AI Chat Application

This repository contains my AI-powered web app. It’s still under active development, but there is already a working alpha version.

Architecture

• Frontend: Next.js web UI
• Backend: Microservices (Auth Service + Chat Service)
• Gateway: NGINX (API Gateway / Reverse Proxy)
• Data Layer: PostgreSQL (persistent storage)
• LLM Runtime: Ollama (local inference with Llama 3)
• Search Engine: SearXNG API (web search for RAG and LLM tools)
• Infrastructure: Docker (containerized services for local development)

Authentication

Authentication is handled by a dedicated Auth Service.
I use JWT for both access and refresh tokens, signed with RS256 (asymmetric keys) and stored in HttpOnly cookies.
Requests are authorized via the access token, and sessions are renewed using the refresh token.

For LLM responses via Ollama, I use Llama 3 (8.03B parameters).

Roadmap

• Improve UI/UX and overall performance
• Build a stronger answer pipeline with web search + document reading using an Agentic AI approach

Note

Run Llama 3 locally with care — it can be memory-intensive.

Author

Created by Denys Bondarchuk. Feel free to reach out or contribute to the project!