Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
149 lines (113 loc) · 3.84 KB

CONTRIBUTING.md

File metadata and controls

149 lines (113 loc) · 3.84 KB

Contributing to NeMoCode

Thanks for your interest in contributing to NeMoCode. This guide covers how to set up a development environment, make changes, and submit pull requests.

Prerequisites

  • Python 3.11 or later
  • uv or pip for package management
  • An NVIDIA API key from build.nvidia.com (for live testing)

Setup

git clone https://github.com/Hmbown/NeMoCode.git
cd NeMoCode
pip install -e ".[dev]"

Set your API key:

export NVIDIA_API_KEY="nvapi-..."

Or use the built-in setup wizard:

nemo setup

Development Workflow

  1. Create a branch from main
  2. Make your changes
  3. Run checks (see below)
  4. Open a pull request

Running Checks

ruff check src/ tests/       # Lint
ruff format src/ tests/       # Format
pytest tests/ -q              # Tests

Pre-commit Hooks

Install pre-commit hooks to run checks automatically:

pip install pre-commit
pre-commit install

Code Style

  • Formatter: ruff (configured in pyproject.toml)
  • Lint rules: E, F, I, W, UP, B
  • Line length: 100
  • Imports: sorted isort-style (ruff handles this)
  • Type hints: use modern X | Y syntax, not Union[X, Y]
  • Docstrings: Google style on all public functions and classes
  • SPDX headers: all .py files must start with: 
    # SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES.
# SPDX-License-Identifier: MIT

Testing

  • Tests live in tests/ mirroring the src/nemocode/ structure
  • Use pytest-asyncio for async tests (auto mode is configured)
  • Mock all external API calls (httpx) — no real network calls in tests
  • Keep tests fast (< 30s total)

Project Structure

src/nemocode/
  cli/         # CLI commands, TUI, rendering
  config/      # Configuration schema and loading
  core/        # Core engine (scheduler, sessions, context)
  providers/   # NIM API providers (chat, embeddings, rerank)
  skills/      # Built-in skills (commit, review)
  tools/       # Tool registry and implementations
  workflows/   # High-level agent orchestration

Adding a New Tool

  1. Create a function in src/nemocode/tools/ with the @tool decorator
  2. Add parameters with type hints (JSON Schema is auto-generated)
  3. Return a JSON string with results or error
  4. Register it in tools/__init__.py if not auto-discovered
from nemocode.tools import tool

@tool(description="One-line description shown to the model")
def my_tool(arg: str, count: int = 1) -> str:
    """My tool does something useful.
    arg: The input string to process
    count: Number of times to process
    """
    return json.dumps({"result": arg * count})

Adding a New Agent

Create a markdown file in your project or config directory:

---
name: my-agent
role: executor
tools: [read_file, edit_file, bash_exec]
---

You are a specialized agent for...
(custom system prompt here)

Commit Messages

Use conventional commits:

feat(tools): add my_tool for XYZ
fix(scheduler): resolve race condition in session save
docs(readme): update installation instructions
refactor(providers): extract base class for chat providers
test(core): add session persistence roundtrip test

Pull Requests

  • PRs must reference an existing issue
  • All checks must pass (lint, format, tests)
  • Keep PRs focused — one logical change per PR
  • Include tests for new functionality
  • Update docstrings for any changed public APIs

NVIDIA Model Policy

Per AGENTS.md:

  • Prefer native NVIDIA Nemotron models over Llama-Nemotron variants
  • Do not introduce Llama-based models as coding backends
  • For Nano 4B, use Nemotron 3 Nano 4B family (nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16)

Questions?

Open a discussion on GitHub.