Overview

cd guides/qwen-agentic-rag
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
# Add your FIRECRAWL_API_KEY to .env (optional for web search)

# Pull a local model (0.8b fits 16GB Macs; larger tags need more RAM)
ollama pull qwen3.5:0.8b

# Build the vector DB (uses ./qdrant_storage by default)
python setup_vectordb.py

# Start the API
python server.py

In another terminal (CLI or web UI):

python client.py --query "What is cross-validation and why is it important?"
python ui.py   # http://127.0.0.1:7860

Project layout

File	Purpose
server.py	LitServe API + CrewAI agents
client.py	Simple HTTP client
ui.py	Gradio web UI
rag_code.py	Embeddings + Qdrant retrieval
tools.py	CrewAI tool for vector search
setup_vectordb.py	One-time knowledge-base setup
TUTORIAL.md	Full step-by-step walkthrough

Hardware notes (16GB RAM)

• Prefer qwen3.5:0.8b or similar small tags (see .env.example)
• Avoid qwen3.6:27b on 16GB machines — it can freeze or crash the system
• First response may be slow while models load

Read the full step-by-step tutorial →{ .md-button .md-button--primary }