This is a small demo project that implements LLM-based chat functionality, augmented with real data (also known as RAG), to give more factually correct answers. The defalut implementation uses data about Domain Driven Design, in the form of text, video transcripts and various articles on the web.
This projects consists of two parts: a Dart based chat backend and a frontend app written in Flutter & Dart.
A simple LLM app that implements RAG functionality, using the framework LangChain.dart.
The backend provides a simple REST api, exposing three endpoints:
- GET /chat/{sessionId} - Gets the chat history
- POST /chat/{sessionId} - Ask a questions
- DELETE //chat/{sessionId} - Clears the conversation
To keep the backend as simple and lightweight as possible it was build using Dart Frog. More complex real-world applications might consider using something like ServerPod.
- Set the
OPENAI_API_KEYenvironment variable. - Run the server locally using
dart_frog dev
See backend/README.md for more details.
A simple Flutter app implementing a basic chat interface, with history. The ResultNotifier package is used for state management and more.
A set of default documents, on the subject of Domain-Driven Design, are included and uses as data for the RAG chain in the backend:
- DDD Reference by Eric Evans - https://www.domainlanguage.com/ddd/reference/
- Domain Driven Design Quickly - https://www.infoq.com/minibooks/domain-driven-design-quickly/
- The InfoQ eMag: Domain-Driven Design in Practice - https://www.infoq.com/minibooks/emag-domain-driven-design/
- Domain-Driven Design in Software Development: A Systematic Literature Review on Implementation, Challenges, and Effectiveness (Ozan Özkan, Önder Babur, Mark van den Brand) - https://arxiv.org/abs/2310.01905