101-02 — Run a governed task

What you will build

A one-shot script that runs pw_chunk_classify via TaskRunner using fake_chat—no network.

Prerequisites

• 101-01 completed (venv + install).

Files you will touch

None required; optional comparison with examples/basic/run_fake_task.py.

Step 1 — Same snippet as maintainers use

From repo root, venv active:

PYTHONPATH=src python3 - <<'PY'
import json

from forge_lcdl.env import LlmEnvProfile
from forge_lcdl.runner import TaskRunner
from forge_lcdl.types import ChatResult

payload = {
    "chunk_results": [
        {
            "chunk_id": "c1",
            "is_question_block": True,
            "confidence": 0.9,
            "reason": "toy fake response",
        }
    ]
}

def fake_chat(messages, **kwargs):
    return ChatResult(True, json.dumps(payload))

profile = LlmEnvProfile(
    kind="certificator",
    base_url="http://localhost",
    api_key="",
    model="fake",
    timeout_sec=30,
    ngrok_bypass=False,
    prefer_json_object_mode=True,
)

runner = TaskRunner(chat=fake_chat)
r = runner.run(
    "pw_chunk_classify",
    "v1",
    {
        "url": "https://example.com",
        "chunks": [{"chunk_id": "c1", "text_snippet": "Sample?"}],
    },
    profile=profile,
)
print(r)
PY

Step 2 — Or run the packaged example

PYTHONPATH=src python3 examples/basic/run_fake_task.py

Expected output

• Printed line contains Ok( and a chunks key inside value (the task merges chunk_results from the model-shaped JSON onto input chunks). See basic task runner example.

Common failures

Verify

PYTHONPATH=src python3 examples/basic/run_fake_task.py | grep -q "chunks" && echo OK

What changed

You understand injectable chat, ChatResult, and TaskRunner.run boundaries.

Concepts

• TaskRunner — holds transport for tests or production adapters.
• Contract path — src/forge_lcdl/contracts/pw_chunk_classify/v1/ (next lesson).

Next step

101-03 — Read a contract.