# AI Doctor - Fix OpenAI 429 Errors, Timeouts & Streaming Issues

Works across OpenAI, Claude, Gemini, OpenClaw, LangChain, and other LLM stacks.

Fix common LLM API errors like **429 rate limits**, **timeouts**, **streaming hangs**, **retry loops**, and **runaway token costs**.

AI Doctor scans your codebase and shows exactly what is breaking - then helps you fix it automatically.

---

## Works with

### Providers
- OpenAI API
- Anthropic Claude API
- Google Gemini API
- Grok
- OpenRouter
- Together AI
- Fireworks AI
- Replicate
- DeepSeek, Mistral, Cohere (OpenAI-compatible)

### Frameworks & SDKs
- OpenClaw
- LangChain
- Vercel AI SDK
- LlamaIndex
- Custom OpenAI-compatible clients
- Node.js / JavaScript backends

---

## Quick start

```bash
npx ai-doctor
```

Takes ~30 seconds. No config needed.

## What AI Doctor detects & fixes

### Rate limits (429 errors)
- Fix OpenAI 429 RESOURCE_EXHAUSTED
- Fix LangChain / OpenClaw retry issues
- Add exponential backoff + jitter
- Handle Retry-After correctly

### Timeouts & hanging requests
- Fix context deadline exceeded
- Fix Vercel AI SDK timeout issues
- Prevent stuck API calls
- Add safe timeout guards

### Streaming issues (SSE problems)
- Fix OpenAI / Claude streaming hangs
- Fix incomplete responses
- Fix SSE not flushing correctly
- Detect broken streaming pipelines

### Retry problems
- Fix retry storms in LangChain / custom code
- Add retry limits
- Prevent infinite loops

### Cost issues
- Prevent runaway token usage
- Fix missing or incorrect max_tokens
- Reduce unexpected cost spikes

### Debugging & traceability
- Add request IDs
- Fix missing correlation IDs
- Improve logging for production debugging

## Common errors this helps with

- 429 RESOURCE_EXHAUSTED
- context deadline exceeded
- socket hang up
- ECONNRESET
- 504 Gateway Timeout
- streaming response stuck / incomplete
- OpenAI timeout errors
- Claude streaming issues
- Gemini API failures
- LangChain retry bugs
- OpenClaw streaming / retry issues

## Why use AI Doctor?

Most AI apps break in the same ways:

- Bad retry logic -> 429 errors + rate limits
- No timeouts -> requests hang forever
- Broken streaming -> partial responses
- No token limits -> runaway costs
- No tracing -> hard to debug production issues

AI Doctor finds all of this automatically.

## Commands

```bash
npx ai-doctor           # Scan + interactive fix
npx ai-doctor --fix     # Auto-fix issues
npx ai-doctor --ci      # Report only (CI mode)
```

## Install

```bash
npm install -g ai-doctor
```

## Documentation

https://github.com/michaelbrinkworth/ai-patch-doctor

## Keywords

- openai 429 error fix
- openai timeout fix
- openai streaming issue
- langchain 429 error
- vercel ai sdk timeout
- openclaw streaming issue
- anthropic api error fix
- gemini api timeout
- llm retry logic
- llm debugging tool
- llm cost control
- ai api errors
- llm observability cli

MIT License - Fix your AI API issues in minutes.