Contoso Chat With Assistants API

This project uses the Azure OpenAI Assistants API to create a chatbot that interacts with users, processes their messages, and performs actions based on the content of the messages.

The project tries to implement the same functionality as the contoso-chat project but it uses Assistants API instead of Prompt Flow

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

• Python 3.9 or higher
• Azure OpenAI resource and a GPT model deployment
• Azure Functions Core Tools
• Azure CosmosDB resource
• Visual Studio Code with Azure Functions extension

Installation

1. Clone the repo

   git clone https://github.com/dfmera/contoso-chat-assistant.git

2. Open project in Visual Studio Code

   cd contoso-chat-assistant
   code .
   

3. Create a Python virtual enviroment
   • 3.1. for Mac OS / Linux

     python3 -m venv .venv
     source .venv/bin/activate
     

   • 3.2. for Windows

     python3 -m venv .venv
     .venv/Scripts/Activate.ps1
     

4. Install Python packages

   pip install -r requirements.txt
   

Create an assistant in Azure OpenAI Studio

• Open Azure OpenAI Studio and go to Assistants (preview)

• Give your assistant a name

• Instructions: Copy the instructions in assistant/customer_prompt.txt

• Deployment: Select a GPT model deployment

• Functions: Copy the function definition in assistant/GetCustomerInfo_definition.json

• Activate code interpreter

• Add the file data/product_info/products.csv

• Save the assistant and copy the assistant id

  

Run Azure Functions API

1. Create a .env file and fill it with the next values

   COSMOS_ENDPOINT=
   COSMOS_KEY=
   AZURE_OPENAI_ENDPOINT=
   AZURE_OPENAI_API_KEY=
   OPENAI_API_VERSION=2024-02-15-preview
   OPENAI_GPT_DEPLOYMENT=
   OPENAI_ASSISTANT_ID=
   

2. Create a CosmosDB database and container by running the notebook in

   /data/customer_info/create-cosmos-db.ipynb
   

   Make sure you run the notebook in the .venv you created

3. Open folder api in VS Code and initiate the Azure Functions project or create a python .venv virtual enviroment

4. Create a local.settings.json file and fill it with the next values

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "",
    "FUNCTIONS_WORKER_RUNTIME": "python",
    "CosmosDB": "<YOUR COSMOSDB CONNECTION STRING>",
    "OPENAI_ASSISTANT_ID": "<YOUR OPENAI_ASSISTANT_ID>",
    "AZURE_OPENAI_ENDPOINT": "<YOUR AZURE_OPENAI_ENDPOINT>",
    "AZURE_OPENAI_API_KEY": "<YOUR AZURE_OPENAI_API_KEY>",
    "OPENAI_API_VERSION": "2024-02-15-preview",
    "OPENAI_GPT_DEPLOYMENT": "<YOUR OPENAI_GPT_DEPLOYMENT>",
    "CUSTOMER_INFO_API": "<YOUR LOCAL GetCustomerInfo FUNCTION URL>"
  }
}

5. Run the Azure Functions API in VS Code or in a terminal

   cd api
   func start

Test Assistants API

1. Test the function ContosoChatAssistant as in this example (replace {port} with your local port):

   POST http://localhost:{port}/api/ContosoChatAssistant
   Content-Type: application/json
   
   {
       "customerId": 1,
       "question": "Can you remember me my last orders?",
       "chat_history": []
   }
   

Test Assistants API in Contose Chat Web

1. Clone contoso-web repository

2. Follow the instructions to run the project localy

3. Replace the value of PROMPTFLOW_ENDPOINT key in .env file with your ContosoChatAssistant local URL

   PROMPTFLOW_ENDPOINT=http://{local_url}/api/ContosoChatAssistant
   

4. Run the project and test the chat

   

To do

1. Automate resource creation with azd up.
2. Test the reading of the products.csv file for product query

Contributions are welcome!