Skip to content

QUICKSTART.md

Latest commit

 

History

History
228 lines (158 loc) · 4.22 KB

QUICKSTART.md

File metadata and controls

228 lines (158 loc) · 4.22 KB

🚀 FlowTech-AI - Quick Start Guide

Start the complete stack in 5 minutes!

⚡ Quick installation

# 1. Clone the repo
git clone https://github.com/FlowTech-Lab/FlowTech-AI.git
cd FlowTech-AI

# 2. Configure (optional if default values OK)
cp .env.example .env
nano .env  # Adjust if necessary

# 3. Launch everything! (requires sudo for permissions)
sudo ./init.sh

# ✅ That's it! Stack ready in 5 minutes

🎯 Available services

After sudo ./init.sh, you have access to:

Service URL Description
OpenWebUI http://localhost:8081 Conversational AI interface
Cursor MCP http://localhost:8000 Cursor code context
Cursor Knowledge http://localhost:8001 Cursor notes search
n8n http://localhost:5678 Automation workflows
Qdrant http://localhost:6333 Vector database
Samba Share \\YOUR_IP\notes Network share for notes

Credentials: Displayed at the end of sudo ./init.sh

Notes Sync Setup

After initialization, you can sync your Markdown notes to make them searchable in Cursor:

# Initial sync
./scripts/sync-notes.sh

# Install hourly auto-sync (optional)
./scripts/install-cron.sh

📖 Full guide: Notes Sync Documentation

🔧 Cursor Configuration

1. Copy MCP config

# Linux/Mac
cp cursor-mcp-config.json ~/.cursor/mcp.json

# Windows
copy cursor-mcp-config.json %USERPROFILE%\.cursor\mcp.json

2. Edit the IP

{
  "mcpServers": {
    "qdrant": {
      "url": "http://YOUR_IP:8000/sse"  // ← Change the IP
    }
  }
}

3. Restart Cursor

4. Test

In Cursor:

@qdrant store "Test MCP connection"
@qdrant find test connection

✅ If it works, you're ready!

📚 Using OpenWebUI RAG

1. Open OpenWebUI

Navigate to http://localhost:8081

2. Create a Chat

Click on "New Chat"

3. Enable Knowledge

Click the "Knowledge" button in the chat interface

4. Upload Documents

  • Click "Upload Files"
  • Select your documents (PDF, DOCX, MD, TXT, etc.)
  • Wait for indexing (automatic)

5. Ask Questions

User: What does the technical documentation say about database setup?
AI: Based on the uploaded documentation, the database setup involves...

Supported Formats

  • 📄 Documents: PDF, DOCX, TXT, MD
  • 💻 Code: PY, JS, TS, JAVA, etc.
  • 🌐 Web: Paste URLs for website indexing

🎯 Use cases

Developer with Cursor

1. Code in Cursor
2. @qdrant store to save snippets
3. @qdrant find to retrieve context
4. RAG enriched with your technical notes

Technical documentation Q&A

1. Upload docs to OpenWebUI
2. Ask questions in natural language
3. Get accurate answers with citations
4. Share knowledge with your team

Team (fork)

1. Clone the repo
2. sudo ./init.sh
3. Configure Cursor
4. Everyone shares the same knowledge base

📊 Verification

Stack operational?

docker compose ps
# ✅ All services "Up (healthy)"

MCP-Qdrant working?

curl http://localhost:8000/sse
# ✅ Returns event stream

Qdrant collections?

curl http://localhost:6333/collections
# ✅ See "cursor-context"

🆘 Troubleshooting

Services not starting

# Check logs
docker compose logs

# Restart cleanly
docker compose down
sudo ./init.sh

Cursor not connecting

# Check IP in ~/.cursor/mcp.json
# Check firewall
sudo ufw allow 8000

# Test endpoint
curl http://YOUR_IP:8000/sse

RAG not finding documents

# Check OpenWebUI logs
docker compose logs openwebui

# Verify Qdrant collection
curl http://localhost:6333/collections

# Re-upload documents in OpenWebUI

📚 Complete documentation

  • Setup: README.md - Detailed installation
  • Architecture: See README.md - Technical overview
  • Services: docker-compose.yml - Service configuration
  • Templates: Notes/_Templates/ - Obsidian templates

🎉 That's it!

You now have:

  • ✅ Complete operational AI stack
  • ✅ Cursor integration (MCP-Qdrant)
  • ✅ OpenWebUI with RAG
  • ✅ Vector database ready
  • ✅ Automation platform (n8n)

Total time: ~5 minutes ⚡

Next steps: See README.md to go further!