Health Influencer Claim Verification

This project provides a Node.js (Express) service that automatically fetches (or stores) health-related posts/tweets from an influencer, detects any health claims they make, de-duplicates repeated claims, and uses OpenAI to categorize and verify the validity of these claims. It stores the results in a PostgreSQL database and exposes a REST API for a front-end or other clients to consume.

Features

1. Influencer Management

   • Create or update influencers (name, handle, follower count) in a PostgreSQL database.
   • Fetch a list of all influencers or retrieve detailed info (including claims) for a specific influencer.

2. Claim Extraction & Verification

   • Pull (mock) tweets or text from the influencer.
   • Use OpenAI GPT to automatically extract health-related claims.
   • De-duplicate repeated claims (a naive approach by default---could integrate advanced methods).
   • Categorize each claim (e.g., Nutrition, Medicine, Mental Health, Fitness, Other).
   • Verify the claim ("Verified," "Questionable," or "Debunked") and assign a confidence score (0--100).

3. Trust Scoring

   • Compute an overall trust score for each influencer, based on the average confidence of their claims.

4. Database Storage

   • Stores influencers and claims in a PostgreSQL database with relationships.
   • Each claim includes text, category, verification status, confidence score, and timestamp.

5. RESTful Endpoints

   • GET /influencers -- Lists all influencers.
   • GET /influencers/:id -- Retrieves a single influencer's details and claims.
   • POST /influencers -- Creates or updates an influencer.
   • POST /analyze/:id -- Runs the OpenAI-based analysis pipeline for a given influencer.

How It Works

1. Create an Influencer

   • POST their name, handle (unique), and follower count to /influencers.

2. Analyze

   • When you call POST /analyze/:id, the service:
     1. Fetches (or simulates) tweets for that influencer.
     2. Uses OpenAI to extract health-related claims from the text.
     3. De-duplicates repeated claims.
     4. Categorizes each claim into broad categories (Nutrition, Medicine, etc.).
     5. Verifies each claim with a "Verified," "Questionable," or "Debunked" label and a confidence score (0--100).
     6. Stores everything in the claims table, updating the influencer's trust_score.

3. View Results

   • GET /influencers to see all influencers, or /influencers/:id to see that influencer's claims in detail.

Tech Stack

• Node.js / Express for the backend API
• PostgreSQL for data storage
• OpenAI for LLM-based claim extraction, categorization, and verification
• Docker Compose (optional) to run PostgreSQL and the Node service together

Project Structure

go

Code kopieren

├── backend/
│   ├── Dockerfile
│   ├── package.json
│   ├── server.js        // Main Express app
│   ├── db.js            // Database connection & table init
│   └── ...
└── docker-compose.yml

Installation & Setup

1. Clone the repository.

2. Set up your environment variables in a .env file or export them in your shell:

   bash

   Code kopieren

   DB_HOST=localhost DB_USER=postgres DB_PASS=postgres DB_NAME=health_db OPENAI_API_KEY=sk-xxx PORT=5001

3. Install dependencies:

   bash

   Code kopieren

   cd backend npm install

4. Run with Docker Compose (recommended):

   bash

   Code kopieren

   docker-compose up --build

   • This spins up Postgres and the Node server on port 5001 (or whichever you configure).

   • Or, to run locally without Docker, ensure Postgres is running separately, then:

     bash

     Code kopieren

     npm start

Usage

1. Create an influencer:

   bash

   Code kopieren

   curl -X POST -H "Content-Type: application/json"\ -d '{"name":"Dr. Health","handle":"drhealth","follower_count":10000}'\ http://localhost:5001/influencers

2. Check influencer list:

   bash

   Code kopieren

   curl http://localhost:5001/influencers

3. Analyze claims for influencer ID 1:

   bash

   Code kopieren

   curl -X POST http://localhost:5001/analyze/1

4. View updated influencer details (including claims):

   bash

   Code kopieren

   curl http://localhost:5001/influencers/1

Customization

• Tweet Ingestion: Currently, tweets are mocked. Integrate real Twitter data or store tweets in a DB table if desired.
• De-duplication: A naive string check is used by default; you can integrate more advanced embedding-based solutions.
• Verification: Modify the OpenAI prompts or use a different model for more accurate claim verification.
• Frontend: Build a React/Vue/Angular dashboard calling these endpoints to visualize influencer data, claims, and trust scores.

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you'd like to change.

License

MIT -- Feel free to use and adapt this project to your own needs.