The demo uses a small, intentionally incomplete Python project. Copy it to a temporary directory so Prime Agent can edit it without changing this checkout.
From the repository root:
cp -R examples/demo_project /tmp/prime-agent-python-demo
uv run --no-project --with prime-agent-python-client examples/stream_prompt.py \
--repository /tmp/prime-agent-python-demo \
--provider github-copilot \
--model gpt-4.1 \
"Implement summarize_orders in order_summary.py and run the unit tests"The script streams assistant text and prints the final Prime Agent session statistics. Prime Agent credentials and model access must already be configured.
Use a fresh copy so the second run starts from the same source:
cp -R examples/demo_project /tmp/prime-agent-superqode-demo
cd /tmp/prime-agent-superqode-demo
uvx --from superqode==0.2.82 superqode harness doctor \
--spec prime-agent.yaml \
--json
uvx --from superqode==0.2.82 superqode harness run \
--spec prime-agent.yaml \
--prompt "Implement summarize_orders in order_summary.py and run the unit tests" \
--provider github-copilot \
--model gpt-4.1 \
--stream
python -m unittest -vThe demo uses only the Python standard library. The initial unit-test failure is intentional; the requested implementation makes it pass.