Skip to content

getting_started_python.md

Latest commit

 

History

History
78 lines (56 loc) · 2.88 KB

getting_started_python.md

File metadata and controls

78 lines (56 loc) · 2.88 KB

Getting Started with CodeNERD for Python

This guide will walk you through setting up CodeNERD for a Python project (like ArangoRDF). CodeNERD treats your code as a "Glass Box," indexing it into a knowledge graph and allowing it to safely execute code in sandboxed Docker containers.

Prerequisites

  1. CodeNERD CLI Installed: Ensure you have built the nerd binary.
  2. Docker Running: CodeNERD uses Docker to spin up ephemeral environments for your Python code.
  3. Python Project: A standard Python project with requirements.txt, setup.py, or pyproject.toml.

1. Initialization

Navigate to your Python project root.

cd ~/my-python-project
nerd init

What happens:

  • .nerd/ directory created: Stores the local database and configuration.
  • Indexing: CodeNERD parses all your .py files using Tree-sitter to understand classes, functions, and imports.
  • Fact Generation: It creates a highly detailed "symbol graph" of your code.
  • Vector Embeddings: It generates embeddings for your code to enable semantic search.

2. Using the Interactive Agent

Launch the TUI (Text User Interface):

nerd

You are now in a chat session with the agent. The agent has context of your entire codebase.

Example Workflows

A. Explaining Code

You: "Explain how the GraphConnection class works and where it is instantiated." Agent: Will query its internal knowledge graph (Mangle) to find symbols and cross-references, then explain it to you.

B. Refactoring

You: "Rename the connect method to establish_connection and update all callers." Agent:

  1. Identifies all references.
  2. Proposes a plan.
  3. Generates a diff.
  4. Asks for your approval (/approve).

C. Running Tests (The "Tactile" Layer)

You: "Run the tests for the connection module." Agent:

  1. Spins up a Docker container mirroring your environment.
  2. Installs dependencies (from requirements.txt).
  3. Executes pytest inside the container.
  4. Reports usage and results back to you.

3. Sandboxed Execution

One of CodeNERD's strongest features for Python is Sandboxed Execution. It doesn't run code on your host machine.

When you ask it to "Run tests" or "Verify this fix":

  1. It creates a Docker container tagged nerd-python-<repo>.
  2. It mounts your code (or clones it).
  3. It creates a virtual environment inside the container.
  4. It executes the command.

This means you can let the agent iterate on fixes without worrying about it messing up your local environment.

Troubleshooting

"No tests found"

  • Ensure you have pytest installed in your project.
  • Check that your test files usually follow test_*.py.

"Docker error"

  • Make sure Docker Desktop (or engine) is running.
  • Run docker ps to see if CodeNERD containers are active.