pypto-tooling

Personal tooling umbrella: agent skills, runbooks, and glue for PyPTO / simpler development on Ascend 910B. Most of the heavy machinery that used to live in this repo moved to dedicated sibling repos — see below.

Repo map

This repo is now a thin umbrella. The three former sub-projects are standalone repos, siblings under hw-native-sys/:

Repo	Contents
../pypto-docker	Dockerfiles (CANN + sim images for pypto / simpler / pypto-lib / pytorch-hccl-tests), image build scripts, docker-entrypoint-cann.sh, and the full image build/runtime README
../mcp-hw-native-sys	MCP server (pip package) operating over the hw-native-sys workspace — repos, tasks, knowledge index, verify ladder
../pypto-profiling	Personal collective benchmark harness (pypto vs simpler vs HCCL L3 collectives)

What stays here

• debugging_skills/ — NPU / docker debugging skills + error-code reference
• task-submit/ — reference doc for the task-submit device-locking CLI
• diagnose_npu.py — 10-point NPU hardware health check (copy into a container and run)
• bz910b-reproduce.md — Ascend 910B server image repro runbook
• personal_setup.md — personal git / env setup notes
• .mcp.json — Claude Code MCP registration for mcp-hw-native-sys

The docker-specific skills moved to pypto-docker:

• pypto-docker/build_skills/ — in-container PyPTO (pypto_core) rebuild workflow
• pypto-docker/dockerfile_skills/ — Dockerfile construction skill + incident logs

MCP registration

.mcp.json registers the hw-native-sys MCP server from the sibling mcp-hw-native-sys repo. Any Claude Code session started in pypto-tooling (or a parent) picks it up; tools appear as mcp__hw-native-sys__<tool_name>.

Sim Dev Iteration Workflow (local code changes)

The full sim image workflow (build, mount, pip install -e, targeted vs full test suites) moved with the images to pypto-docker/README.md. The short form:

# One-time: build the sim image (~15-30 min)
docker build -t pypto3-hw-native-sys:sim -f Dockerfile.hw-native-sys.sim.ubuntu22.04 .

# Every code change: mount workspace + pip install -e (~2-5 min)
docker run --rm \
  -v /path/to/hw-native-sys/pypto:/opt/pypto \
  pypto3-hw-native-sys:sim \
  bash -c "pip install --no-build-isolation -e '/opt/pypto[dev]' 2>&1 | tail -1 && \
           pytest tests/ut -n auto --maxprocesses 8 -v"

Key rules:

• Never rebuild the image for code changes — -v mount + pip install -e is the iteration loop.
• Always run ruff inside Docker — the host .ruff_cache gets root-owned from previous Docker runs.
• Test targeted suites first, then full suite before pushing.