API Reference¶

Complete API documentation for pygrad.

Module Overview¶

pygrad provides a numpy-style API for building knowledge graphs from Python repositories.

import pygrad as pg

# Core functions
await pg.add("https://github.com/owner/repo")
result = await pg.search("https://github.com/owner/repo", "How to use X?")
datasets = await pg.list()
await pg.delete("https://github.com/owner/repo")
await pg.visualize("./graph.html")

API Modules¶

Module	Description
Core	Main API functions (`add`, `search`, `list`, `delete`, `visualize`)
Repository	Repository cloning and identification utilities
Processor	Python code processing and XML generation
Parser	TreeSitter-based Python parsing

Quick Reference¶

Core Functions¶

Function	Description
`pg.add(url)`	Add a repository to the knowledge graph
`pg.search(url, query)`	Search a repository's knowledge graph
`pg.list()`	List all indexed datasets
`pg.delete(url)`	Delete a repository from the knowledge graph
`pg.visualize(path)`	Export the knowledge graph as HTML
`pg.get_dataset(name)`	Get a dataset by name

Low-Level APIs¶

Class/Function	Module	Description
`RepoTreeSitter`	`parser`	Parse Python files with TreeSitter
`PythonRepositoryProcessor`	`processor`	Process repositories into XML
`process_repository()`	`processor`	High-level repository processing
`clone_repository()`	`repository`	Clone a Git repository
`get_repository_id()`	`repository`	Extract repository identifier from URL

Type Hints¶

All functions are fully typed. Enable type checking in your IDE for the best experience:

from pygrad import add, search, list, delete, visualize
from pygrad import ClassInfo, FunctionInfo  # Data classes
from pygrad import RepoTreeSitter  # Parser
from pygrad import PythonRepositoryProcessor  # Processor

Async/Await¶

All core functions are async and must be awaited:

import asyncio
import pygrad as pg

async def main():
    await pg.add("https://github.com/owner/repo")
    result = await pg.search("https://github.com/owner/repo", "query")
    print(result)

asyncio.run(main())

Configuration¶

pygrad uses environment variables for configuration. See the Configuration Guide for details.

Variable	Description	Default
`LLM_PROVIDER`	LLM provider (`ollama`, `openai`)	`ollama`
`LLM_MODEL`	Model name	`qwen3-coder:30b`
`EMBEDDING_PROVIDER`	Embedding provider	`ollama`
`EMBEDDING_MODEL`	Embedding model	`embeddinggemma:latest`