What Are the Dependencies for Graphify? A Complete Guide to Mandatory and Optional Packages

Graphify depends on NetworkX, NumPy, RapidFuzz, and a comprehensive suite of Tree-sitter parsers as mandatory requirements, with optional extras for databases, document processing, LLM providers, and multimedia handling.

Graphify is a Python library developed by Graphify-Labs that transforms source code into queryable knowledge graphs. Understanding the dependencies for Graphify is essential for installation planning, as the project splits its requirements into lean core packages and feature-specific optional groups declared in pyproject.toml.

Mandatory Core Dependencies

The base installation of Graphify requires four categories of packages that provide graph algorithms, numerical computing, fuzzy matching, and multi-language parsing capabilities.

NetworkX (>=3.4)

NetworkX provides the fundamental graph data structures and algorithms that power Graphify's knowledge representation. According to the Graphify-Labs/graphify source code, this library handles the construction and traversal of code dependency graphs in graphify/__init__.py.

NumPy (>=1.21)

NumPy enables high-performance array operations required for numerical computations within the graph analysis pipeline. The >=1.21 constraint ensures compatibility with modern array broadcasting features used throughout the codebase.

RapidFuzz (>=3.0)

RapidFuzz supplies fast fuzzy string matching capabilities, allowing Graphify to identify approximate matches between symbol names and references across different source files.

Tree-sitter Language Parsers

Graphify includes a family of Tree-sitter parsers covering dozens of programming languages. These mandatory parsers include support for Python, JavaScript, TypeScript, Go, Rust, Java, Groovy, C, C++, Ruby, C#, Kotlin, Scala, PHP, Swift, Lua, Zig, PowerShell, Elixir, Objective-C, Julia, Verilog, Fortran, Bash, and JSON. These packages enable Abstract Syntax Tree (AST) construction within the graphify/extractors/ directory.

Optional Dependency Groups

Graphify organizes additional functionality into optional dependency groups under the [project.optional-dependencies] table in pyproject.toml (lines 50-84). Install these using pip install graphifyy[<group>] syntax.

MCP and HTTP Transport

For Model Context Protocol (MCP) support and HTTP server capabilities, Graphify requires:

• mcp – The core MCP implementation
• starlette>=1.3.1 – ASGI framework used by graphify/serve.py for HTTP transport

Database Connectors

Graphify supports multiple graph and relational database backends:

• neo4j – For Neo4j graph database integration
• falkordb – FalkorDB connector support
• psycopg[binary] – PostgreSQL binary driver for relational storage

Document Processing

To parse non-code documents and convert them to graph nodes:

• pypdf>=6.12.0 – PDF text extraction
• markdownify – Markdown to text conversion
• python-docx – Microsoft Word document handling
• openpyxl – Excel spreadsheet parsing

File System and Visualization

Additional utilities include:

• watchdog – File system watching for live graph updates
• matplotlib – Graph visualization and plotting capabilities
• numpy>=2.0 – Required specifically for Python 3.13+ when using visualization features

Community Detection

The graspologic package provides advanced community detection algorithms, though this dependency is only available on Python < 3.13.

Multimedia and Speech Processing

For audio and video processing capabilities:

• faster-whisper – Speech-to-text transcription (requires Python >= 3.11)
• yt-dlp – YouTube and media platform downloading

LLM Provider Integrations

Graphify supports multiple large language model backends through optional packages:

• openai – OpenAI API and compatible endpoints (including Ollama and Kimi)
• tiktoken – Token counting for OpenAI and Gemini models
• anthropic – Claude API integration
• boto3 – AWS Bedrock access

Language-Specific Extras

Additional Tree-sitter parsers for specialized languages:

• jieba – Chinese text tokenization support
• tree-sitter-sql – SQL parsing capabilities
• tree-sitter-pascal, tree-sitter-dm, tree-sitter-hcl – Pascal, DM, and HCL language support

Where Dependencies Are Declared

All Graphify dependencies are centrally managed in the pyproject.toml file at the repository root. Lines 13-43 define mandatory requirements, while lines 50-84 specify the optional dependency groups. The graphify/__init__.py file exposes the public API and registers CLI entry points (graphify, graphify-mcp).

Installing Graphify with Specific Dependencies

To install only the core package with mandatory dependencies:

pip install graphifyy

To include specific features, use the extras syntax. For example, to install with Neo4j support and PDF processing:

pip install "graphifyy[neo4j,pdf]"

Multiple extras can be combined. This example installs database connectors, document handling, and LLM support:

pip install "graphifyy[neo4j,psycopg,documents,openai]"

After installation, verify the setup by building a graph from source code:

from graphify import Graphify

# Initialize with project root

g = Graphify(root_path="my_project/")

# Build the knowledge graph using the installed parsers

g.build()

# Query relationships (requires appropriate backend if using LLM features)

result = g.query("What functions call `process_data`?")
print(result)

How Dependencies Power Graphify's Architecture

The dependency structure directly maps to Graphify's internal architecture. The graphify/extractors/ directory contains language-specific modules that rely on the Tree-sitter parser packages to ingest source code. NetworkX and NumPy operate within the core graph building logic to construct and analyze code relationships.

When using MCP transport features, the graphify/serve.py module imports starlette to implement the ASGI HTTP server. Database connectors like neo4j and psycopg enable the persistence layer, while document processing packages extend the ingestion capabilities beyond raw source code.

Summary

• Graphify dependencies split into mandatory core packages (NetworkX, NumPy, RapidFuzz, Tree-sitter) and optional feature groups.
• Mandatory packages enable multi-language code parsing and graph construction without additional installation size.
• Optional groups cover databases (Neo4j, PostgreSQL, FalkorDB), documents (PDF, Word, Excel), LLM providers (OpenAI, Anthropic, AWS), and multimedia (Whisper, yt-dlp).
• Install specific combinations using pip install graphifyy[<group1>,<group2>] syntax.
• All dependencies are declared in pyproject.toml lines 13-84, with architecture files like graphify/serve.py and graphify/extractors/ consuming these packages.

Frequently Asked Questions

Do I need all optional dependencies to use Graphify?

No. Graphify is designed with a minimal core installation. You only need to install optional dependencies when using specific features like Neo4j storage, PDF ingestion, or LLM querying. The base package handles source code parsing and basic graph operations without extras.

What Python versions are supported by Graphify?

Graphify supports Python 3.9 through 3.13, though specific optional dependencies have version constraints. The graspologic package only works on Python < 3.13, while faster-whisper requires Python >= 3.11. The core mandatory dependencies work across all supported versions.

Can I use Graphify without installing database drivers?

Yes. Database connectors like neo4j, falkordb, and psycopg are completely optional. Graphify stores graphs in memory using NetworkX by default. You only need database drivers if you want to persist or query graphs using external storage systems.

How do I install Graphify with PDF support?

Install the PDF extra using pip install "graphifyy[pdf]" or combine it with other features like pip install "graphifyy[pdf,neo4j]". This installs pypdf>=6.12.0, markdownify, python-docx, and openpyxl, enabling Graphify to parse PDFs, Word documents, Excel files, and Markdown into the knowledge graph.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}