3,561 prompts from top AI collections
← Back to Market
[UNTESTED] [pavanvishnu08/Structured-Extraction-API] 1 endpoint(s)
Gemini API Proxy/Gemini
2,143 characters
Status: [UNTESTED] Endpoints found: http://127.0.0.1:8000/api/v1/extract Source: https://github.com/pavanvishnu08/Structured-Extraction-API # Structured Extraction API A FastAPI service that turns unstructured text (emails, messages, resumes) into **validated, predictable JSON** using Google Gemini's schema-constrained output. Every model response is re-validated with Pydantic v2 before it reaches a client, and failures are reported with safe, consistent errors. > Status: all features below are implemented and covered by an automated test suite that runs against a mocked provider. The service has **not** been deployed or run against the live Gemini API as part of this repository's creation; see [Deployment](#deployment) and [Known limitations](#known-limitations). ## Problem statement Businesses receive information as free text: candidate emails, support tickets, reports. Extracting fields by hand is slow and inconsistent. This API accepts text, asks an LLM to extract a fixed set of fields, validates the result against a strict schema, and returns the same JSON shape every time (or a clear error). Example input: ```text I'm Rahul Sharma. My email is rahul@example.com. I have 3 years of Python experience and I'm looking for a Backend Developer role in Hyderabad. ``` Example output (`model` and `processing_time_ms` are measured at runtime; the values below are illustrative): ```json { "success": true, "data": { "name": "Rahul Sharma", "email": "rahul@example.com", "years_of_experience": 3, "skills": ["Python"], "desired_role": "Backend Developer", "location": "Hyderabad" }, "metadata": { "model": "gemini-3.5-flash-lite", "processing_time_ms": 450, "request_id": "9f1c2d3e4a5b46788899aabbccddeeff" } } ``` ## Features - `POST /api/v1/extract`, `POST /api/v1/extract/batch`, `GET /api/v1/schema`, `GET /health`, `GET /ready` - Strict `CandidateProfile` validation (no silent type coercion, length limits, case-insensitive skill de-duplication) - Prompt designed to avoid invention and to treat input as untrusted data (with validation as the real safety net) -
Download .txt