Alice is a tool to create educational chatbots grounded in knowledge bases assembled from course materials: uploaded documents (PDF, DOCX, PPTX, H5P, SCORM, images, plain text) and content synced from Moodle.
Educators log in via OIDC, create a chatbot, pick a persona, and attach a knowledge base by uploading files or syncing a Moodle course. They then share the chatbot via a URL (public, password-protected, or SSO-only).
Learners open the URL and chat. Each answer is grounded in the knowledge base and returned with citations to the source documents.
Educators manage all their chatbots from a single dashboard.
Learners chat with the bot and get answers grounded in the course materials, with citations.
Four services:
frontend/: Angular UI: dashboard for managing chatbots and knowledge bases, plus the learner-facing chat interface.api/: FastAPI backend. Owns chatbots, knowledge bases, users, jobs, and Moodle integrations. Persists to PostgreSQL, stores uploads in S3 (Garage), publishes ingestion jobs to RabbitMQ.services/rag-pipeline/: Haystack-based AI service. Parses documents via a Docling Serve instance, embeds and indexes chunks, and answers queries against an OpenAI-compatible LLM endpoint.sync/: Background worker. Consumes RabbitMQ jobs, pulls files from S3 or Moodle, and drives the RAG pipeline ingestion.
- Docker Compose for development and production
- Angular frontend
- FastAPI for the Python API
- Alembic for database migrations
- SQLModel for the Python ORM
- UV for Python package management
- PostgreSQL as the SQL database
- Garage for S3 storage (currently included in Docker Compose)
- Qdrant as the vector store
- Valkey for ingestion job tracking and RAG session storage
- RabbitMQ for message queueing
- FastStream for the Python client integrated with FastAPI
- Pika for the Python client
- Haystack for RAG pipeline management
- Docling Serve for file parsing
- An OpenAI compatible endpoint for LLM/Embeddings inference
Documentation is available Here or locally in the docs/ folder.
This project is licensed under the GNU AFFERO GENERAL PUBLIC LICENSE Version 3, 19 November 2007, you may obtain the source code on Github. Please check the CREDITS.md file at the root of the repository for more details about the licences and contributors.
Created by LMDDC.