Pplasmaintokenbay.hashnode.dev·Jul 28 · 9 min readHow to Connect n8n to Any OpenAI-Compatible API (Without a Custom Node)Most AI providers describe their APIs as OpenAI-compatible. That usually means you can send an OpenAI-style request to a different base URL, but it does not mean every SDK, endpoint, or feature behave00
Pplasmaintokenbay.hashnode.dev·Jul 23 · 10 min readStructured Outputs Across LLM Providers: A Compatibility Test“OpenAI-compatible” sounds reassuring until you rely on something beyond a basic text completion. I ran into this while working with structured outputs. The same request shape could work perfectly wit00
Pplasmaintokenbay.hashnode.dev·Jul 21 · 14 min readBuild a Multi-Model Evaluation CLI in Node.jsTesting a prompt against one model in a chat interface tells me almost nothing about how it will behave in production. The output may look good, but I still do not know: whether another model follows00
Pplasmaintokenbay.hashnode.dev·Jul 16 · 14 min readBuild an LLM API Health Checker with Node.js and ZodA normal API health check usually asks one question: Did the server return a successful HTTP status? That is useful, but it is not enough for an LLM API. An endpoint can return 200 OK while still be00
Pplasmaintokenbay.hashnode.dev·Jul 15 · 8 min readMy LLM Provider Returned 200. The Workflow Still Failed.The request looked completely normal. The server returned 200 OK. The response body was valid JSON. The latency was acceptable. Nothing appeared in the error logs. So I marked the provider integration00
