**TrustGraph streamlines the delivery and management of complex AI environments, acting as a comprehensive provisioning platform for your containerized AI tools, pipelines, and integrations.**
Deploying state-of-the-art AI requires managing a complex web of models, frameworks, data pipelines, and monitoring tools. TrustGraph simplifies this by providing a unified, open-source solution to **provision complete, trusted AI environments** anywhere you need them – from cloud instances and on-premises servers to edge devices.
***Unified Provisioning:** Define and deploy complete AI environments, including models, dependencies, and tooling, as a single, manageable unit. Stop managing piecemeal installations.
***No-code TrustRAG Pipelines:** Deploy full end-to-end RAG pipelines using unique TrustGraph algorithms leveraging both Knowledge graphs and VectorDBs.
***Environment-Agnostic Deployment:** Provision consistently across diverse infrastructures (Cloud, On-Prem, Edge, Dev environments). Build once, provision anywhere.
***Trusted & Secure Delivery:** Focuses on providing a secure supply chain for AI components.
***Simplified Operations:** Radically reduce the complexity and time required to stand up and manage sophisticated AI stacks. Get operational faster.
***Open Source & Extensible:** Built with transparency and community collaboration in mind. Easily inspect, modify, and extend the platform to meet your specific provisioning needs.
***Component Flexibility:** Avoid component lock-in. TrustGraph integrates multiple options for all system components.
The `TrustGraph CLI` installs the commands for interacting with TrustGraph while running along with the Python SDK. The `Configuration Builder` 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.
TrustGraph is endlessly customizable by editing the `YAML` launch files. The `Configuration Builder` provides a quick and intuitive tool for building a custom configuration that deploys with Docker, Podman, Minikube, AWS, Azure, Google Cloud, or Scaleway. There is a `Configuration Builder` for the both the lastest and stable `TrustGraph` releases.
The `Configuration Builder` 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:
The `-v` flag will destroy all data on shut down. To restart the system, it's necessary to keep the volumes. To keep the volumes, shut down without the `-v` flag:
```
docker compose down
```
With the volumes preserved, restarting the system is as simple as:
If added to the build in the `Configuration Builder`, the `Test Suite` will be available at port `8888`. The `Test Suite` has the following capabilities:
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:
> As more integrations have been added, the number of possible combinations of configurations has become quite large. It is recommended to use the `Configuration Builder` to build your deployment configuration. Each directory contains `YAML` configuration files for the default component selections.
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](https://github.com/apache/pulsar/) serves as this pub/sub backbone. Pulsar acts as the data broker managing data processing queues connected to procesing modules.
TrustGraph incorporates **TrustRAG**, an advanced RAG approach that leverages automatically constructed Knowledge Graphs to provide richer and more accurate context to LLMs. Instead of relying solely on unstructured text chunks, TrustRAG understands and utilizes the relationships *between* pieces of information.
**How TrustRAG Works:**
1.**Automated Knowledge Graph Construction:**
* TrustGraph processes source data to automatically **extract key entities, topics, and the relationships** connecting them.
* It then maps these extracted **semantic relationships and concepts to high-dimensional vector embeddings**, capturing the nuanced meaning beyond simple keyword matching.
2.**Hybrid Retrieval Process:**
* When a query is received, TrustRAG first performs a **cosine similarity search** on the vector embeddings to identify potentially relevant concepts and relationships within the knowledge graph.
* This initial vector search **pinpoints relevant entry points** within the structured Knowledge Graph.
3.**Context Generation via Subgraph Traversal:**
* Based on the ranked results from the similarity search, TrustRAG dynamically **generates relevant subgraphs**.
* It starts from the identified entry points and traverses the connections within the Knowledge Graph. Users can configure the **number of 'hops'** (relationship traversals) to expand the contextual window, gathering interconnected information.
* This structured **subgraph**, containing entities and their relationships, forms a highly relevant and context-aware input prompt for the LLM that is endlessly configurable with options for the number of entities, relationships, and overall subgraph size.
One of the biggest challenges currently facing RAG architectures is the ability to quickly reuse and integrate knowledge sets. **TrustGraph** solves this problem by storing the results of the document ingestion process in reusable Knowledge Cores. Being able to store and reuse the Knowledge Cores means the process has to be run only once for a set of documents. These reusable Knowledge Cores can be loaded back into **TrustGraph** and used for TrustRAG.
When a Knowledge Core is loaded into TrustGraph, the corresponding graph edges and vector embeddings are queued and loaded into the chosen graph and vector stores.
- 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:
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.
