Skip to content

AGENTS.md

Latest commit

 

History

History
43 lines (35 loc) · 1.66 KB

AGENTS.md

File metadata and controls

43 lines (35 loc) · 1.66 KB

AGENTS.md — Data Fabrication

Project Purpose

Data Fabrication is a Python Platform SDK challenge service. Miners submit ZIP packages containing the full Python harness code; the service executes the harness, evaluates the generated agentic coding dataset, stores results in SQLite, and exposes Platform-compatible weights.

Architecture

data-fabrication/
├── src/data_fabrication/
│   ├── app.py                 # FastAPI app entrypoint
│   ├── config.py              # Runtime settings
│   ├── db.py                  # SQLite wrapper
│   ├── repository.py          # Persistence and leaderboard queries
│   ├── routes.py              # Public Platform proxy routes
│   ├── weights.py             # Weight computation
│   ├── evaluator/             # ZIP artifacts, JSONL parsing, scoring, AST checks, execution
│   └── sdk/                   # Vendored Platform-compatible helpers
├── tests/
├── pyproject.toml
└── Dockerfile

Commands

python -m pip install -e ".[dev]"
ruff check src tests
ruff format --check src tests
mypy --config-file pyproject.toml src
pytest tests
python -m compileall src

Guidelines

  • Keep the runtime Python 3.12+ compatible.
  • Prefer Platform-compatible routes: /health, /version, /internal/v1/get_weights.
  • Use the vendored Docker executor for untrusted harness execution in production.
  • /submit must accept ZIP harness packages only, never direct datasets or loose code.
  • Direct subprocess execution is for local development/tests only and must remain opt-in.
  • Never log shared tokens, miner secrets, or raw bearer headers.