yogendrachukka01/adaptiveagent

AdaptiveAgent is a starter kit for building an agentic RAG system, which means an AI assistant that answers questions using your own private documents instead of just general knowledge. The project describes itself as production grade, secure, and explainable, and it is built around a reasoning pipeline orchestrated with LangGraph, paired with a Next.js frontend, support for multiple vector stores, evaluation using an LLM as a judge, and automatic fallback to a cloud model when needed. The whole stack is meant to run locally through Docker. A single docker compose command sets up a Postgres database, Redis, and Ollama for local model inference, and after about two minutes for models to download you get a frontend at one local address, an API documentation page at another, and a health check endpoint. There is also a local development path with hot reload, where the backend runs through FastAPI with uvicorn and the frontend runs through the standard Next.js dev server. The README gives detailed setup instructions for Windows, macOS, and Linux separately, each with an automated setup script and a manual step by step option. On Windows this includes PowerShell commands and winget installs, on macOS it covers Homebrew and Apple Silicon support with Ollama running natively, and on Linux it covers apt packages and adding your user to the docker group so Docker can run without sudo. Prerequisites listed are Python 3.11 or newer, Node.js 20 or newer, Docker 24 or newer with Compose v2, and Git 2.30 or newer. A verification script checks that all these tools and the project's own dependencies are installed correctly. For day to day development, a Makefile provides shortcuts for tasks like starting the backend or frontend, running tests, linting, formatting code, and managing the Docker stack, alongside the equivalent manual commands using tools like ruff, pytest, mypy, and the TypeScript compiler. Badges at the top of the README show the project uses Python 3.11 plus, Next.js 15, and is under the MIT license, a permissive license that lets you reuse and modify the code freely.