News RAG API Service

This service provides a REST API for retrieving and analyzing news articles using RAG (Retrieval Augmented Generation) with OpenAI's GPT models.

Setup

1. Clone the repository
2. Copy .env.default to .env and fill in your API keys:

cp .env.default .env

3. Install dependencies:

pip install -r requirements.txt

Environment Variables

Create a .env file with the following variables:

• OPENAI_API_KEY: Your OpenAI API key
• APP_API_KEY: API key for authenticating requests to this service

Running the Service

python app.py

The service will start on http://localhost:8000

API Endpoints

1. Retrieve News Articles

Endpoint: POST /api/retrieval

Retrieves relevant news articles based on a query.

# Using curl
curl -X POST "http://localhost:8000/api/retrieval" \
     -H "x-api-key: YOUR_APP_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"query": "latest AI developments"}'

# Using Python
import requests
import json

response = requests.post(
    "http://localhost:8000/api/retrieval",
    headers={
        "x-api-key": "YOUR_APP_API_KEY",
        "Content-Type": "application/json"
    },
    json={"query": "latest AI developments"}
)
print(json.dumps(response.json(), indent=2))

2. Generate Completion

Endpoint: POST /api/completion

Generates an AI response based on conversation history and relevant news articles.

# Using curl
curl -X POST "http://localhost:8000/api/completion" \
     -H "x-api-key: YOUR_APP_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"messages": [{"role": "user", "content": "What are the latest developments in AI?"}]}'

# Using Python
import requests
import json

response = requests.post(
    "http://localhost:8000/api/completion",
    headers={
        "x-api-key": "YOUR_APP_API_KEY",
        "Content-Type": "application/json"
    },
    json={
        "messages": [
            {"role": "user", "content": "What are the latest developments in AI?"}
        ]
    }
)
print(json.dumps(response.json(), indent=2))

Response Formats

Retrieval Response

{
  "articles": [
    {
      "title": "Article Title",
      "url": "https://example.com/article",
      "date": "2024-03-21"
    }
  ]
}

Completion Response

{
  "response": "Generated response with citations [Source 1] and analysis..."
}

Advanced Usage

Date Filtering

You can include date-related queries in your questions:

curl -X POST "http://localhost:8000/api/retrieval" \
     -H "x-api-key: YOUR_APP_API_KEY" \
     -H "Content-Type: application/json" \
     -d '{"query": "AI news from last week"}'

Conversation Context

The completion endpoint maintains conversation context:

messages = [
    {"role": "user", "content": "What are the latest AI developments?"},
    {"role": "assistant", "content": "According to recent news..."},
    {"role": "user", "content": "Tell me more about those developments"}
]

response = requests.post(
    "http://localhost:8000/api/completion",
    headers={
        "x-api-key": "YOUR_APP_API_KEY",
        "Content-Type": "application/json"
    },
    json={"messages": messages}
)

Error Handling

The API returns standard HTTP status codes:

• 200: Successful request
• 401: Invalid API key
• 500: Internal server error with stack trace

Error responses include a detail message explaining the error.