Add RAG + Ollama quick reference guide

Introduces docs/RAG_QUICK_REFERENCE.md with setup instructions, architecture overview, troubleshooting, deployment options, and migration notes for the RAG + Ollama AI system. This guide provides concise steps for local/private AI deployment using Xenova embeddings, Supabase pgvector, and Ollama models.

Author: DongDuong2001

docs/RAG_QUICK_REFERENCE.md

@@ -0,0 +1,273 @@
+# 🚀 RAG + Ollama Quick Reference
+
+## ✅ What We Use Now
+
+**Single AI Architecture: RAG + Ollama**
+
+- **Embeddings**: Xenova/all-MiniLM-L6-v2 (384 dimensions)
+- **Vector DB**: Supabase pgvector
+- **AI Model**: Ollama (deepseek-r1:7b or compatible)
+- **Cost**: $0 for AI (infrastructure costs only)
+- **Privacy**: 100% local/private
+
+---
+
+## 🛠️ Quick Setup
+
+### 1. Install Ollama
+
+**Windows:**
+```powershell
+# Download from https://ollama.com
+# Run installer
+```
+
+**macOS:**
+```bash
+brew install ollama
+```
+
+**Linux:**
+```bash
+curl -fsSL https://ollama.com/install.sh | sh
+```
+
+### 2. Pull Model
+
+```bash
+ollama pull deepseek-r1:7b
+```
+
+### 3. Verify
+
+```bash
+ollama list
+# Should show: deepseek-r1:7b
+```
+
+### 4. Run Dev Server
+
+```powershell
+pnpm dev
+# or
+.\start-dev.ps1
+```
+
+---
+
+## 🧠 How RAG Works
+
+```
+User: "How do I add a collaborator?"
+     ↓
+1. Convert question to 384D vector
+     ↓
+2. Search knowledge_base table (pgvector)
+     ↓
+3. Find top 3 similar docs (similarity > 0.7)
+     ↓
+4. Build context: [relevant docs] + user question
+     ↓
+5. Send to Ollama (deepseek-r1:7b)
+     ↓
+6. AI responds with context-aware answer
+```
+
+---
+
+## 📊 Status Indicators
+
+| Indicator | Meaning |
+|-----------|---------|
+| 🟣 Purple pulse + "RAG-Enhanced" | RAG context found & used |
+| 🟢 Green pulse + "Ollama" | Ollama running without RAG |
+| Error message | Ollama not running |
+
+---
+
+## 🔧 Environment Variables
+
+```env
+# Required for AI
+OLLAMA_URL=http://localhost:11434
+OLLAMA_MODEL=deepseek-r1:7b
+
+# Required for RAG
+NEXT_PUBLIC_SUPABASE_URL=your-url
+NEXT_PUBLIC_SUPABASE_ANON_KEY=your-key
+```
+
+---
+
+## 🗄️ Database Setup
+
+```sql
+-- Enable pgvector
+CREATE EXTENSION IF NOT EXISTS vector;
+
+-- Create knowledge base table
+CREATE TABLE knowledge_base (
+  id UUID PRIMARY KEY DEFAULT uuid_generate_v4(),
+  content TEXT NOT NULL,
+  embedding vector(384),
+  metadata JSONB,
+  category TEXT,
+  source TEXT,
+  title TEXT,
+  created_at TIMESTAMP DEFAULT NOW()
+);
+
+-- Add vector similarity index
+CREATE INDEX ON knowledge_base 
+USING ivfflat (embedding vector_cosine_ops);
+```
+
+---
+
+## 📚 Index Documentation
+
+```bash
+# Index all docs for RAG
+node scripts/index-knowledge.js
+
+# This creates embeddings for:
+# - README.md
+# - docs/*.md
+# - Code comments
+# - Feature descriptions
+```
+
+---
+
+## 🐛 Troubleshooting
+
+### "AI not responding"
+
+```bash
+# 1. Check if Ollama is running
+ollama list
+
+# 2. Verify model is downloaded
+ollama pull deepseek-r1:7b
+
+# 3. Test Ollama directly
+ollama run deepseek-r1:7b
+# Type: "Hello"
+# Should respond
+# Type: /bye to exit
+
+# 4. Check environment variables
+echo $OLLAMA_URL  # Should be http://localhost:11434
+```
+
+### "RAG not working"
+
+1. **Check Supabase connection**
+   - Verify `NEXT_PUBLIC_SUPABASE_URL`
+   - Verify `NEXT_PUBLIC_SUPABASE_ANON_KEY`
+
+2. **Check knowledge_base table**
+   ```sql
+   SELECT COUNT(*) FROM knowledge_base;
+   -- Should have documents indexed
+   ```
+
+3. **Re-index documentation**
+   ```bash
+   node scripts/index-knowledge.js
+   ```
+
+### "Slow responses"
+
+- Use smaller model: `ollama pull llama3.2` (3B instead of 7B)
+- Close other apps to free RAM
+- Consider GPU for faster inference
+
+---
+
+## 📈 Production Deployment
+
+### Option A: Vercel + VPS
+
+1. **Deploy Next.js to Vercel**
+   ```bash
+   vercel --prod
+   ```
+
+2. **Run Ollama on VPS** (DigitalOcean, AWS, etc.)
+   ```bash
+   # On VPS
+   curl -fsSL https://ollama.com/install.sh | sh
+   ollama pull deepseek-r1:7b
+   
+   # Set up reverse proxy (nginx)
+   # Configure SSL
+   ```
+
+3. **Update Vercel env vars**
+   ```env
+   OLLAMA_URL=https://ai.yourdomain.com
+   ```
+
+### Option B: All Local (Development)
+
+```env
+OLLAMA_URL=http://localhost:11434
+OLLAMA_MODEL=deepseek-r1:7b
+```
+
+---
+
+## 📁 Key Files
+
+| File | Purpose |
+|------|---------|
+| `app/api/chat/route.ts` | RAG-enhanced chat API |
+| `lib/services/rag-service.ts` | RAG embedding & retrieval |
+| `app/dashboard/ai-tools/page.tsx` | Chat UI |
+| `scripts/index-knowledge.js` | Knowledge indexer |
+| `docs/RAG_SYSTEM.md` | Full RAG documentation |
+| `docs/OLLAMA_SETUP.md` | Ollama setup guide |
+
+---
+
+## ✅ Benefits Summary
+
+| Aspect | Benefit |
+|--------|---------|
+| **Privacy** | 100% local, no external API calls |
+| **Cost** | $0 for AI (only infrastructure) |
+| **Accuracy** | Context from YOUR docs, not generic |
+| **Offline** | Works without internet |
+| **Unlimited** | No rate limits or quotas |
+| **Fast** | Vector search <100ms |
+
+---
+
+## 🎯 What Changed (Migration)
+
+### ❌ Removed
+- DeepSeek API integration
+- Google Gemini API integration
+- Multi-provider fallback logic
+- Cloud API environment variables
+
+### ✅ Added
+- RAG service with Xenova embeddings
+- Supabase pgvector integration
+- Knowledge base indexer
+- Context-aware AI responses
+
+---
+
+## 📚 Learn More
+
+- [RAG System Documentation](./RAG_SYSTEM.md)
+- [Ollama Setup Guide](./OLLAMA_SETUP.md)
+- [Production Deployment](./PRODUCTION_DEPLOYMENT_OLLAMA.md)
+- [Migration Summary](./RAG_MIGRATION_SUMMARY.md)
+
+---
+
+**Last Updated:** January 17, 2026  
+**Architecture:** RAG + Ollama (100% Private)