Llama Stack Workshop Materials

Starting the model server

For localhost development, use Ollama or you can use a remote model server such as vLLM via Model-as-a-Service solution MaaS s

ollama serve

Pull down your needed models. For LLM tool invocations you ofte need a larger model such as Qwen 14B. The way you know is to test your app/agent + model + model-server-configuration.

ollama pull llama3.2:3b
ollama pull qwen3:14b-q8_0

cd llama-stack-scripts

export LLAMA_STACK_BASE_URL=http://localhost:8321
export INFERENCE_MODEL=ollama/llama3.2:3b
# export INFERENCE_MODEL=vllm/llama-4-scout-17b-16e-w4a16
export LLAMA_STACK_LOG_FILE=logs/llama-stack-server.log
export LLAMA_STACK_LOGGING="tools=DEBUG,tool_runtime=DEBUG,providers=DEBUG,server=info"

If using Ollama

export OLLAMA_URL=http://localhost:11434 

If using MaaS

export VLLM_API_TOKEN=blah

export VLLM_URL=https://llama-4-scout-17b-16e-w4a16-maas-apicast-production.apps.prod.rhoai.rh-aiservices-bu.com:443/v1

python3.12 -m venv .venv
source .venv/bin/activate

uv run python -V

Install Dependencies

uv run --with llama-stack llama stack list-deps starter | xargs -L1 uv pip install

Run the Llama Stack server attaching itself to ollama

uv run --with llama-stack llama stack run starter

Inspect the server by running the scripts in llama-stack-scripts

Note: Llama Stack persists state to ~/.llama/ specifically ~/.llama/distributions/starter when using the run starter command above. If you want a clean start:

rm -rf ~/.llama/distributions/starter

Start Customer Backend

Assumes Postgres is up and running with appropriate database pre-created. See the deeper dive README.MD

cd fantaco-customer-main
open README.md

Run the Customer REST API

java -jar target/fantaco-customer-main-1.0.0.jar

Quick test of Customer REST API

curl -sS -L "$CUST_URL/api/customers?companyName=Around" | jq

Start Finance Backend

cd fantaco-finance-main
open README.md

Run the Finance REST API

java -jar target/fantaco-finance-main-1.0.0.jar

Quick test for Finance REST API

curl -sS -X POST $FIN_URL/api/finance/orders/history \
  -H "Content-Type: application/json" \
  -d '{
    "customerId": "AROUT",
    "limit": 10
  }' | jq

Customer MCP

cd fantaco-mcp-servers/customer-mcp
source .venv/bin/activate

python customer-api-mcp-server.py

Finance MCP

cd fantaco-mcp-servers/finance-mcp
source .venv/bin/activate

python finance-api-mcp-server.py

Using mcp-inspector to test the MCP Servers

Agents with Llama Stack

The agents-llama-stack/ directory contains numbered example scripts that demonstrate progressive agent capabilities. Run these scripts in order 1 through 6 to learn the concepts step-by-step:

1. 1_hello_world_agent_no_stream.py - Basic agent without streaming
2. 1_hello_world_agent_streaming.py - Basic agent with streaming
3. 3_list_customer_tools.py / 3_list_finance_tools.py - List available MCP tools
4. 4_agent_customer_mcp.py / 4_agent_finance_mcp.py - Single-domain agents
5. 5_agent_customer_and_finance.py - Multi-domain agent
6. 6_multi_turn_agent.py - Multi-turn conversational agent

cd agents-llama-stack
source .venv/bin/activate
python 1_hello_world_agent_no_stream.py
# Continue with 2, 3, 4, 5, 6...

Simple Agent LangGraph

Follow the REAME.md in simple-agent-langgraph