GestaltDB Documentation

GestaltDB is a pure Python graph database toolkit for attributed graphs. It stores nodes, edges, labels, typed adjacency records, and indexes on embedded key-value backends such as LMDB, LevelDB, and RocksDB/PyRex.

Start with the quickstart if you want to create a graph and run a query. Use the topic pages for backend selection, serializers, Cypher syntax, ingestion, sampling, and benchmarks.

User Guide

• Installation
  • Python Version
  • Install With uv
  • Install With pip
  • Install From GitHub
  • Notebook Kernel Example
  • Optional Dependencies
  • Pre-Release Installs
• Quickstart
  • Create a Graph
  • Nodes and Edges
  • Labels and Indexes
  • Cypher Query
  • Typed Traversal
  • Bulk Inserts
  • Close the Store
  • Next Steps
• Storage Backends
  • Backend Selection Summary
  • LMDB Backend
  • LevelDB Backend
  • RocksDB Backend
  • Indexes
  • Transactions
  • Backend Selection Pattern
  • Cleanup
• Serializers
  • Pickle Serializer
  • JSON Serializer
  • MessagePack Serializer
  • Protobuf Serializer
  • Using a Serializer With GraphDB
• Typed Traversal and Sampling
  • Create a Typed Graph
  • Query Typed Neighbors
  • Query Typed Edges
  • Sample Neighbors
  • Object-Based Sampling Patterns
  • Dictionary-Based Sampling Patterns
  • Cypher Sampling Procedure
  • Sample a Materialized Subgraph
  • Rebuild Typed Adjacency
• Cypher Queries
  • Basic Result Shape
  • Node Scans
  • Typed Relationship Traversal
  • Filtering
  • Projection and Result Shaping
  • Chained MATCH Clauses
  • Sampling Procedure
  • Current Limitations
• Performance and Benchmarks
  • Practical Recommendations
  • Install Benchmark Dependencies
  • Backend Benchmarks
  • Columnar Ingestion Benchmarks
  • Structured JSON Entity Ingestion
  • End-to-End Ingestion Insights
  • Transaction-Capable RocksDB Benchmark
  • RocksDB Tuning and Compaction Benchmarks
  • External Graph Database Benchmarks
  • Benchmark Caveats
• Notebook Usage
  • Choose the Correct Kernel
  • Minimal Notebook Cell
  • Close Stores Between Runs
  • Troubleshooting plyvel

API Reference

• API Reference
  • Graph Models and Database
  • Sampling Configuration
  • Columnar Ingestion
  • Cypher Queries
  • Key-Value Stores
  • Serializers

Quick Example

from gestaltdb.graphdb import Edge, GraphDB, Node
from gestaltdb.kvstores import LMDBStore
from gestaltdb.serializers import PickleSerializer

graph_db = GraphDB(LMDBStore(path="graph_lmdb_example"), PickleSerializer())

alice = Node(node_id="alice", labels=["Person"], properties={"name": "Alice"})
bob = Node(node_id="bob", labels=["Person"], properties={"name": "Bob"})
graph_db.put_node(alice)
graph_db.put_node(bob)

graph_db.put_edge(Edge(
    edge_id="alice-knows-bob",
    source="alice",
    target="bob",
    properties={"type": "knows", "since": 2024},
))

result = graph_db.query('MATCH (a:Person {name: "Alice"}) MATCH (a)-[:knows]->(b) RETURN a.id, b.name')
print(result.records)

graph_db.close()