Demo CRM agent wired for Veris simulations.
Prerequisites: Python ≥3.11, OpenAI key, Veris key, ngrok (authtoken configured)
git clone https://github.com/veris-ai/mini_crm.git
cd mini_crmuv sync && source .venv/bin/activateCopy .env.example to .env and fill in the required values:
cp .env.example .envEdit .env with your actual values:
OPENAI_API_KEY=<your-openai-key>
VERIS_API_KEY=<your-veris-api-key>
⚠️ Make sure ngrok is configured on your CLI!You must have an ngrok account and your authtoken set up locally.
To verify, run:ngrok config checkTo get started you may follow the ngrok getting started guide (steps 1 and 2 are sufficient).
veris setup-local start --local-url http://localhost:8000The setup command launches the FastAPI server and makes it publicly accessible via ngrok. The simulator command watches and runs scenarios.
Open a new terminal, then:
cd mini_crm
source .venv/bin/activateCreate a new agent card:
veris sim create --agent-path agent_card.json --agent-version v1.0.0Launch the simulation:
veris sim v3launch --watch Watch the simulation run, you can check the status by going to the correspoding .veris/runs/RUN_ID file prompted in the terminal.
Give it 3-4 minutes to run, if scenarios are not yet finished, you can kill the simulation by running:
veris sim kill <simulation_id>To learn more explore the veris-ai/veris-cli docs.
uvicorn app.main:app --reload
curl http://localhost:8000/health
curl -X POST http://localhost:8000/chat \
-H "Content-Type: application/json" \
-d '{"message": "Qualify Acme Capital."}'Direct runs mutate db/leads.json; simulator runs capture the same tool calls without touching disk.
app/agent.py– agent definition and toolsapp/main.py– FastAPI surface instrumented by Verisdb/leads.json– sample TinyDB store.veris/– generated simulator config afterveris init
Need more depth? See TROUBLESHOOTING.md