LLMCompatStart a test

OpenAI-Compatible API Troubleshooting Guides

Focused field notes for finding whether an OpenAI-compatible integration failed at the URL, network, authentication, model, server, or SDK layer.

Published
6 field notes
Format
Answer, evidence, verification

Published field notes

Start with the layer closest to the error, then verify the result against the endpoint.

Configuration · 7 minHow to Configure an OpenAI-Compatible API Base URLSeparate the origin, API root, and operation path. Then diagnose 401, 404, and model errors.Read guide Tool calling · 8 minOpenAI-Compatible Tool Calling: Test and Fix tool_callsCheck the tool definition, tool_calls response, JSON arguments, call ID, and tool-result round trip.Read guide Streaming · 7 minOpenAI-Compatible Streaming: Test and Fix SSE ResponsesCheck response headers, SSE frames, JSON events, completion markers, buffering, and early disconnects.Read guide Browser · 7 minOpenAI-Compatible API CORS Errors: Diagnose and Fix ThemCheck preflight responses, allowed origins, request headers, browser access, and safe proxy options.Read guide Structured output · 8 minOpenAI-Compatible JSON Mode: Test response_format and Fix Invalid JSONSeparate parameter support, JSON syntax, object shape, and application-level field validation.Read guide Local runtimes · 9 minLocal OpenAI-Compatible APIs: Ollama, vLLM, LM Studio, and llama.cppCompare API roots, server commands, model IDs, authentication, and browser CORS settings.Read guide

Need evidence from the endpoint?

Run the six-probe suite to distinguish configuration problems from capability and response-shape failures.

Open the compatibility tester