Flask Based API for Document Retrieval

Flask-Based API for Document Retrieval with Pinecone, Caching, Rate Limiting, and Background Scraping

Project Overview

This project is a Flask-based API designed to retrieve documents using Pinecone for vector search. It includes features like:

• Caching for faster retrieval
• Rate limiting to control API usage
• Background scraping to periodically update the database
• Dockerization for easy deployment and scalability

The application uses:

• Pinecone for vector-based document retrieval
• Hugging Face Transformers (BERT) for generating text embeddings
• Flask-SQLAlchemy for user management and tracking API usage
• Flask-Caching for caching API results
• Flask-Limiter for rate-limiting users
• Docker for packaging the app into a containerized environment

---

Approach and Project Flow

1. Setting Up the Flask API

We started by setting up the basic Flask application and API endpoints:

• /health: A simple endpoint to check if the API is running.
• /search: An endpoint to query Pinecone with text embeddings and retrieve results.

2. Embedding Generation with BERT

For each query, we generate embeddings using a pre-trained BERT model (via Hugging Face’s transformers library). These embeddings are used to perform vector searches using Pinecone.

3. Integration with Pinecone

We integrated Pinecone, a vector database, to store and query document embeddings. This allows efficient and fast retrieval of documents based on similarity search.

4. Rate Limiting and User Management

We implemented rate limiting using Flask-Limiter to restrict users from making more than 5 requests per minute:

• Users are tracked using a SQLite database with Flask-SQLAlchemy.
• If a user exceeds the rate limit, the API returns an HTTP 429 error (Too Many Requests).

5. Caching for Faster Retrieval

We added caching using Flask-Caching. Caching ensures that identical queries are served from memory, reducing the need to hit the database and vector search engine repeatedly. Cached results expire after 5 minutes.

6. Background Scraping

We implemented a background scraper that can scrape a user-provided website for articles or data and update the Pinecone index with new documents:

• Scraping is handled by BeautifulSoup.
• The scraping task runs in the background on a separate thread and updates the Pinecone index periodically.

7. Dockerization

We Dockerized the project using a Dockerfile. This allows the project to be easily deployed in any environment with consistent behavior across different systems.

---

Features

1. Document Retrieval: Retrieve documents based on similarity search using embeddings.
2. Rate Limiting: Prevent API abuse by limiting requests to 5 per minute per user.
3. Caching: Cache the results of similar queries for faster response times.
4. User Management: Track the number of API calls made by each user.
5. Background Scraping: Scrape websites in the background to continuously update the Pinecone index.
6. Dockerization: Easily run and deploy the application using Docker.

---

Project Structure

project/
├── app.py               # Main Flask application
├── database.py          # Database setup for user management
├── cache.py             # Caching configuration
├── limiter.py           # Rate limiting configuration
├── utils.py             # Utility functions (embedding, Pinecone query)
├── scraping.py          # Background scraping logic
├── requirements.txt     # Python dependencies
├── Dockerfile           # Docker configuration
├── .env                 # Environment variables (not committed to version control)
├── .dockerignore        # Ignore unnecessary files in the Docker build
└── README.md            # Project documentation

Key Files:

• app.py: Contains the Flask application and all API routes.
• database.py: Handles the setup and schema for user management using SQLite.
• cache.py: Manages caching for faster response times.
• limiter.py: Implements rate-limiting functionality.
• utils.py: Provides helper functions for generating embeddings and querying Pinecone.
• scraping.py: Contains the logic for background scraping and updating the Pinecone index.
• Dockerfile: Used to build and run the application in a Docker container.

---

Setup and Installation

Prerequisites:

• Python 3.9+
• Docker

Step 1: Clone the Repository

git clone <repository-url>
cd project

Step 2: Set Up a Virtual Environment (Optional but Recommended)

python -m venv venv
source venv/bin/activate  # On Windows, use venvScriptsactivate

Step 3: Install the Dependencies

pip install -r requirements.txt

Step 4: Set Up Environment Variables

Create a .env file in the project root and add your Pinecone API key and environment:

PINECONE_API_KEY=your_pinecone_api_key
PINECONE_ENVIRONMENT=your_pinecone_environment

Step 5: Initialize the Database

To set up the database, run the following code:

>>> from app import db, app
>>> with app.app_context():
>>>     db.create_all()

Step 6: Run the Application

python app.py

The app will be running at http://localhost:5000.

Docker Setup

Step 1: Build the Docker Image

docker build -t flask-app .

Step 2: Run the Docker Container

docker run -p 5000:5000 flask-app

Now, your app will be running at http://localhost:5000.

API Endpoints

1. Health Check

URL: /health Method: GET Description: Checks if the API is running. Response:

json
Copy code
{
  "status": "API is running"
}

2. Search

URL: /search Method: POST Description: Search documents based on text queries. Request Body:

json
Copy code
{
  "query": "Your search query",
  "user_id": "user123",
  "top_k": 3
}

Response: Returns a list of matching documents based on the query.

3. Start Scraping URL: /start_scraping Method: POST Description: Starts the background scraping process for a specific site. Request Body:

json
Copy code
{
  "url": "https://example.com"
}

Response:

json

{
  "message": "Started scraping for https://example.com"
}

Troubleshooting

Common Issues:

1. Rate Limit Exceeded: If you hit the rate limit, the API will return a 429 error.
2. Caching Delay: If cached results are returned, you might need to wait 5 minutes before new results appear.
3. Logs: The application logs all requests and errors in api.log. Background scraping logs are written to scraping.log.

Future Enhancements

1. Authentication: Adding API key-based authentication for added security.
2. Improved Error Handling: More detailed error messages for invalid queries or scraping failures.
3. Support for Multiple Scraping Sites: Enhance the scraper to handle multiple sites in parallel.