AstraAI is a small AI coding assistant I built to learn how large language models can help generate and understand code.
It’s a basic full-stack project where you can type a prompt like
“create a FastAPI route for user login”
and it generates the code with explanations.
The project has a FastAPI backend, a React frontend, and an AI layer that connects to language models.
It also uses a vector database to remember context between messages so conversations feel continuous.
Everything runs locally and can be deployed using Docker.
- Generate backend or frontend code from plain text
- Explain existing code in simple terms
- Store and recall previous prompts using embeddings
- Works with both online and local models
- Runs fully offline if needed
- Simple chat interface for interaction
- Frontend: React, TypeScript
- Backend: FastAPI, Python
- Database: PostgreSQL
- AI Layer: LangChain, FAISS
- Other Tools: Redis, Docker, WebSockets
- Clone the repo
git clone https://github.com/GojoSnippet/astraai cd astraai
Start the backend
cd backend pip install -r requirements.txt uvicorn app.main:app --reload
Start the frontend
cd ../frontend npm install npm start
What I Learned
How language models handle context and prompts
How to connect FastAPI with a React frontend
How vector databases improve response relevance
How to containerize and run everything locally with Docker
About
I’m Sai Kumar Domakonda, a student who likes working on AI-based tools, backend systems, and small developer utilities. This project was mainly built to practice integrating LLMs with real applications and learn end-to-end system design