Custom RAG to recommend recipes based on user and environmental constraints.
Chat-Gptouille is a full-stack proof-of-concept chatbot that replicates core ChatGPT functionality using locally hosted language models via Ollama. Built with a Flutter front end, a FastAPI backend, and a PostgreSQL database, the entire system runs seamlessly through Docker Compose. The project illustrates how modern development tools and local AI models can be combined into a cohesive, privacy-respecting conversational AI experience—entirely free of cloud dependencies.
Local-first architecture using Ollama to serve LLMs like qwen2.5 and all-minilm
Mobile-friendly interface built in Flutter for cross-platform deployment
Real-time chat system with persistent user accounts and message history
Fully containerized setup for reproducibility and ease of deployment
End-to-end demo and installation videos to guide through setup and usage
Frontend: Flutter (Dart) mobile application
Backend: FastAPI (Python) REST API
Database: PostgreSQL with SQLAlchemy ORM
Containerization: Docker Compose for orchestration
LLM Serving: Ollama for local model inference
Admin UI: Adminer for database management
The API layer handles all application logic:
RESTful endpoints for message handling (/chat/send) and embedding management (/chat/load)
Request/response validation with Pydantic
Integration with Ollama to query local language models
A containerized PostgreSQL instance stores:
User credentials and profiles
Chat session metadata
Message history
SQLAlchemy models map the data structure
Adminer provides a visual web interface for data inspection
The mobile app delivers an intuitive chat interface:
User authentication and session management
Real-time messaging with UI updates
Clean, responsive design built with Flutter/Dart
Communicates with backend via HTTP client
Docker Compose unifies services into a single deployment command
Health checks, automatic restarts, and persistent volumes ensure stability
Environment files (.env) make configuration modular and secure
Chat-Gptouille offers a compelling example of how to combine modern tools and local LLMs to build a real-time, privacy-first chatbot experience. It's an ideal foundation for developers aiming to create custom conversational agents without relying on external APIs or cloud-based models.
