DBT Graph Loader

Transform your DBT project's lineage and metadata into queryable knowledge graphs

DBT Graph Loader is a Python library that loads DBT (Data Build Tool) metadata into graph databases, enabling you to explore, query, and visualize your data lineage as an interactive knowledge graph.

🚀 Features

• 🔄 Multiple Graph Databases: Native support for Neo4j and FalkorDB
• 📊 Complete DBT Coverage: Models, sources, tests, macros, seeds, snapshots, and operations
• 🔗 Rich Relationships: Dependencies, references, macro usage, and test coverage mapping
• 📁 Flexible Input: Load from manifest.json and catalog.json files or strings
• ⚡ Incremental Updates: Diff two manifests and apply only what changed — no full reload needed
• 🛠️ Easy CLI: Simple command-line interface for batch operations
• 🐍 Python API: Programmatic access for integration into data pipelines
• 📈 Graph Analytics: Built-in statistics and insights about your data lineage
• 🐳 Docker Ready: Easy containerization and deployment

📦 Installation

Using Poetry (Recommended)

poetry add dbt-graph-loader

Using pip

pip install dbt-graph-loader

Development Installation

# Clone the repository
git clone https://github.com/ponderedw/dbt-graph-loader.git
cd dbt-graph-loader

# Install with Poetry
poetry install

# Or with pip
pip install -e .

🎯 Quick Start

1. Generate DBT Metadata Files

First, ensure you have the required DBT files:

cd your-dbt-project
dbt compile  # Generates manifest.json
dbt docs generate  # Generates catalog.json (optional but recommended)

2. Load into Neo4j

# Using CLI
dbt-graph-loader neo4j \
    --uri bolt://localhost:7687 \
    --username neo4j \
    --password your_password \
    --manifest target/manifest.json \
    --catalog target/catalog.json

3. Load into FalkorDB

# Using CLI
dbt-graph-loader falkordb \
    --host localhost \
    --port 6379 \
    --graph-name my_dbt_lineage \
    --manifest target/manifest.json \
    --catalog target/catalog.json

📋 Supported DBT Resources

Resource Type	Description	Properties Captured
Models	DBT models and their transformations	Materialization, dependencies, descriptions, tags
Sources	External data sources	Freshness rules, schemas, descriptions
Seeds	CSV files loaded as tables	File metadata, configurations
Snapshots	Slowly changing dimension tables	Strategies, unique keys, timestamps
Tests	Data quality tests	Severity levels, test parameters, attached nodes
Macros	Reusable SQL code blocks	Arguments, package info, usage patterns
Operations	Pre/post hooks and run operations	Execution context, dependencies

🔗 Graph Relationships

The loader creates rich relationships between your DBT resources:

• DEPENDS_ON: Direct dependencies between any resources
• REFERENCES: Model-to-model references via ref() functions
• USES_MACRO: Macro usage relationships
• TESTS: Test-to-resource relationships

🛠️ Usage

Command Line Interface

Neo4j Options

dbt-graph-loader neo4j --help

Options:
  --uri TEXT        Neo4j connection URI (required)
  --username TEXT   Neo4j username (required)  
  --password TEXT   Neo4j password (required)
  --manifest TEXT   Path to manifest.json (required)
  --catalog TEXT    Path to catalog.json (optional)

FalkorDB Options

dbt-graph-loader falkordb --help

Options:
  --host TEXT          FalkorDB host (default: localhost)
  --port INTEGER       FalkorDB port (default: 6379)
  --graph-name TEXT    Graph name (default: dbt_graph)
  --username TEXT      FalkorDB username (optional)
  --password TEXT      FalkorDB password (optional)
  --manifest TEXT      Path to manifest.json (required)
  --catalog TEXT       Path to catalog.json (optional)
  --incremental-run    Only apply changes between old and new manifest (default: false)
  --old-manifest TEXT  Path to the previous manifest.json (required when --incremental-run is set)

Incremental update

When --incremental-run is set, the loader diffs the two manifests by node checksum and applies only the minimum set of changes:

• Removed nodes — deleted from the graph along with all their relationships
• Changed nodes — properties updated in-place; outgoing relationships refreshed
• Added nodes — inserted with their relationships
• Unchanged nodes — not touched

dbt-graph-loader falkordb \
    --host localhost \
    --manifest target/manifest.json \
    --incremental-run \
    --old-manifest target/manifest_previous.json

This is significantly faster than a full reload for large projects where only a subset of models changes between runs.

Python API

Neo4j Integration

from dbt_graph_loader.loaders.neo4j_loader import DBTNeo4jLoader

# Initialize the loader
loader = DBTNeo4jLoader(
    neo4j_uri="bolt://localhost:7687",
    username="neo4j",
    password="your_password"
)

try:
    # Load from files
    loader.load_dbt_to_neo4j_from_files(
        manifest_path="target/manifest.json",
        catalog_path="target/catalog.json"
    )
    
    # View statistics
    loader.get_graph_stats()
    
finally:
    loader.close()

FalkorDB Integration

from dbt_graph_loader.loaders.falkordb_loader import DBTFalkorDBLoader

# Initialize the loader
loader = DBTFalkorDBLoader(
    host="localhost",
    port=6379,
    graph_name="dbt_lineage",
    username="your_username",  # if auth enabled
    password="your_password"   # if auth enabled
)

try:
    # Load from files
    loader.load_dbt_to_falkordb(
        manifest_path="target/manifest.json",
        catalog_path="target/catalog.json"
    )
    
