Python Packages

TrustGraph is distributed as a collection of Python packages available on PyPI. Each package provides specific functionality while maintaining minimal dependencies to avoid large installation footprints.

Installation

All packages are available on PyPI and can be installed using pip:

pip install trustgraph-base
pip install trustgraph-flow
pip install trustgraph-cli
# ... and so on

Note that all packages depend on trustgraph-base as a dependency so there is no need to install it unless you are using that package in isolation.

Version strategy

The TrustGraph release process creates containers and packages with the same version number. For best results, match the container and package version numbers.

Package Structure

Package	Description	Major Dependencies	Scripts
`trustgraph-base`	Minimal base classes and API support	`pulsar-client`, `prometheus-client`	None
`trustgraph-flow`	Core AI processing pipeline capabilities	`anthropic`, `openai`, `langchain`, `neo4j`, `milvus`, `pinecone`, `qdrant`, `fastembed`, `ollama`, `cohere`, `mistralai`	60+ processing scripts
`trustgraph-cli`	Command-line interface for client-side operations	`requests`, `aiohttp`, `rdflib`, `tabulate`, `websockets`	49 `tg-*` CLI commands
`trustgraph-embeddings-hf`	HuggingFace embeddings support	`torch`, `transformers`, `sentence-transformers`, `huggingface-hub`	`embeddings-hf`
`trustgraph-bedrock`	AWS Bedrock integration	`boto3`	`text-completion-bedrock`
`trustgraph-vertexai`	Google Vertex AI integration	`google-cloud-aiplatform`	`text-completion-vertexai`
`trustgraph-ocr`	OCR processing capabilities	`boto3`, `pdf2image`, `pytesseract`	`pdf-ocr`
`trustgraph-mcp`	Model Context Protocol server	`mcp`, `websockets`	`mcp-server`

Package Dependencies

Core Dependencies

trustgraph-base: Minimal foundation with messaging and metrics
trustgraph-flow: Full AI pipeline with extensive ML/AI library support
trustgraph-cli: Client tools with web and RDF support

Specialized Dependencies

embeddings-hf: PyTorch and HuggingFace ecosystem
bedrock/vertexai: Cloud provider SDKs
ocr: Image processing and OCR libraries
mcp: Model Context Protocol implementation

Python Requirements

All packages require Python 3.8 or higher and are licensed under GPLv3+.

Architecture

The modular design allows users to install only the components they need:

Install trustgraph-base for basic API integration
Add trustgraph-flow for full AI processing capabilities
Include trustgraph-cli for command-line management
Add specialized packages (embeddings, cloud providers, OCR) as needed