Show HN: Query ClickHouse Databases from ChatGPT Custom GPTs via OpenAPI (github.com)

A lightweight FastAPI service provides an OpenAPI-compliant HTTP bridge that lets ChatGPT custom GPTs query ClickHouse data warehouses using natural language. To use it you run the bridge (uvicorn/Docker), add its openapi.json as a new action in chat.openai.com/gpts/editor, set Bearer token authentication (API_BEARER_TOKEN), and then ask natural-language questions — ChatGPT translates them to SQL and the bridge executes those queries against ClickHouse via its HTTP interface. The service exposes auto-generated Swagger/ReDoc docs (/docs, /redoc), returns ClickHouse JSONCompact by default, enables CORS, and uses token-based auth with sensible HTTP status codes for success, query errors, auth failures, and connection issues. Technically, the bridge is built on a modern Python stack (FastAPI, Pydantic, httpx), targets Python 3.11+, and supports containerized deployment (Docker). It’s a thin proxy to ClickHouse’s HTTP endpoint and exposes an OpenAPI schema so LLMs and orchestration tools can call it programmatically. For the AI/ML community this simplifies exploratory data analysis, prompt-driven feature inspection, and integrating warehouse queries into LLM workflows or agents — but it also underscores the need for careful auth, query permissions, and monitoring since natural-language-to-SQL automation can surface sensitive data or generate costly queries.