nelgraph 1.0.3

GraphRAG knowledge base for codebases

nelgraph 🚀

An autonomous, zero-configuration Knowledge Graph builder and semantic search engine optimized for local codebases and AI testing agents. It ingests source code, AST Call Graphs, and Git history into a hybrid Graph-Vector database (Neo4j + ChromaDB) using DeepSeek V4-Flash.

🛠️ Installation

pip install nelgraph

🚀 CLI Usage

1. Initialize GraphRAG for your project

Navigate to your project directory and run:

nelgraph init

During initialization, it will prompt you for your OpenRouter API key, configure the local .env and .graphrag_data/ directory, start Neo4j inside a local Docker container, build structural nodes and indexes, and automatically install a git post-commit hook so the graph auto-syncs.

2. Manual Synchronization

If you want to manually trigger incremental synchronization:

nelgraph sync

3. Check Status

View current database metrics, indexed function counts, and enrichment coverage:

nelgraph status

4. Run Watcher

Run a file watcher that auto-syncs the graph in the background when files change:

nelgraph watch

🔌 Programmatic Python API

You can import nelgraph directly into your scripts or AI testing agents:

import nelgraph

# Configure (if .env is not present or needs custom config)
nelgraph.configure(
    codebase_path="/absolute/path/to/project",
    openrouter_api_key="your-openrouter-key"
)

# 1. Get snapshot of prioritized functions
snapshot = nelgraph.get_snapshot()
print(f"Total functions: {snapshot['total']}")

# 2. Retrieve detailed context for a function
ctx = nelgraph.get_function_context("process_order")
print(ctx.get("function", {}).get("raw_code"))

# 3. Mark function as tested/verified
nelgraph.mark_tested("process_order")