Graphify: Transform Unstructured Data into Knowledge Graphs

Overview

Graphify turns code, documentation, PDFs, screenshots, and notes into a connected knowledge graph. Its FalkorDB export can load that graph directly into FalkorDB for persistent querying. Use Graphify when you want to:

build a graph from an existing repository, docs set, or mixed corpus
expose concept and relationship structure across files and folders
generate a reusable graph for downstream retrieval, agent memory, and exploration
keep an artifact that can be loaded into FalkorDB for long-lived querying

This is a best-fit workflow for graph-first repository analysis, documentation indexing, and knowledge-base construction.

Install Graphify

Graphify runs as a skill in AI coding assistants. The examples below cover Claude Code and Codex; see the Graphify README for other supported assistants. Install the CLI with the FalkorDB integration and register the skill:

pip install "graphifyy[falkordb]"
graphify install

pip install "graphifyy[falkordb]"
graphify install --platform codex

Then open your AI coding assistant in any directory and run the skill on a folder or repository:

/graphify ./my-project

$graphify ./my-project

Graphify supports additional modes such as --wiki, --watch, and --mcp depending on how you want to consume the graph.

Load the graph into FalkorDB

Start FalkorDB locally:

docker run -p 6379:6379 falkordb/falkordb:latest

Then run Graphify with the FalkorDB push option. The target graph is named graphify by default:

/graphify ./my-project --falkordb-push falkordb://localhost:6379

$graphify ./my-project --falkordb-push falkordb://localhost:6379

Graphify uses MERGE, so pushing again does not duplicate unchanged entities. It does not remove entities that disappeared from the source corpus; clear or replace the target graph before pushing when deletions must be reflected. If you need a portable Cypher artifact instead, use --falkordb; Graphify writes the statements to graphify-out/cypher.txt. Direct push is recommended because FalkorDB executes one Cypher statement at a time.

Querying the imported graph

After the graph is in FalkorDB, you can use regular FalkorDB queries to explore the data, for example:

MATCH (n)-[r]->(m)
RETURN type(r), count(r)
ORDER BY count(r) DESC
LIMIT 20;

MATCH (n)
WHERE n.name CONTAINS "query"
RETURN n
LIMIT 50;

Tips

For large corpora, keep the Graphify output as a reusable artifact. Re-push it for additive updates, or replace the target graph when source entities were removed.

Reference

Frequently Asked Questions

What is Graphify?

Graphify turns code, documentation, PDFs, screenshots, and notes into a connected knowledge graph. It runs as a skill inside AI coding assistants, and its FalkorDB export loads that graph into FalkorDB so you can query it persistently with Cypher.

How do I install Graphify with FalkorDB support?

Install and register Graphify with the FalkorDB extra: pip install "graphifyy[falkordb]" && graphify install. For Codex, register with graphify install --platform codex. The [falkordb] extra pulls in the falkordb SDK, which the exporter needs for --falkordb-push.

Which AI coding assistants does Graphify work with?

The examples on this page cover Claude Code and Codex. Other assistants are supported too — see the Graphify README for the current list and the matching install command.

How do I load a Graphify knowledge graph into FalkorDB?

Start FalkorDB with docker run -p 6379:6379 falkordb/falkordb:latest, then run the skill with the push option, for example /graphify ./my-project --falkordb-push falkordb://localhost:6379. The target graph is named graphify by default.

Should I use --falkordb-push or --falkordb?

Use --falkordb-push in most cases: it writes straight into FalkorDB, which is faster because FalkorDB executes one Cypher statement at a time. Use --falkordb when you need a portable artifact instead — Graphify writes the statements to graphify-out/cypher.txt.

Does pushing Graphify to FalkorDB twice create duplicate nodes?

No. Graphify uses MERGE, so re-pushing unchanged entities does not duplicate them. However, it does not remove entities that disappeared from the source — clear or replace the target graph before pushing when deletions must be reflected.

How do I query a graph that Graphify created?

Use ordinary Cypher. For example, MATCH (n)-[r]->(m) RETURN type(r), count(r) ORDER BY count(r) DESC summarizes the relationship types Graphify extracted from your corpus.

​Overview

​Install Graphify

​Load the graph into FalkorDB

​Querying the imported graph

​Tips

​Reference

​Frequently Asked Questions