    # Load from strings (useful for APIs)
    with open("target/manifest.json") as f:
        manifest_str = f.read()
    with open("target/catalog.json") as f:
        catalog_str = f.read()
        
    loader.load_dbt_to_falkordb_from_strings(manifest_str, catalog_str)
    
    # View statistics
    loader.get_graph_stats()
    
finally:
    loader.close()

Convenience Functions

from dbt_graph_loader import load_to_neo4j, load_to_falkordb, incremental_update_falkordb

# Simple Neo4j loading
load_to_neo4j(
    uri="bolt://localhost:7687",
    username="neo4j",
    password="password",
    manifest_path="target/manifest.json",
    catalog_path="target/catalog.json"
)

# Simple FalkorDB loading (full reload)
load_to_falkordb(
    host="localhost",
    port=6379,
    graph_name="dbt_lineage",
    manifest_path="target/manifest.json",
    catalog_path="target/catalog.json"
)

# Incremental FalkorDB update
incremental_update_falkordb(
    host="localhost",
    port=6379,
    graph_name="dbt_lineage",
    old_manifest_path="target/manifest_previous.json",
    new_manifest_path="target/manifest.json",
    catalog_path="target/catalog.json"  # optional
)

Or directly via the loader:

from dbt_graph_loader.loaders.falkordb_loader import DBTFalkorDBLoader

loader = DBTFalkorDBLoader(host="localhost", port=6379)
loader.incremental_update_from_files(
    old_manifest_path="target/manifest_previous.json",
    new_manifest_path="target/manifest.json",
    catalog_path="target/catalog.json"
)

🔍 Example Queries

Once your DBT metadata is loaded, you can query the graph using Cypher (Neo4j) or OpenCypher (FalkorDB).

Neo4j Cypher Examples

// Find all models that depend on a specific source
MATCH (m:Model)-[:DEPENDS_ON]->(s:Source {name: "raw_data.customers"})
RETURN m.name, m.materialized, m.description

// Get the complete downstream lineage from a model
MATCH path = (start:Model {name: "dim_customers"})-[:DEPENDS_ON*]->(downstream)
RETURN path

// Find models without any tests
MATCH (m:Model)
WHERE NOT EXISTS {
    MATCH (t:Test)-[:TESTS]->(m)
}
RETURN m.name, m.schema, m.materialized

// Identify the most referenced models
MATCH (m:Model)<-[:REFERENCES]-(referencing)
RETURN m.name, count(referencing) as reference_count
ORDER BY reference_count DESC
LIMIT 10

// Find macro usage patterns
MATCH (m:Model)-[:USES_MACRO]->(macro:Macro)
RETURN macro.name, count(m) as usage_count
ORDER BY usage_count DESC

// Discover circular dependencies (if any)
MATCH path = (n)-[:DEPENDS_ON*]->(n)
WHERE length(path) > 1
RETURN path

FalkorDB OpenCypher Examples

// Models by materialization type
MATCH (m:Model)
RETURN m.materialized, count(m) as model_count
ORDER BY model_count DESC

// Source freshness analysis
MATCH (s:Source)
WHERE s.freshness_warn_after IS NOT NULL
RETURN s.name, s.freshness_warn_after, s.freshness_error_after

// Test coverage by schema
MATCH (m:Model)
OPTIONAL MATCH (t:Test)-[:TESTS]->(m)
RETURN m.schema, 
       count(m) as total_models,
       count(t) as total_tests,
       round(100.0 * count(t) / count(m), 2) as test_coverage_pct
ORDER BY test_coverage_pct DESC

🐳 Docker Integration

FastAPI Integration Example

from fastapi import FastAPI, UploadFile, File
from dbt_graph_loader.loaders.neo4j_loader import DBTNeo4jLoader
import os

app = FastAPI()

@app.post("/upload-dbt-metadata/")
async def upload_dbt_metadata(
    manifest_file: UploadFile = File(...),
    catalog_file: UploadFile = File(...)
):
    manifest_content = await manifest_file.read()
    catalog_content = await catalog_file.read()
    
    loader = DBTNeo4jLoader(
        neo4j_uri=os.getenv("NEO4J_URI"),
        username=os.getenv("NEO4J_USERNAME"),
        password=os.getenv("NEO4J_PASSWORD")
    )
    
    try:
        loader.load_dbt_to_neo4j_from_strings(
            manifest_content.decode('utf-8'),
            catalog_content.decode('utf-8')
        )
        return {"status": "success", "message": "DBT metadata loaded"}
    finally:
        loader.close()

📊 Graph Schema

Node Properties

Models

• unique_id, name, database, schema, materialized
• description, tags, package_name, path, enabled
• language, checksum, access, relation_name

Sources

• unique_id, name, source_name, identifier
• database, schema, description, loader
• freshness_warn_after, freshness_error_after, columns

Tests

• unique_id, name, column_name, severity, enabled
• test_name, test_kwargs, package_name

Macros

• unique_id, name, package_name, path
• description, arguments

Seeds

• unique_id, name, database, schema, path
• delimiter, materialized, enabled

Snapshots

• unique_id, name, database, schema, strategy
• unique_key, updated_at, materialized

🧪 Development

Setup Development Environment

# Clone repository
git clone https://github.com/ponderedw/dbt-graph-loader.git

# Install dependencies
poetry install

# Build package
poetry build

📋 Prerequisites

For Neo4j

• Neo4j 4.0+ (local installation or cloud)
• Python 3.8+

For FalkorDB

• FalkorDB instance (Redis-compatible graph database)
• Python 3.8+

DBT Requirements

• DBT project with generated manifest.json (required)
• Generated catalog.json (optional but recommended for richer metadata)