Latest archive
Recently updated troubleshooting pages, kept as a crawlable archive so new and refreshed fixes stay easy to discover.
Fix MCP tools in Claude Code returning user denied error even when permissions.allow contains correct server entry, no permission prompt appears Includes evidence for Claude Code troubleshooting demand.
tools spuriously rejected as 'user denied' without permission prompt Fix Claude Code Chrome MCP tools failing on all domains when navigating or reading pages Includes evidence for Claude Code troubleshooting demand.
"Navigation to this domain is not allowed" / "Permission denied for reading pages on this domain" 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 Fix Vercel deployment build failure caused by exceeding the limit of 12 Serverless Functions per deployment Includes evidence for Vercel troubleshooting demand.
No more than 12 Serverless Functions can be added to a Deployment Fix Vercel deployment build failure due to out of memory during build step Includes evidence for Vercel troubleshooting demand.
build exited with code 1 (OOM during Vite/vanilla-extract build bundle) Fix Claude Code on Windows where git worktree remove fails with Permission denied because MCP servers aren't shut down before worktree operations Includes evidence for Claude Code troubleshooting demand.
worktree removal fails with Permission denied due to MCP shutdown ordering Developer using Anthropic streaming API loses connection mid-stream (TransportError) and can't catch it with standard except anthropic.APIConnectionError blocks — must know to also catch httpx.TransportError Includes evidence for Anthropic API troubleshooting demand.
Mid-stream httpx.TransportError leaks through as bare httpx exceptions because SSE body iteration has no try/except — customers standard retry ladder misses mid-stream drops User ran /login in Claude Code on Windows, sees success message but tool remains unauthenticated and rejects all commands afterward Includes evidence for Claude Code troubleshooting demand.
/login shows "Login successful" but immediately reverts to "Not logged in". Any subsequent command fails with "Not logged in". User upgraded to Docker 29.5.0 and rootless mode stopped working — docker socket file disappears, can't run any docker commands Includes evidence for Docker troubleshooting demand.
failed to connect to the docker API at unix:///run/user/1000/docker.sock; check if the path is correct and if the daemon is running: dial unix /run/user/1000/docker.sock: connect: no such file or directory Developer calls langgraph.with_structured_output() expecting it to preserve pre-bound tools, but they are silently dropped — functional regression breaking LLM function calling workflows Includes evidence for LangChain troubleshooting demand.
[langchain-openai] with_structured_output() silently drops previously bound tools and lacks support for OpenAI native tool bindings 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 Developer using client.responses.create(background=True) cannot handle failure cases programmatically because the SDK returns 200 with failed payload instead of raising a typed exception Includes evidence for OpenAI API troubleshooting demand.
Background responses failures lack a stable code/name that maps to an exception class — HTTP poll returns 200 OK with status=failed but no typed exception is raised Fix ESM action crashes after @actions/glob v0.7.0 upgrade — broken import of Minimatch named export Includes evidence for GitHub Actions troubleshooting demand.
SyntaxError: The requested module 'minimatch' does not provide an export named 'Minimatch' (in @actions/glob v0.7.0) Fix Claude Code MCP server authentication returning HTTP 431 error on Linux platform Includes evidence for Claude Code troubleshooting demand.
HTTP ERROR 431 — MCP server authentication failure (browser restart!) Fix Docker rootless mode broken after upgrading to 29.5.0 — docker.sock missing Includes evidence for Docker troubleshooting demand.
failed to connect to the docker API at unix:///run/user/1000/docker.sock: dial unix /run/user/1000/docker.sock: connect: no such file or directory Fix OpenAI Batch API 404 error when using GPT-5 mini or nano models with chat completions endpoint Includes evidence for OpenAI API troubleshooting demand.
The model gpt-5-mini-2025-08-07-batch does not exist or you do not have access (HTTP 404 on /v1/chat/completions batch endpoint) Fix or work around Anthropic API rejecting valid but complex JSON schemas for structured outputs Includes evidence for Anthropic API troubleshooting demand.
400 invalid_request_error: "The compiled grammar is too large, which would cause performance issues" Fix OpenAI Python SDK failing to surface a typed exception for failed background Runs, preventing correct retry/backoff handling Includes evidence for OpenAI API troubleshooting demand.
status='failed', error.code='server_error' with HTTP 200 OK — no mapped SDK exception raised Fix AWS Bedrock SSE error events being swallowed with 200 status, making errors undetectable Includes evidence for Anthropic API troubleshooting demand.
SSE stream produces server-side errors but HTTP status_code stays 200 — errors invisible to caller Handle Anthropic Bedrock server-side errors (5xx/4xx responses) during streaming properly; understand transient error retry behavior and HTTP status exception handling Includes evidence for Anthropic API troubleshooting demand.
non-200 stream events raise ValueError instead of APIStatusError Fix Azure OpenAI enterprise S0 tier rate limiting for GPT-4.1 models; understand token rate limits vs free tier restrictions and how to increase default rate limits Includes evidence for OpenAI API troubleshooting demand.
RateLimitError — Requests to the ChatCompletions_Create Operation under Azure OpenAI API version 2025-01-01-preview have exceeded token rate limit of your current OpenAI S0 pricing tier Fix Claude Code OAuth authentication failure after auto-update preventing login to AI coding assistant Includes evidence for Claude Code troubleshooting demand.
