🦉 Caddy

Caddy is a Retrieval-Augmented Generation (RAG) platform combining:

• model: A FastAPI-based LLM service backed by OpenSearch
• scraper: A semi-redundant scraper for uploading large sites in batches
• frontend: A collection, resource and user management UI

📦 Components

Folder	Description
model	Qdrant integration and document ingestion
scraper	Scraping, summarization, doc parsing
frontend	Resource, collection and user management UI

🚀 Getting Started

With Docker:

cp .env.example .env

And populate the .env file with the necessary vars. Then run this to bring up the backing services:

make run

Following the backing services coming up (qdrant, postgres and minio), run the model and frontend in the way that you see fit, either through your IDE of via the make commands:

make run_frontend
make run_backend

Important note: you should allocate at least 4GB for Docker or performance will be horrible and qdrant may struggle.

Working with dependencies

If you change a dependency in pyproject.toml you will need to rebuild the app if you are running the app in docker.

Do this using

docker compose build model --no-cache

To drop data volumes, which you may want to do if you change the database init script in ./model/scripts/postgres-init.sql or some qdrant settings, you'll need to drop the volumes. docker volume rm will not work for volumes created by docker compose, so use the following command:

# drop all volumes including Qdrant, postgres etc
docker compose down --volumes

📂 Features

• Semantic and lexical document search
• Qdrant-backed vector store
• PDF/DOCX/web content ingestion
• Authenticated scraping support
• AWS/Google/Azure storage, local directory support
• FastAPI backend + Astro frontend
• Data Ingestion & Collection Management

🧪 Testing

make run_tests

🧠 Architecture

• FastAPI (REST API)
• Qdrant (vector DB)
• PostgreSQL (metadata)
• Minio/S3 (document storage)
• Astro (admin UI)

🧾 License

MIT