development.md

Working on petsitter

Running the tests, and driving petsitter from other code.

← back to the README

---

Running from a checkout

./petsitter          # the server, using src/ directly; no install needed

Dependencies are in pyproject.toml: httpx, starlette, uvicorn, click, and detect-secrets (used by Secrets Protector).

Running Tests

# Activate virtual environment
source .venv/bin/activate

# Install test dependencies
pip install -e ".[test]"

# Run tests
pytest tests/ --ignore=tests/test_playground_e2e.py --ignore=tests/test_registry_e2e.py

pyproject.toml sets pythonpath = ["src"], so the unit tests run against the checkout.

test_proxy.py::TestProxyHandler::test_config_magic_returns_diag and test_use_route.py::TestUsePathEndpoint::test_use_route_config_magic_streams_diag currently fail.

The two end-to-end suites are scripts, not pytest tests (pytest can't collect them), and need more set up:

# boots a server against a stub upstream and drives it in Chromium;
# needs petsitter installed (pip install -e .) and Playwright
pip install playwright && playwright install chromium
python tests/test_playground_e2e.py

# index parsing, checksums, pkg: loading; needs the index repo's crawl.py
git clone https://github.com/day50-dev/tricks tricks-index
python tests/test_registry_e2e.py

Example: Using with an Agentic Framework

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8080/v1",
    api_key="not-needed"
)

response = client.chat.completions.create(
    model="any-model-name",
    messages=[{"role": "user", "content": "List files in /tmp"}],
    tools=[{"type": "function", "function": {"name": "get_weather", "parameters": ...}}]
)

Live updates

The dashboard gets everything live over one server-sent event stream:

curl -N http://localhost:8080/api/stream
# data: {"sid": "..."}   then {"topic": ..., "data": ...} per message

Subscribe to topics with POST /api/stream/<sid> and {"subscribe": ["logs", "pause", "live:<extension id>"]}. Each topic sends its backlog first.

Every request gets an ID when it arrives; it tags its log lines, and tricks read it as self.request_id.