CrewAI Demo

Note

This repository contains a demo of using CrewAI for content generation with human-in-the-loop capabilities, along with an API wrapper to serve the CrewAI application.

Table of Contents

• Overview
• Project Structure
• Components
• Getting Started
• Documentation
• LLM Provider Support
• License

Overview

This project demonstrates how to build AI agent workflows using the CrewAI framework and expose them as a RESTful API. Key features include:

• Human-in-the-Loop (HITL) approval processes
• Asynchronous execution of AI agent workflows
• Webhook notifications for job status updates
• Support for multiple LLM providers (OpenRouter, Azure OpenAI)

Project Structure

Click to see the project structure

crewai_demo/                  # Root project directory
│
├── crewai_app/               # CrewAI application
│   ├── __init__.py           # Package initialization
│   └── orchestrator.py       # Main orchestrator for CrewAI using @CrewBase pattern
│
├── api_wrapper/              # API wrapper service
│   ├── __init__.py           # Package initialization
│   ├── api_wrapper.py        # Main API service
│   └── api_client.py         # Client for the API
│
├── docs/                     # Documentation
│   ├── api_wrapper_documentation.md
│   ├── azure_openai_integration.md
│   ├── HITL_WORKFLOW.md
│   ├── HITL_IMPLEMENTATION.md
│   └── curl_examples.md
│
├── scripts/                  # Utility scripts
│   ├── run_api.sh            # Script to run the API service
│   ├── check_jobs.sh         # Script to check job status
│   ├── curl_workflow.sh      # Example curl commands for workflows
│   ├── direct_mode_example.sh # Example for direct mode
│   ├── hitl_mode_example.sh  # Example for HITL mode
│   └── webhook_receiver.py   # Simple webhook receiver for testing
│
├── .env.sample              # Environment variable template
├── .env.azure.sample        # Azure-specific environment template
└── requirements.txt         # Project dependencies

Components

CrewAI Application

The crewai_app directory contains the core CrewAI application, which is responsible for content generation using AI agents. The main component is:

• orchestrator.py: The main orchestrator that defines the CrewAI crew, agents, and tasks using the @CrewBase decorator pattern. All CrewAI functionality is consolidated in this file.

API Wrapper

The api_wrapper directory contains the API service that wraps the CrewAI application and exposes it via a REST API. The main components are:

• api_wrapper.py: The main FastAPI application that provides endpoints for interacting with the CrewAI application.
• api_client.py: A client for the API that can be used to interact with the API programmatically.

Documentation

The project includes comprehensive documentation in the docs/ directory:

• API Wrapper Documentation - How to use the API wrapper
• Azure OpenAI Integration - Using Azure OpenAI with the application
• HITL Workflow - Human-in-the-Loop workflow guide
• HITL Implementation - Implementation details for HITL
• cURL Examples - Example API requests using curl

Scripts

The scripts directory contains utility scripts for working with the application:

• run_api.sh - Script to start the API server
• check_jobs.sh - Script to check the status of jobs
• curl_workflow.sh - Example workflow using curl commands
• direct_mode_example.sh - Example of using direct mode
• hitl_mode_example.sh - Example of using HITL mode
• webhook_receiver.py - Simple webhook receiver for testing notifications

Getting Started

Click to expand setup instructions

1. Clone the repository

   git clone https://github.com/yourusername/crewai_demo.git
   cd crewai_demo

2. Install dependencies

   pip install -r requirements.txt

3. Configure environment variables

   cp .env.sample .env
   # Edit .env with your API keys and configuration

4. Run the API wrapper

   bash scripts/run_api.sh
   # Or directly with uvicorn:
   # uvicorn api_wrapper.api_wrapper:app --host 0.0.0.0 --port 8888 --reload

5. Make a test request

   curl -X POST http://localhost:8888/kickoff \
     -H "Content-Type: application/json" \
     -d '{
       "crew": "ContentCreationCrew",
       "inputs": {
         "topic": "Artificial Intelligence"
       }
     }'

   For more examples, see the scripts in the scripts/ directory or the cURL Examples documentation.

LLM Provider Support

Important

The application supports multiple LLM providers. Make sure to configure the appropriate environment variables.

OpenRouter Configuration (Default)

# In .env file
LLM_PROVIDER=openrouter
OPENAI_API_KEY=your_api_key_here
OPENAI_API_BASE=https://openrouter.ai/api/v1
OPENROUTER_MODEL=openai/gpt-4o-mini

Note: OpenRouter uses the OPENAI_API_KEY environment variable for authentication.

Azure OpenAI Configuration

# In .env file
LLM_PROVIDER=azure
AZURE_OPENAI_API_KEY=your_azure_api_key_here
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
AZURE_OPENAI_API_VERSION=2023-05-15
AZURE_OPENAI_DEPLOYMENT_ID=gpt-35-turbo-0125

See the Azure OpenAI Integration documentation for more details.

License

MIT License