Extract API Specification Table from Code

The most common issue when having ChatGPT generate API documentation is hallucinated parameters not found in the code. Given raw code without strict constraints, AI often infers standard REST conventions and adds fields like page, limit, or sort. Developers referencing such docs end up sending invalid payloads and receiving 400 Bad Request errors.

The Task paragraph strictly specifies "only the endpoints explicitly present in the code" to prevent this. Explicitly listing the target fields—path, method, request, response, error—serves the same purpose. Without these constraints, AI defaults to writing just the path and a brief description, completely omitting the request body structure. Documenting error responses is equally critical, as success-only documentation is barely useful for frontend integration.

Specifying the framework upfront is essential because routing syntax and decorators vary widely across frameworks. Knowing the framework is required to determine whether @RequestParam is a query param or body, or if req.params represents path variables. Including the "Location" column in the Format ensures this distinction is preserved in the table.

The Constraints requiring "Not identifiable from code" and the (Inferred) tag act as essential guardrails for automatic API spec generation. If blanks are left open, AI tends to hallucinate plausible values, but defining fallback terminology forces it to acknowledge unknowns. Wrapping the code in """ prevents code comments (e.g., "TODO: ignore this") from being misinterpreted as prompt instructions.

While it produces a table, it frequently introduces non-existent parameters like page and limit, while omitting parameter locations and required flags, leading to failed API calls. Error responses are often skipped entirely, forcing frontend engineers to ask about error schemas manually. Pasting multiple endpoints at once also results in inconsistent formats across endpoints.