graph-ted-db

Local property-graph storage for Python. Your data lives in a folder on disk; open it from your process, query it in-process, and optionally sync that folder like any other files. No database server to run for the default path.

Package: graph-ted-db · Import: graph_ted_db · Repo: graph-ted/graph-ted-db · License: MIT

Status: the public API and on-disk format below are what we ship. pip install graph-ted-db is the intended install; until the package is on PyPI, use a checkout (pip install -e .) or a wheel path.

Install

pip install graph-ted-db

Requires Python 3.10+.

Quick start

from graph_ted_db import GraphStore, init_graph

init_graph("./my-graph", name="demo", exist_ok=True)
g = GraphStore.open("./my-graph")

alice = g.make_node(labels=["Person"], props={"name": "Alice"})
bob = g.make_node(labels=["Person"], props={"name": "Bob"})
g.make_edge(type="KNOWS", from_id=alice.id, to_id=bob.id)

print(g.get_node(alice.id).props["name"])

CLI:

graph-ted-db init ./my-graph --name demo --exist-ok
graph-ted-db put-node ./my-graph --label Person --prop name=Alice
graph-ted-db ls-nodes ./my-graph

graphted-db is an alias for the same command.

Optional HTTP (localhost only by default; see HTTP docs):

graph-ted-db serve ./my-graph
# in another terminal:
curl -s http://127.0.0.1:8099/health
curl -s http://127.0.0.1:8099/cypher -H 'Content-Type: application/json' \
  -d '{"query":"MATCH (n:Person) RETURN n.name AS name"}'

Features

• Folder = database — nodes, edges, and embeddings as sharded JSONL; readable, backupable, syncable
• In-process API — put / get / iter / delete without opening a network port
• openCypher queries — g.execute(...) runs a documented subset of openCypher for graph-pattern queries (openCypher subset)
• Optional HTTP — localhost endpoint for openCypher queries when another process needs the store (HTTP docs)
• Sync-friendly — multiple devices can write at once; last-write-wins per record; conflict-copy shards merged on read
• NetworkX-style helpers — add_node, add_edge, neighbors, and related methods on the same GraphStore (no algorithm suite)

What it is and isn't

It is:

• A Python library that stores a property graph (nodes, edges, properties, embeddings) as plain JSONL files in one folder.
• In-process: open, read, write and query from your own Python process. No server to install or run.
• A documented subset of openCypher, plus NetworkX-style helpers.
• Sync-friendly across devices: multiple devices can add and edit at the same time; concurrent edits to the same record resolve to the latest version. The folder can live in OneDrive, Dropbox or an rclone bisync folder (see Sharing).
• Small: no runtime dependencies beyond the Python standard library.

It isn't:

• A database server, a hosted service or a multi-tenant system. The optional HTTP endpoint is for local processes on the same machine.
• Encrypted. Files are ordinary files; protect them with disk encryption and OS permissions.
• A complete Cypher implementation or a Neo4j replacement. Unsupported clauses fail with an error rather than being ignored.
• A graph-algorithm library. Use NetworkX or similar on exported data.
• Built for very large graphs. The whole graph is loaded into memory when it opens; see the tested sizes in Overview.
• Stable across minor versions before 1.0 (see Versioning).
• Telemetry-enabled. It makes no network calls of its own.

When to use it

Good fit for prototypes, local tools, agents, and small trusted groups that want a property graph next to the app.

Choose a server graph database when you need multi-tenant hosting, fine-grained remote auth, or a complete query language and enterprise feature set out of the box.

Security

Data is stored as ordinary files. The library does not encrypt the graph folder. Security matches your environment: disk or volume encryption, OS account permissions, backups, sync tools, and whether anything listens on the network.

• Airgapped host + encrypted volume — can be very strong; treat that environment as your boundary.
• Synced or shared folders — anyone with access to the folder can read what you stored (including properties).
• HTTP serve — defaults to localhost; binding beyond loopback requires an explicit host and a token. That protects the port, not the files on disk.

Choose storage and network to match how sensitive the data is.

Documentation

Docs: graph-ted.com/graph-ted-db/docs, built from the docs/ folder in this repository. Preview locally: see PREVIEW.md.

Trademarks

openCypher is a trademark of Neo4j, Inc. Other names are trademarks of their respective owners; no endorsement implied. See Trademarks.

Versioning

Pre-1.0 (0.1.0). Until 1.0.0, minor versions may include breaking changes; they are noted in release notes. From 1.0.0, breaking changes bump the major version.

Development

pip install -e ".[dev]"
pytest

Feedback and issues

• Bugs, enhancement requests, docs problems: open an issue using one of the forms.
• Questions and ideas: Discussions.
• Security vulnerabilities: report privately as described in SECURITY.md, never in a public issue.

Contributing: see CONTRIBUTING.md.

Related

graph-ted-db is the local store for the graph-ted toolkit (graph-ted.com). This repository is the database library only.