ContextQMD
Back to Datahub

DataHub Agent Context

datahub-agent-context/README.md

1.5.0.34.7 KB
Original Source

DataHub Agent Context

DataHub Agent Context provides a collection of tools and utilities for building AI agents that interact with DataHub metadata. This package contains MCP (Model Context Protocol) tools that enable AI agents to search, retrieve, and manipulate metadata in DataHub. These can be used directly to create an agent, or be included in an MCP server such as Datahub's open source MCP server.

Features

Installation

Base Installation

shell
python3 -m pip install --upgrade pip wheel setuptools
python3 -m pip install --upgrade datahub-agent-context

With LangChain Support

For building LangChain agents with pre-built tools:

shell
python3 -m pip install --upgrade "datahub-agent-context[langchain]"

Prerequisites

This package requires:

  • Python 3.9 or higher
  • acryl-datahub package

Quick Start

Basic Example

These tools are designed to be used with an AI agent and have the responses passed directly to an LLM, so the return schema is a simple dict, but they can be used independently if desired.

python
from datahub.ingestion.graph.client import DataHubGraph
from datahub_agent_context.mcp_tools.search import search
from datahub_agent_context.mcp_tools.entities import get_entities

# Initialize DataHub graph client
client = DataHubClient.from_env()

# Search for datasets
with client as client:
    results = search(
        query="user_data",
        filters={"entity_type": ["dataset"]},
        num_results=10
    )

# Get detailed entity information
with client as client:
    entities = get_entities(
        urns=[result["entity"]["urn"] for result in results["searchResults"]]
    )

LangChain Integration

Build AI agents with pre-built LangChain tools:

python
from datahub.sdk.main_client import DataHubClient
from datahub_agent_context.langchain_tools import build_langchain_tools
from langchain.agents import create_agent

# Initialize DataHub client
client = DataHubClient.from_env()

# Build all tools (read-only by default)
tools = build_langchain_tools(client, include_mutations=False)

# Or include mutation tools for tagging, descriptions, etc.
tools = build_langchain_tools(client, include_mutations=True)

# Create agent
agent = create_agent(model, tools=tools, system_prompt="...")

See examples/langchain/ for complete LangChain agent examples including:

Available Tools

Search Tools

  • search() - Search across all entity types with filters and sorting
  • search_documents() - Search specifically for Document entities
  • grep_documents() - Grep for patterns in document content

Entity Tools

  • get_entities() - Get detailed information about entities by URN
  • list_schema_fields() - List and filter schema fields for datasets

Lineage Tools

  • get_lineage() - Get upstream or downstream lineage
  • get_lineage_paths_between() - Get detailed paths between two entities

Query Tools

  • get_dataset_queries() - Get SQL queries for datasets or columns

Mutation Tools

  • add_tags(), remove_tags() - Manage tags
  • update_description() - Update entity descriptions
  • set_domains(), remove_domains() - Manage domains
  • add_owners(), remove_owners() - Manage owners
  • add_glossary_terms(), remove_glossary_terms() - Manage glossary terms
  • add_structured_properties(), remove_structured_properties() - Manage structured properties
  • save_document() - Save or update a Document.

User Tools

  • get_me() - Get information about the authenticated user

Architecture

The package is organized into the following modules:

  • mcp_tools/ - Core MCP tool implementations
    • base.py - Base GraphQL execution and response cleaning
    • search.py - Search functionality
    • documents.py - Document search and grep
    • entities.py - Entity retrieval
    • lineage.py - Lineage querying
    • queries.py - Query retrieval
    • tags.py, descriptions.py, domains.py, etc. - Mutation tools
    • helpers.py - Shared utility functions
    • gql/ - GraphQL query definitions

Development

Setup

shell
# Clone the repository
git clone https://github.com/datahub-project/datahub.git
cd datahub/datahub-agent-context

# Set up development environment
./gradlew :datahub-agent-context:installDev

# Run tests
./gradlew :datahub-agent-context:testFull

# Run linting
./gradlew :datahub-agent-context:lintFix

Testing

The package includes comprehensive unit tests for all tools:

shell
# Run full test suite
./gradlew :datahub-agent-context:testFull

Support