How to Export the Code-Review Graph to Neo4j, GraphML, or Obsidian

Export the knowledge graph to Neo4j Cypher, GraphML, or Obsidian Vault using the GraphStore API and CLI commands provided by the code-review-graph project.

The code-review-graph project provides multiple export formats for its knowledge base, enabling integration with graph databases, visualization tools, and personal knowledge management systems. All formats build on the same low-level data representation returned by export_graph_data(store), which extracts nodes, edges, and community structures from the central GraphStore object.

---

Export Formats Overview

The project supports four distinct export targets:

• Neo4j (Cypher) — Generates executable Cypher statements for Neo4j graph database import
• GraphML — Standard XML-based graph format compatible with Gephi, yEd, and NetworkX
• Obsidian Vault — Markdown-based knowledge base with WikiLinks for interactive exploration
• JSON/SVG — Additional programmatic and visualization formats

Each format is implemented in code_review_graph/exports.py and accessible via both Python API and CLI.

---

Exporting to Neo4j (Cypher)

The Neo4j exporter produces a self-contained Cypher script that recreates the entire graph in a Neo4j instance.

How It Works

The export_neo4j_cypher() function in [code_review_graph/exports.py](https://github.com/tirth8205/code-review-graph/blob/main/code_review_graph/exports.py) performs two main operations:

1. Node creation — Each node becomes a Cypher CREATE statement with:

   • Label derived from the node's kind field (Class, Function, Module, etc.)
   • Properties formatted by _cypher_props() including qualified_name, name, file_path, language, and optional community_id

2. Edge creation — For every relationship, a MATCH-CREATE pair links nodes by their escaped qualified_name values using _cypher_escape()

Programmatic Export

from pathlib import Path
from code_review_graph.graph import GraphStore
from code_review_graph.exports import export_neo4j_cypher

# Load an existing GraphStore

store = GraphStore.load(Path("path/to/data_dir"))
out_file = Path("graph.cypher")

export_neo4j_cypher(store, out_file)

The resulting graph.cypher contains statements like:

CREATE (:Class {qualified_name: 'my_pkg.MyClass', name: 'MyClass', 
                file_path: 'my_pkg.py', language: 'python'});

MATCH (a {qualified_name: 'my_pkg.MyClass'}), 
      (b {qualified_name: 'my_pkg.helper'})
CREATE (a)-[:CALLS]->(b);

Import into Neo4j using:

cat graph.cypher | cypher-shell -u neo4j -p <password>

CLI Command

code-review-graph export --format cypher

This writes graph.cypher to the default data directory and prints the full path.

---

Exporting to GraphML

The GraphML exporter generates standard XML that preserves all node/edge properties and community assignments for external graph analysis tools.

Key Implementation Details

• Located in [code_review_graph/exports.py](https://github.com/tirth8205/code-review-graph/blob/main/code_review_graph/exports.py) alongside other format handlers
• Uses the same export_graph_data(store) foundation as Neo4j and Obsidian exports
• Compatible with Gephi, Cytoscape, yEd, and NetworkX

Usage

from code_review_graph.exports import export_graphml

export_graphml(store, Path("graph.graphml"))

code-review-graph export --format graphml

---

Exporting to Obsidian Vault

The Obsidian exporter materializes the knowledge graph as a navigable Markdown-based vault where relationships become bidirectional WikiLinks.

How It Works

The export_obsidian_vault() function performs five transformation steps:

1. Slug generation — Node names are normalized to filesystem-safe identifiers via _obsidian_slug() (exports.py:46-49)

2. Adjacency collection — Both incoming and outgoing edges are gathered for each node's "Connections" section

3. Markdown composition — Each node becomes slug.md containing:

   • YAML front-matter (kind, file, language, community, tags)
   • Human-readable heading
   • ## Connections section with [[wikilink|DisplayName]] entries

4. Community summaries — _COMMUNITY_<id>.md files document each community's size, cohesion, and dominant language

5. Index generation — _INDEX.md provides global statistics and navigation

Programmatic Export

from pathlib import Path
from code_review_graph.graph import GraphStore
from code_review_graph.exports import export_obsidian_vault

store = GraphStore.load(Path("path/to/data_dir"))
vault_dir = Path("my-obsidian-vault")

export_obsidian_vault(store, vault_dir)

Resulting vault structure:

my-obsidian-vault/
├── my-class.md          # Individual node pages

├── my-function.md
├── _COMMUNITY_1.md      # Community overview

├── _INDEX.md            # Global navigation

└── .obsidian/           # (create manually if needed)

Open the directory in Obsidian to use Graph view and Backlinks for interactive exploration.

CLI Command

code-review-graph export --format obsidian

---

Common Export Pipeline

All formats share a unified data extraction flow:

from code_review_graph.visualization import export_graph_data

# Core extraction used by every exporter

data = export_graph_data(store)

# → {"nodes": [...], "edges": [...], "communities": [...]}

The GraphStore in [code_review_graph/graph.py](https://github.com/tirth8205/code-review-graph/blob/main/code_review_graph/graph.py) persists this structure to disk, while format-specific functions in [code_review_graph/exports.py](https://github.com/tirth8205/code-review-graph/blob/main/code_review_graph/exports.py) handle serialization.

---

CLI Reference

Command	Purpose	Output
code-review-graph export --format cypher	Neo4j import script	graph.cypher
code-review-graph export --format graphml	Standard graph XML	graph.graphml
code-review-graph export --format obsidian	Markdown vault directory	obsidian_vault/
code-review-graph export --format json	Raw JSON serialization	graph.json
code-review-graph export --format svg	Static vector visualization	graph.svg

The export dispatch logic resides in [code_review_graph/cli.py](https://github.com/tirth8205/code-review-graph/blob/main/code_review_graph/cli.py#L1915-L1925).

---

Summary

• All exports build on export_graph_data() — The unified extraction in visualization.py ensures consistency across formats
• Neo4j export generates executable Cypher — Node creation followed by MATCH-CREATE edge statements with proper escaping
• Obsidian export creates navigable Markdown — WikiLinks preserve graph topology for PKM workflows
• GraphML provides tool interoperability — Standard XML for Gephi, NetworkX, and other analysis platforms
• Both API and CLI access — Use export_neo4j_cypher(), export_obsidian_vault(), export_graphml() directly, or invoke via code-review-graph export --format <name>

---

Frequently Asked Questions

Does the tool support batch exporting multiple formats at once?

No. As implemented in code_review_graph/cli.py, the export command accepts a single --format argument. Run the command multiple times with different format values to generate all desired outputs, or call the programmatic API functions in sequence within a Python script.

Can I customize the Cypher output for existing Neo4j schemas?

The current export_neo4j_cypher() implementation in exports.py generates standalone CREATE statements with hardcoded property names. For schema customization, post-process the output or modify the _cypher_props() helper at lines 73-88 to inject additional labels or transform property keys before emission.

What Obsidian plugins work best with the exported vault?

The generated vault is optimized for Graph view (core plugin) and Backlinks panel. Enable Tags pane to filter by language or community front-matter fields. Community plugins like Dataview can query the YAML metadata for advanced filtering, though the base export requires no plugins.

How are qualified names escaped in Cypher to handle special characters?

The _cypher_escape() function at lines 73-77 of exports.py handles escaping by processing backslashes and quotes in node identifiers. This ensures that qualified names containing dots, brackets, or other Cypher-sensitive characters parse correctly when imported.

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