AI Answer Engine

This is a simple answer engine test project leveraging PydanticAI for educational purposes.

⚙️ What is an Answer Engine?

An answer engine is a tool designed to give you direct, detailed answers to your questions. By searching the web and synthesizing information into clear, up-to-date responses. Unlike traditional search engines, which make you sift through a list of links, this system delivers the insights you’re looking for in a single, easy-to-read response.

An example of a great answer engine is PerplexityAI.

🤖 How does it work?

Pydantic AI agent answers questions using the search tool provided to it to search the web using Tavily and infer the answer by LLM models.

🚀 Usage

1. Rename .env.example file to .env in the root directory and set the environment variables.

2. Run the FastAPI application:

   uvicorn app.api:main --reload

3. You can now access the API at http://127.0.0.1:8000/search.

Sample API collections are available in the docs directory. you can view and work with them using Bruno.

🔎 Example

curl -X 'POST' \
  'http://127.0.0.1:8000/search' \
    -H 'accept: application/json' \
    -H 'Content-Type: application/json' \
    -d '{
        "query": "How's the weather like today in Tehran?"
    }'

{
    "answer": "The weather in Tehran today is partly cloudy with a temperature of -2.7°C. The wind is coming from the north-northwest at a speed of 14.0 kph, and the humidity is at 58%. The conditions make it feel like -7.5°C. There is no precipitation, and the visibility is about 10 kilometers."
}

📄 License

This project is licensed under the MIT License.