docker/genai-stack

The GenAI Stack is a project designed to help users build their own GenAI application quickly. It includes demo applications that can serve as inspiration or a starting point. The project requires configuration through a .env file, which includes variables for the Ollama LLM API URL, Neo4j database URL and credentials, LLM model tag, embedding model, and various API keys for different services. Users can use any LLM available via Ollama on MacOS and Linux, or use GPT-3.5 and GPT-4 on all platforms with their own OpenAI API keys.

To get started, users can create a .env file from the env.example template and configure the variables accordingly. The project includes applications such as a Support Bot, Stack Overflow Loader, PDF Reader, Standalone Bot API, and Standalone Bot UI. The Support Bot is a fullstack Python application that answers support questions based on recent entries, provides summarized answers with sources, and demonstrates the difference between RAG (Reasoning and Generation) disabled and enabled modes. The Stack Overflow Loader imports recent Stack Overflow data into a knowledge graph, embeds questions and answers, and stores them in a vector index. The PDF Reader reads local PDFs and asks questions about their contents using vector similarity search. The Standalone Bot API is a standalone HTTP API with streaming (SSE) and non-streaming endpoints that exposes the functionality to answer questions in the same way as the Support Bot. Lastly, the Static front-end is a separate application built using modern best practices (Vite, Svelte, Tailwind) that has the same features as the Support Bot but with instant auto-reload on changes. The project also includes a database that can be explored at http://localhost:7474.