top-coder-ai-skills-debugger

Shared debugger library for Top Coder AI Skills. Python library for AI-driven debugging of Node.js (CDP) and Python (DAP) programs — breakpoints, stepping, and variable inspection without a GUI.

Features

• Node.js — CDP over WebSocket (built-in inspector, no extra npm deps).
• Python — DAP via debugpy (launch and attach).
• Single API — DebugSession: start, breakpoints, continue, step, evaluate, stack, variables.
• Two modes — In-memory (e.g. MCP server) or file-backed with a daemon (e.g. CLI scripts that run as separate processes).
• LLM-friendly — All methods return plain text suitable for agent consumption.

Install

pip install top-coder-ai-skills-debugger
# or
uv add top-coder-ai-skills-debugger
# or
poetry add top-coder-ai-skills-debugger

Requirements: Python ≥3.11.

Quick start (in-memory)

import asyncio
from debugger_core.session import DebugSession

async def main():
    session = DebugSession()
    print(await session.start("app.js", "node"))
    print(await session.add_breakpoint("app.js", 10))
    print(await session.resume())       # run until breakpoint
    print(await session.inspect("x"))    # evaluate expression
    print(await session.get_stack())
    print(await session.stop())

asyncio.run(main())

All methods return strings (status messages, stack traces, variable dumps, or error text).

Session modes

For file-backed usage, always call stop() when done so the daemon and debuggee are torn down.

API overview

Development

From the package directory:

# Build wheel + sdist
make build

# Publish to PyPI (after configuring auth)
make publish

# Clean and reinstall from dist (sanity check)
make clean && make test-install

See the repo root and docs/PUBLISHING.md for workspace setup and publishing from the monorepo.

License

MIT. See the top-level repo for full license and source links.