pages/assets/upload Fix token acquisition failure when setting up GitHub Copilot as a backend model in LiteLLM proxy Includes evidence for LiteLLM troubleshooting demand.
Error polling for access token when configuring github copilot model Fix LiteLLM proxy /v1/models endpoint so wildcard team BYOK routes properly expand and discover models stored under litellm_credential_name Includes evidence for LiteLLM troubleshooting demand.
/v1/models wildcard expansion fails to find team-scoped wildcard deployment models; team.models = ['openai/*'] does not resolve correctly when check_provider_endpoint=true Enable reliable error handling for async/background OpenAI API calls by providing consistent error codes and names that map to library exception classes Includes evidence for OpenAI API troubleshooting demand.
Background task responses fail without stable error_code/error_type values that can be mapped to Python exception classes for programmatic handling Resolve Ollama model loading failure caused by Apple Silicon MLX runner being incorrectly selected on Linux systems with NVIDIA GPU Includes evidence for Ollama troubleshooting demand.
Error: failed to load model: 500 Internal Server Error: mlx runner failed: Error: failed to initialize MLX Fix resource leak in langchain-openrouter where each model instantiation creates unclosed httpx connections, especially problematic with FastAPI dependency injection and LangGraph factory graphs Includes evidence for LangChain troubleshooting demand.
ChatOpenRouter.__init__ creates brand-new httpx.Client/AsyncClient per instance with no caching; under per-request instantiation patterns this leaks TLS keep-alive sockets and pool state to openrouter.ai Fix sporadic Docker Swarm overlay networking failures in docker-compose when defining services on overlay networks, where the network disappears between initialization steps Includes evidence for Docker troubleshooting demand.
Error response from daemon: failed to set up container networking: could not find a network matching network mode <overlay-network-name>: network <overlay-network-name> not found