OAuth Request Failed — This isn't working right now. You can try again later Fix Cline plugin failing to load in JetBrains IDEs due to health check timeout; alternative Cursor-like IDE error for same category Includes evidence for Cursor troubleshooting demand.
Healthcheck timed out — Failed to load Cline in IntelliJ/JetBrains Fix OpenAI 429 insufficient_quota error in n8n workflows despite having $18+ credits and low token usage; RPM limit bypass strategies Includes evidence for OpenAI API troubleshooting demand.
429 – You exceeded your current quota, please check your plan and billing details. Type: insufficient_quota Fix OpenAI Files API 400 bad request when uploading PDF and referencing in streaming responses; understand file input limitations with Azure OpenAI Includes evidence for OpenAI API troubleshooting demand.
Uploading PDF via Files API and using in Streaming gives 400 bad request Fix Cloudflare 524 timeout errors when serving large files through Cloudflare proxy, especially for AI/ML embedding workloads Includes evidence for Cloudflare troubleshooting demand.
524 Error from Cloudflare while loading big file for embedding — origin server timed out 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. Resolve unexpected GitHub Copilot rate limit message even when user has active PRO subscription Includes evidence for GitHub Copilot troubleshooting demand.
Oops, you reached the rate limit. Please try again later. Fix GitHub Copilot CLI authentication failure where no credentials are detected in any checked location Includes evidence for GitHub Copilot troubleshooting demand.
Error: No authentication information found Fix Vercel deployment failure caused by Vercel runtime resolver using require() instead of ESM path for middleware modules Includes evidence for Vercel troubleshooting demand.
MIDDLEWARE_INVOCATION_FAILED on Vercel deploy: Cannot find module — local build artifact does not reference ESM path Fix intermittent streaming crashes when calling Anthropic Claude via AWS Bedrock cross-region inference profiles; rate-limited responses appear as HTTP 200 error frames causing TypeError instead of catchable APIStatusError Includes evidence for Anthropic API troubleshooting demand.
AttributeError: 'NoneType' object has no attribute 'model' — Bedrock cross-region profile returns HTTP 200 with error payload type='rate_limit_error', SDK decodes as BetaRawMessageStartEvent with message=None Fix Docker daemon unexpectedly crashing with crypto hash function unavailable panic, preventing any container operations until daemon restart Includes evidence for Docker troubleshooting demand.
panic: crypto: requested hash function #0 is unavailable — GitHub goroutine crash in github.com/opencontainers/go-digest.Algorithm.Hash triggered by Docker image digest computation Developers running local Ollama models via Codex App notice extreme slowdowns; root cause is Codex App not reading model parameter num_ctx and defaulting to max context sizes Includes evidence for Ollama troubleshooting demand.
Codex App sends requests with context_window=128000-262144 tokens instead of model's configured num_ctx (e.g., 32768) — causing severe generation slowdowns on local models Developer needs to distinguish temporary rate limits from permanent quota exhaustion to trigger different retry logic (e.g., switch backup API key vs exponential backoff) Includes evidence for OpenAI API troubleshooting demand.
RateLimitError: 429 insufficient_quota — both rate-limit and quota-exhaustion map to same RateLimitError class Fix HTTP 500 MIDDLEWARE_INVOCATION_FAILED on Vercel production deployment caused by @swc/helpers module resolution failure when using Next.js 16.2.x App Router with proxy.ts middleware convention and Sentry installed Includes evidence for Vercel troubleshooting demand.
MIDDLEWARE_INVOCATION_FAILED: Cannot find module '/var/task/node_modules/@swc/helpers/esm/_interop_require_default.js' — next@16.2.x + proxy.ts + @sentry/nextjs Fix Claude Code SDK crashing immediately on startup during GitHub Actions CI, no API calls made Includes evidence for Claude Code troubleshooting demand.
error: Claude Code process exited with code 1; SDK execution error: 14 | depsCount Fix Cloudflare Wrangler failing to bundle Next.js 16 apps that use proxy.ts instead of deprecated middleware.ts Includes evidence for Cloudflare Workers troubleshooting demand.
wrangler Next.js integration does not recognize proxy.ts; still depends on old middleware.ts convention Fix Next.js 16.2.2 standalone deployment running out of memory due to cached streamed fetches leaking arrayBuffers under load Includes evidence for Vercel / Next.js troubleshooting demand.
cached internal streamed fetches cause unbounded arrayBuffers growth and OOM Fix streaming tool calls having incomplete function arguments when speculative decoding produces multiple tool_calls entries per chunk Includes evidence for OpenAI Python SDK troubleshooting demand.
accumulate_delta drops tool_call fragments when one chunk has multiple entries at the same index Developer using anthropic SDK streaming sees overloaded_error arrive as HTTP 200 SSE event; SDK creates bare APIStatusError(200) instead of OverloadedError(529); fallback/retry logic breaks because status check >= 500 never matches Includes evidence for Anthropic API troubleshooting demand.
Mid-stream SSE overloaded_error returns status_code=200 instead of 529 — fails fallback retry logic