Ragify - Self-Hosted RAG Container

All-in-one container for semantic document search. Index your docs, search with AI embeddings.

Includes: Ollama + nomic-embed-text, Qdrant vector DB, REST API, Web UI, MCP server.

Quick Start

docker pull ghcr.io/strawberry-code/ragify:latest-tika
docker run -d --name ragify -p 8080:8080 -v ragify_data:/data \
  ghcr.io/strawberry-code/ragify:latest-tika

Open http://localhost:8080 - upload files and search.

Image Variants

Image	Size	Description
ragify:latest	~3GB	Text and code files only
ragify:latest-tika	~4GB	Recommended - PDF, DOCX, XLSX, and 1000+ formats via Apache Tika

Production Setup with OAuth

For production, enable GitHub OAuth to restrict access.

1. Create GitHub OAuth App

1. Go to GitHub Developer Settings
2. Click "New OAuth App"
3. Set:
   • Application name: Ragify
   • Homepage URL: https://your-domain.com
   • Authorization callback URL: https://your-domain.com/oauth/github-callback
4. Copy Client ID and Client Secret

2. Configuration Files

docker-compose.yml

services:
  ragify:
    image: ghcr.io/strawberry-code/ragify:latest-tika
    container_name: ragify
    ports:
      - "8080:8080"
    volumes:
      - ragify_data:/data
      - ./users.yaml:/config/users.yaml:ro
    env_file:
      - .env
    restart: unless-stopped

volumes:
  ragify_data:

users.yaml - Authorized GitHub usernames

authorized_users:
  - username: your-github-username
  - username: teammate-username

.env - Environment variables

AUTH_CONFIG=/config/users.yaml
GITHUB_CLIENT_ID=Ov23li...
GITHUB_CLIENT_SECRET=abc123...
BASE_URL=https://your-domain.com

3. Start

docker compose up -d

Environment Variables

Variable	Default	Description
AUTH_CONFIG	-	Path to users.yaml (enables OAuth)
GITHUB_CLIENT_ID	-	GitHub OAuth App Client ID
GITHUB_CLIENT_SECRET	-	GitHub OAuth App Secret
BASE_URL	http://localhost:8080	Public URL for OAuth callbacks
API_PORT	8080	API and Web UI port
OLLAMA_MODEL	nomic-embed-text	Embedding model
CHUNK_SIZE	400	Target chunk size in tokens
CHUNK_MAX_TOKENS	1500	Maximum chunk size (safe margin for nomic-embed-text 2048 limit)
EMBEDDING_BATCH_SIZE	10	Chunks per embedding API call (reduce if Ollama errors)

Features

Web UI

• Upload files via drag & drop
• Create and manage collections
• Search with semantic results
• View indexing job status

REST API

# List collections
curl http://localhost:8080/api/collections

# Search
curl -X POST http://localhost:8080/api/search \
  -H "Content-Type: application/json" \
  -d '{"query": "how to configure", "collection": "docs", "limit": 5}'

# Upload file
curl -X POST http://localhost:8080/api/upload \
  -F "file=@document.pdf" \
  -F "collection=docs"

MCP Server (Claude Desktop / Claude Code)

The container exposes an MCP endpoint for Claude integration.

Claude Code config (~/.claude.json):

{
  "mcpServers": {
    "ragify": {
      "type": "streamable-http",
      "url": "https://your-domain.com/mcp/sse",
      "headers": {
        "Authorization": "Bearer <your-oauth-token>"
      }
    }
  }
}

Processing Pipeline

┌─────────────────────────────────────────────────────────────────────┐
│                        RAGIFY PIPELINE                               │
│                                                                      │
│  ┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────────┐   │
│  │  Upload  │───►│  Tika    │───►│ Chunking │───►│  Embedding   │   │
│  │  (file)  │    │ Extract  │    │ 2-stage  │    │ nomic-embed  │   │
│  └──────────┘    └──────────┘    └──────────┘    └──────┬───────┘   │
│                                                          │          │
│                                                          ▼          │
│  ┌──────────┐    ┌──────────┐                    ┌──────────────┐   │
│  │  Search  │◄───│  Qdrant  │◄───────────────────│    Store     │   │
│  │  Query   │    │  Vector  │                    │   Vectors    │   │
│  └──────────┘    └──────────┘                    └──────────────┘   │
└─────────────────────────────────────────────────────────────────────┘

Pipeline Steps

1. Upload - Files uploaded via Web UI or API
2. Extraction - Apache Tika extracts text from PDF, DOCX, XLSX, etc. (tika variant only)
3. Chunking - Two-stage semantic chunking:
   • Stage 1: Macro chunks with Chonkie (1024 tokens)
   • Stage 2: Fine chunks with Semchunk (512 tokens, 50 overlap)
   • Filter: Validates chunk quality, re-chunks if > 8192 tokens
4. Embedding - Ollama generates 768-dim vectors with nomic-embed-text
5. Storage - Vectors stored in Qdrant with metadata (file hash, URL, title)
6. Deduplication - SHA-256 file hash prevents re-indexing unchanged files

Supported Formats

Base image: .txt, .md, .py, .js, .ts, .java, .go, .rs, .c, .cpp, .json, .yaml, .xml, .html, .css

Tika image (additional): .pdf, .docx, .xlsx, .pptx, .odt, .rtf, .epub, and 1000+ more

Health Check

curl http://localhost:8080/health
# {"status": "healthy", "ollama": "ok", "qdrant": "ok"}

Volumes

Path	Description
/data	Qdrant storage (persist this!)
/config/users.yaml	Authorized users (read-only)
/tmp/collections	Uploaded files (temporary, 15-day retention)

CLI Access (Inside Container)

docker exec -it ragify bash

# Index a directory
python3 ragify.py index /data/docs

# Query
python3 ragify.py query "search term"

# List collections
python3 ragify.py list

Troubleshooting

Container won't start

docker logs ragify

Check for:

• Missing AUTH_CONFIG when OAuth env vars are set
• Invalid users.yaml format

PDF files not processed

Make sure you're using the -tika image variant:

docker pull ghcr.io/strawberry-code/ragify:latest-tika

OAuth callback error

Verify BASE_URL matches your actual domain and GitHub OAuth App callback URL.

License

MIT License - See LICENSE

Contributing

See CONTRIBUTING.md for development setup and guidelines.