exoclaw-channel-codemod 0.1.0

Codemod that vendors HKUDS/nanobot channels into exoclaw-channel-* packages

exoclaw-channel-codemod

The codemod that vendors HKUDS/nanobot channels into exoclaw — used as a build-time and test-time dependency by every exoclaw-channel-{slack,telegram,discord,email,matrix,whatsapp} package.

What it does

Given a HKUDS/nanobot channel source file (vendor/upstream.py) and optional ports (vendor/upstream_test.py) snapshotted at a pinned commit (vendor/SHA), plus optional channel-specific patches (patches/00NN-*.patch), it produces:

• exoclaw_channel_<name>/channel.py — the channel module ready to ship
• tests/test_channel.py — the upstream test suite re-targeted at the local module

Transforms are mechanical (import rewrites, signature tweaks, bus capture, monkeypatch string-target rewrites). Patches are applied after the codemod for channel-specific tweaks that can't be generalized.

Public API

from exoclaw_channel_codemod import regenerate

regenerate(pkg_dir)  # one package

Pure-Python and idempotent (only writes when content changed).

Maintainer scripts

• sync.sh <name> [--apply] — fetch latest upstream, diff, optionally regenerate
• scaffold-channel.sh <name> — bootstrap a new channel package by snapshotting upstream

Why this is its own package

• Build isolation: hatchling builds each exoclaw-channel-* wheel in an isolated env. The hatch hook in those packages declares exoclaw-channel-codemod as a build dependency so it's importable inside the sandbox.
• Audit boundary: the codemod is the trusted transform. Versioning it as a real package makes "what was used to derive this generated file" explicit.

Audit boundary

Vendored channel packages commit only vendor/upstream*.py + vendor/SHA + patches/. The generated channel.py and test_channel.py are gitignored — they're derivatives. Anyone reviewing a channel package can verify "this wheel = upstream@SHA + this codemod version + listed patches" by checking out the source and running the test suite (which materializes the codemod output via conftest.py).