Julep Assistant

An intelligent AI support assistant built with the Julep platform to help developers understand and use Julep effectively. This project demonstrates best practices for building AI applications with Julep, including web crawling, document indexing, and conversational AI capabilities.

Features

• Intelligent Documentation Crawling: Automatically crawls and indexes Julep documentation
• RAG-powered Conversations: Uses Retrieval-Augmented Generation for accurate, context-aware responses
• Interactive Chat Interface: Built with Chainlit for a smooth user experience
• Feedback System: Collects and validates user feedback for continuous improvement
• Workflow Examples: Demonstrates complex Julep workflows including web crawling and document processing

Project Structure

julep-assistant/
├── agent.yaml              # Agent configuration (name, instructions, model)
├── task/                   # Julep task definitions
│   ├── main.yaml          # Main workflow task
│   ├── crawl.yaml         # Web crawling sub-task
│   └── full_task.yaml     # Complete task with all steps
├── scripts/               # Utility scripts
│   ├── crawler.py         # Standalone web crawler
│   └── indexer.py         # Document indexing utility
├── chainlit-ui/           # Web interface
│   ├── app.py            # Main Chainlit application
│   ├── feedback/         # Feedback handling system
│   └── requirements.txt  # Python dependencies
└── julep-assistant-notebook.ipynb  # Interactive notebook demo

Prerequisites

• Python 3.8+
• Julep API key (get one at platform.julep.ai)
• Spider API key for web crawling

Installation

1. Clone the repository:

git clone https://github.com/yourusername/julep-assistant.git
cd julep-assistant

2. Install dependencies:

pip install -r chainlit-ui/requirements.txt

3. Set up environment variables:

# Create a .env file in the project root
cp .env.example .env

# Edit .env and add your API keys:
# JULEP_API_KEY=your_julep_api_key_here
# AGENT_UUID=ce7be83e-db8b-4ba9-808e-7cade6812e98  # Or create your own
# SPIDER_API_KEY=your_spider_api_key_here  # Optional, for web crawling

Usage

Running the Chat Interface

cd chainlit-ui
chainlit run app.py

This will start the web interface at http://localhost:8000

Using the Jupyter Notebook

Open julep-assistant-notebook.ipynb to explore:

• Creating and configuring Julep agents
• Defining and executing tasks
• Setting up RAG-powered conversations
• Monitoring task executions

Running Scripts Directly

Web Crawler:

python scripts/crawler.py --url https://docs.julep.ai --max-pages 100

Document Indexer:

python scripts/indexer.py

How It Works

1. Agent Configuration: The assistant is configured with specific instructions and knowledge about Julep
2. Document Indexing: Crawls and indexes Julep documentation for RAG
3. Hybrid Search: Uses both vector and text search for optimal retrieval
4. Contextual Responses: Generates accurate answers based on retrieved documentation

Key Components

Agent (agent.yaml)

• Defines the assistant's personality, capabilities, and instructions
• Uses Claude Sonnet 3.5 model for high-quality responses

Tasks (task/*.yaml)

• main.yaml: Entry point workflow
• crawl.yaml: Web crawling sub-task with Spider integration
• full_task.yaml: Complete workflow including crawling and indexing

Chainlit UI

• Interactive chat interface
• Session management
• Feedback collection and validation
• Real-time streaming responses

Environment Variables

Variable	Description	Required
JULEP_API_KEY	Your Julep API key	Yes
AGENT_UUID	UUID for the Julep agent	No (uses default)
JULEP_ENV	Julep environment (production/development)	No (defaults to production)
SPIDER_API_KEY	Spider API key for web crawling	No (only for crawling tasks)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Built with Julep AI - The platform for building stateful AI applications
• UI powered by Chainlit
• Web crawling by Spider

Support

For questions about this project, please open an issue on GitHub.

For Julep-specific questions, visit the Julep Documentation or join the Julep Discord.