graphify-go

graphify-go

A high-performance Go port of graphify — turn any codebase into a navigable knowledge graph with community detection and an honest audit trail.

graphify-go is a complete reimplementation of the original Python-based graphify in Go, delivering 20x faster performance with zero Python dependency. Single binary, no setup required.

Performance Comparison

Benchmarked on a real React Native project (721 files, 783 nodes, 8,403 edges) on Apple M-series.

Why Go?

The original graphify by @safishamsi is an excellent tool for building knowledge graphs from codebases. This Go port was created to:

• Eliminate Python dependency — no pip install, no virtual environments, no version conflicts
• Leverage native tree-sitter — tree-sitter is written in C; Go's CGO bindings are significantly faster than Python's ctypes bridge
• Single binary distribution — go install and you're done, on any platform
• Parallel extraction — Go's goroutines enable true concurrent file processing

Features

• 18 languages supported via tree-sitter AST parsing: Python, JavaScript, TypeScript, Go, Java, Rust, C, C++, C#, Ruby, Swift, Kotlin, Scala, PHP, Lua, and more
• Louvain community detection — pure Go implementation, no external libraries
• Three outputs: interactive HTML (vis.js), queryable JSON, audit report (Markdown)
• SHA256 file caching — incremental re-extraction on subsequent runs
• Cross-file import resolution — traces imports across module boundaries
• Confidence tagging — every edge is marked EXTRACTED, INFERRED, or AMBIGUOUS
• AI agent integration — works as a skill in Claude Code, Codex, and Gemini CLI

Install

go install github.com/techinpark/graphify-go/cmd/graphify@latest

Or build from source:

git clone https://github.com/techinpark/graphify-go.git
cd graphify-go
go build -o graphify ./cmd/graphify/

Usage

# Analyze current directory
graphify .

# Analyze a specific path
graphify ./src

# Incremental update (only re-extract changed files)
graphify . --update

# Skip HTML visualization
graphify . --no-viz

# Deep extraction mode
graphify . --mode deep

Output

graphify-out/
├── graph.html         # Interactive vis.js visualization — open in browser
├── graph.json         # NetworkX-compatible JSON — queryable, persistent
├── GRAPH_REPORT.md    # Audit report with God Nodes, Surprising Connections
└── cache/             # SHA256 cache for incremental re-runs

graph.html — Interactive Visualization

Dark-themed interactive graph powered by vis.js:

• Nodes sized by degree, colored by community
• Click to inspect node details and neighbors
• Search nodes by name
• Community legend with toggle visibility
• Dashed edges for INFERRED/AMBIGUOUS relationships

GRAPH_REPORT.md — Audit Report

• God Nodes — most connected core abstractions
• Surprising Connections — unexpected cross-file/cross-community edges
• Communities — auto-detected module clusters with cohesion scores
• Knowledge Gaps — isolated nodes, thin communities, high ambiguity
• Suggested Questions — what this graph is uniquely positioned to answer

AI Agent Integration

graphify-go works as a skill/command in major AI coding assistants. Install once, then use natural language or slash commands to analyze any codebase.

Claude Code

# Install the skill
graphify install

# Then in Claude Code, type:
/graphify .

This registers SKILL.md at ~/.claude/skills/graphify-go/ and adds a trigger to ~/.claude/CLAUDE.md. When you type /graphify, Claude Code runs the binary and presents the results.

OpenAI Codex

# Install for Codex
graphify install --platform codex

# Then in Codex, type:
$graphify .

This registers the skill at ~/.codex/skills/graphify-go/ and creates an AGENTS.md entry.

Google Gemini CLI

Gemini CLI reads project instructions from GEMINI.md. Add graphify-go manually:

# 1. Build or install graphify
go install github.com/techinpark/graphify-go/cmd/graphify@latest

# 2. Add to GEMINI.md in your project root
cat >> GEMINI.md << 'EOF'

## graphify-go

To analyze this codebase's architecture, run:

```bash
graphify .

Then read graphify-out/GRAPH_REPORT.md for god nodes, communities, and surprising connections. Open graphify-out/graph.html in a browser for interactive visualization. EOF

Then in Gemini CLI, ask: *"Run graphify and summarize the architecture"*

### Uninstall

```bash
graphify uninstall                    # Claude Code
graphify uninstall --platform codex   # Codex

How It Works with AI Agents

All agents execute the same graphify binary and produce identical outputs. The SKILL.md instructs the agent to:

1. Run graphify <path> — the full pipeline completes in under 1 second
2. Read graphify-out/GRAPH_REPORT.md and present a summary
3. Suggest opening graph.html for interactive exploration

Architecture

cmd/graphify/          CLI entry point (cobra)
internal/
├── detect/            File discovery & classification
├── extract/           Tree-sitter AST extraction (18 languages)
├── graph/             Graph data structures & operations
├── cluster/           Louvain community detection (pure Go)
├── analyze/           God nodes, surprising connections, questions
├── export/            HTML (vis.js), JSON, Markdown report
├── cache/             SHA256-based file caching
└── skill/             AI agent skill installation

Supported Languages

Credits

This project is a Go port of the original graphify by Safi Shamsi. The original project introduced the concept of turning any folder into a queryable knowledge graph with community detection and an honest audit trail. All credit for the core idea, graph schema, and output format goes to the original author.

License

MIT