ContextClue Graph Builder is an open-source toolkit for extracting knowledge graphs from semi-structured and unstructured data such as PDFs, reports, and tabular files.

It enables engineers, businesses, researchers, and developers to transform raw documents into graph structures for analytics, search, chatbots, and digital twin applications.

📄 Document → Graph Extract tabular information from documents and load it into graph structures.

⚙️ Flexible Configuration Define headers, file paths, entity labels, and relationship types.

🚀 FastAPI Backend Deploy graph extraction as a REST API service (Docker-ready).

🔄 Runtime Graph Retention Graphs persist between API calls while the service is running.

🔮 Future Roadmap

• Smarter chunking & embeddings
• Integration with graph DBs and vector DBs
• Relationship discovery across multiple data sources
• Knowledge graph visualization dashboards

💼 Business Use Cases

ContextClue Graph Builder goes beyond raw graph extraction—it powers enterprise-grade knowledge systems.

1. Industrial Engineering & Manufacturing

• Convert CAD, ERP, PLM, and planning data into unified, searchable knowledge graphs.
• Enable digital twin navigation: interactive exploration of components, processes, and relationships.
• Provide graph-based operational intelligence for predictive performance and system optimization.

2. Maintenance, Repair & Operations (MRO)

• Automotive, aerospace, energy, and logistics sectors use ContextClue to:
  • Reduce downtime with faster diagnostics.
  • Support predictive maintenance.
  • Increase efficiency of maintenance workflows.

3. Knowledge Assistants

• Integrate with chat platforms like Slack to build internal assistants.
• Example: Addeptalk (powered by ContextClue) connects Google Drive docs to Slack, enabling employees to ask natural-language questions and receive contextual answers.

4. Domain-Specific Applications

• Finance & Legal: Compliance document automation, audit preparation.
• Healthcare & Research: Extract structured knowledge from scientific papers and clinical reports.
• Developers & IT: Summarize technical docs, generate structured code knowledge, power RAG-based bots.

Following good practices we suggest you create a separate virtual environment for working with Graph builder package.

Note that the graph_builder requires python 3.12 or higher.

poetry install

Note that this is a short example taken from examples/example1.ipynb, for more information please refer to it.

from entity_graph.graph_extractor.entities_graph_extractor import EntitiesGraphExtractor

# Initialize the Extractor
extractor = EntitiesGraphExtractor()

test_data_path = "" # replace it with a correct data path

# Specify extraction configuration
config = {
    "extraction_type": "table_from_header",
    "filename": test_data_path + "coffee_machines.pdf",
    "header": ['Manufacturer', 'Coffee Machine Name', 'Machine ID', 'Production Year', 'Machine Type', 'Power (W)', 'Pressure (bar)', 'Water Tank Capacity (L)', 'Additional Features'],
}

# Load table to graph
extractor.load_table_from_file(
    config,
    "coffee_machines.pdf",
    "Machines",
    "instances",
)

In the folder containing the docker-compose.yml file, run the commands:

docker compose build

Once the image is built:

docker compose up

Make sure to create the .env file in the directory based on the .env_example file with the needed environmental values.

In the current version of the application, graphs are retained between requests but not preserved across API restarts.

This means that each time the API is restarted, the graphs must be rebuilt.

📌 Roadmap

• Improved data chunking and embeddings
• Database and vector database infrastructure
• Advanced relational analysis between sources
• Interactive knowledge graph visualization

We welcome community contributions!

_Fork this repo

Create a branch (feature/my-feature)

Commit changes (git commit -m "Add feature")

Push branch (git push origin feature/my-feature)

Open a Pull Request 🎉

Please include tests for new functionality._

Integrated chatbot with RAG