Glean Agent Examples

Example implementations showing how to use Glean to access your company's knowledge base with common AI agent frameworks.

Overview

This repository demonstrates how to integrate Glean with different AI agent frameworks. These examples show how to:

• Search and retrieve information from your Glean knowledge base
• Use Glean's chat capabilities for natural language interactions

Available Integrations

LangChain

Uses LangChain's built-in Glean integrations:

• GleanSearchRetriever for semantic search over your knowledge base
• ChatGlean for conversational interactions
• Example shows both standalone usage and as part of an agent

LangGraph

Shows how to build stateful agents with Glean:

• Maintains conversation context across interactions
• Demonstrates workflow control with Glean's search and chat
• Handles complex multi-step queries

OpenAI Assistants

Shows how to use Glean with OpenAI's Assistant framework:

• Uses Glean's MCP server to expose search and chat as Assistant tools
• Demonstrates function calling with Glean's capabilities
• Shows how to maintain context in conversations

Getting Started

In order to setup and run these examples, follow these steps:

1. Prerequisites

   • Python 3.13+
   • Glean API credentials (subdomain and API token with the chat and search scopes)
   • OpenAI API key (for OpenAI and LangChain examples)

2. Installation

   • install Go-Task:

     # macOS with Homebrew
     brew install go-task/tap/go-task
     
     # Linux/macOS with curl
     sh -c "$(curl --location https://taskfile.dev/install.sh)" -- -d
     
     # Windows with Scoop
     scoop install task

   • install uv

     # macOS with Homebrew
     brew install uv
     
     # Linux/macOS with curl
     curl -LsSf https://astral.sh/uv/install.sh | sh
     
     # Windows
     powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

   • Then install the package and set up your environment:

     # install dependencies
     task install
     
     # set up environment variables
     
     export GLEAN_SUBDOMAIN=your-instance
     export GLEAN_API_TOKEN=your_token
     export OPENAI_API_KEY=your_key  # If using OpenAI examples

3. Running Examples

   First, start the server:

   # Start the server
   task serve

   Then in a new terminal, you can run examples:

   # List available examples
   task list
   
   # Run a specific example
   task run -- langchain_search "What are our company policies?"

   The server will handle requests from multiple examples, so you only need to start it once.

Architecture

Each example follows a consistent pattern:

1. Authenticates with Glean using your credentials
2. Sets up the agent framework with appropriate tools/capabilities
3. Handles queries by searching or chatting through Glean
4. Returns formatted responses

The examples are structured to be:

• Easy to understand and modify
• Production-ready with proper error handling
• Maintainable with clear separation of concerns

API Documentation

• Glean API Documentation
• LangChain Glean Integration
• OpenAI Assistants API