TrustGraph is a fully agentic AI data engineering platform for complex unstructured data. Extract your documents to knowledge graphs and vector embeddings with customizable data extraction agents. Deploy AI agents that leverage your data to generate explainable AI responses.

• 📄 Document Extraction: Bulk ingest documents such as .pdf,.txt, and .md
• 🪓 Adjustable Chunking: Choose your chunking algorithm and parameters
• 🔁 No-code LLM Integration: Anthropic, AWS Bedrock, AzureAI, AzureOpenAI, Cohere, Google AI Studio, Google VertexAI, Llamafiles, Ollama, and OpenAI
• 📖 Entity, Topic, and Relationship Knowledge Graphs
• 🔢 Mapped Vector Embeddings
• ❔No-code GraphRAG Queries: Automatically perform a semantic similiarity search and subgraph extraction for the context of LLM generative responses
• 🤖 Agent Flow: Define custom tools used by a ReAct style Agent Manager that fully controls the response flow including the ability to perform GraphRAG requests
• 🎛️ Production-Grade reliability, scalability, and accuracy
• 🔍 Observability: get insights into system performance with Prometheus and Grafana
• 🗄️ AI Powered Data Warehouse: Load only the subgraph and vector embeddings you use most often
• 🪴 Customizable and Extensible: Tailor for your data and use cases
• 🖥️ Configuration UI: Build the YAML configuration with drop down menus and selectable parameters

There are four ways of interacting with TrustGraph:

• REST API
• Python SDK
• TrustGraph CLI
• Configuration UI

The TrustGraph CLI installs the commands for interacting with TrustGraph while running along with the Python SDK. The Configuration UI enables customization of TrustGraph deployments prior to launching. The REST API can be accessed through port 8088 of the TrustGraph host machine with JSON request and response bodies.

pip3 install trustgraph-cli==0.15.6

TrustGraph is endlessly customizable by editing the YAML launch files. The Configuration UI provides a quick and intuitive tool for building a custom configuration that deploys with Docker, Podman, Minikube, or Google Cloud. There is a Configuration UI for the both the lastest and stable TrustGraph releases.

• Stable Configuration UI (0.15.6) 🚀
• Latest Configuration UI (0.16.5) 🚀

The Configuration UI has three sections:

• Component Selection ✅: Choose from the available deployment platforms, LLMs, graph store, VectorDB, chunking algorithm, chunking parameters, and LLM parameters
• Customization 🧰: Customize the prompts for the LLM System, Data Extraction Agents, and Agent Flow
• Finish Deployment 🚀: Download the launch YAML files with deployment instructions

The Configuration UI will generate the YAML files in deploy.zip. Once deploy.zip has been downloaded and unzipped, launching TrustGraph is as simple as navigating to the deploy directory and running:

docker compose up -d

When finished, shutting down TrustGraph is as simple as:

docker compose down -v

• REST API Notebooks
• Python SDK Notebooks

TrustGraph releases are available here. Download deploy.zip for the desired release version.

TrustGraph is fully containerized and is launched with a YAML configuration file. Unzipping the deploy.zip will add the deploy directory with the following subdirectories:

• docker-compose
• minikube-k8s
• gcp-k8s

Each directory contains the pre-built YAML configuration files needed to launch TrustGraph:

Once a configuration launch file has been selected, deploy TrustGraph with:

Docker:

docker compose -f <launch-file.yaml> up -d

Kubernetes:

kubectl apply -f <launch-file.yaml>

TrustGraph is designed to be modular to support as many LLMs and environments as possible. A natural fit for a modular architecture is to decompose functions into a set of modules connected through a pub/sub backbone. Apache Pulsar serves as this pub/sub backbone. Pulsar acts as the data broker managing data processing queues connected to procesing modules.

• For processing flows, Pulsar accepts the output of a processing module and queues it for input to the next subscribed module.
• For services such as LLMs and embeddings, Pulsar provides a client/server model. A Pulsar queue is used as the input to the service. When processed, the output is then delivered to a separate queue where a client subscriber can request that output.

TrustGraph extracts knowledge documents to an ultra-dense knowledge graph using 3 automonous data extraction agents. These agents focus on individual elements needed to build the knowledge graph. The agents are:

• Topic Extraction Agent
• Entity Extraction Agent
• Relationship Extraction Agent

The agent prompts are built through templates, enabling customized data extraction agents for a specific use case. The data extraction agents are launched automatically with the loader commands.

PDF file:

tg-load-pdf <document.pdf>

Text or Markdown file:

tg-load-text <document.txt>

Once the knowledge graph and embeddings have been built or a knowledge core has been loaded, RAG queries are launched with a single line:

tg-query-graph-rag -q "Write a blog post about the 5 key takeaways from SB1047 and how they will impact AI development."

Invoking the Agent Flow will use a ReAct style approach the combines GraphRAG and text completion requests to think through a problem solution.

tg-invoke-agent -v -q "Write a blog post about the 5 key takeaways from SB1047 and how they will impact AI development."

🚀 Full Deployment Guide 🚀

Developing for TrustGraph