https://github.com/azurecosmosdb/multi-agent-swarm
A sample exploring ergonomic, lightweight multi-agent orchestration in Python using Azure Cosmos DB with OpenAI Swarm
https://github.com/azurecosmosdb/multi-agent-swarm
Last synced: 12 months ago
JSON representation
A sample exploring ergonomic, lightweight multi-agent orchestration in Python using Azure Cosmos DB with OpenAI Swarm
- Host: GitHub
- URL: https://github.com/azurecosmosdb/multi-agent-swarm
- Owner: AzureCosmosDB
- License: mit
- Created: 2024-12-12T05:48:06.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-04-01T17:19:42.000Z (about 1 year ago)
- Last Synced: 2025-06-06T01:06:05.391Z (about 1 year ago)
- Language: Python
- Size: 29.5 MB
- Stars: 19
- Watchers: 2
- Forks: 11
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE.md
Awesome Lists containing this project
README
# Multi-agent AI sample with Azure Cosmos DB
A sample personal shopping AI Chatbot that can help with product enquiries, making sales, and refunding orders by transferring to different agents for those tasks.
Features:
- **Multi-agent**: [OpenAI Swarm](https://github.com/openai/swarm) to orchestrate multi-agent interactions with [Azure OpenAI](https://learn.microsoft.com/azure/ai-services/openai/overview) API calls.
- **Transactional data management**: planet scale [Azure Cosmos DB database service](https://learn.microsoft.com/azure/cosmos-db/introduction) to store transactional user and product operational data.
- **Retrieval Augmented Generation (RAG)**: [vector search](https://learn.microsoft.com/azure/cosmos-db/nosql/vector-search) in Azure Cosmos DB with powerful [DiskANN index](https://www.microsoft.com/en-us/research/publication/diskann-fast-accurate-billion-point-nearest-neighbor-search-on-a-single-node/?msockid=091c323873cd6bd6392120ac72e46a98) to serve product enquiries from the same database.
- **Gradio UI**: [Gradio](https://www.gradio.app/) to provide a simple UI ChatBot for the end-user.
## Backend agent activity
Run the CLI interactive session to see the agent handoffs in action...
## Front-end AI chat bot
Run the AI chat bot for the end-user experience...
## Overview
The personal shopper example includes four main agents to handle various customer service requests:
1. **Triage Agent**: Determines the type of request and transfers to the appropriate agent.
2. **Product Agent**: Answers customer queries from the products container using [Retrieval Augmented Generation (RAG)](https://learn.microsoft.com/azure/cosmos-db/gen-ai/rag).
3. **Refund Agent**: Manages customer refunds, requiring both user ID and item ID to initiate a refund.
4. **Sales Agent**: Handles actions related to placing orders, requiring both user ID and product ID to complete a purchase.
## Prerequisites
- [Azure Cosmos DB account](https://learn.microsoft.com/azure/cosmos-db/) - configured for Serverless and [vector search](https://learn.microsoft.com/azure/cosmos-db/nosql/vector-search).
- [Azure OpenAI Account](https://learn.microsoft.com/azure/ai-services/openai/overview) endpoint with embedding `text-3-embedding-large` and GPT `gpt-4o-mini` models deployed.
- [Azure Developer CLI - AZD](https://learn.microsoft.com/azure/developer/azure-developer-cli/overview) to deploy Azure Services and configure environment variables.
## How to run locally
Clone the repository:
```shell
git clone https://github.com/AzureCosmosDB/multi-agent-swarm
cd multi-agent-swarm
```
Create and activate a virtual environment (Linux/Mac):
```shell
python -m venv venv
source venv/bin/activate
```
```shell
# Install AZD
curl -fsSL https://aka.ms/install-azd.sh | bash
```
Create and activate a virtual environment (Windows):
```shell
python -m venv venv
venv\Scripts\activate
```
```shell
# Install AZD
powershell -ex AllSigned -c "Invoke-RestMethod 'https://aka.ms/install-azd.ps1' | Invoke-Expression"
```
Install dependencies:
```shell
pip install -r requirements.txt
```
Deploy the Azure Services via AZD and inject the service names into the .env file
```shell
azd up
```
From your terminal or IDE, run below and click on URL provided in output:
```shell
python src/app/ai_chat_bot.py
```
To see the agent handoffs, you can also run as an interactive Swarm CLI session using:
```shell
python src/app/multi_agent_service.py
```
There are three products you can ask about: a hat, wool socks and shoes. For more details on data like users and products in this sample, see the `initialize_database()` function in [azure_cosmos_db.py](./src/app/azure_cosmos_db.py)
Here are a series of user prompts you can enter to watch this multi-agent application in action. Enter these one at a time and watch how the app responds. Feel free to explore with your own prompts.
```text
I need some help
I'm looking for footwear
Do you have any shoes
I'd like to purchase these
My user id is 1 and product id is 9
These are the wrong size. Can i return them?
My user id is 1
```