This demo originally started out using the code from the Azure AI Agent QuickStart, but has expanded to show different agent tooling features:
- Perform a simple calculation (no tools)
- Make function calls using Function Calling Tools
- Create a function using code interpreter
- Examine a CSV file and create diagrams from it using code interpreter
- Examine a CSV file and provide analysis using code interpreter
- Perform RAG functionality using AI search
- Use real-time public web data using Bing Grounding
- Return usage stats about in/out tokens
To use the demo you'll need to create an Azure AI Foundry project by following the QuickStart.
If you'd like to use the AI Search/RAG functionality in the demo, you'll find details about the setup in the AI Search tooling document. You'll need to:
- Add the health documents in
files
to an Azure Blob Storage container. - Import the health documents into an Azure AI Search index using the Azure AI Search Importer.
- Create a connection to the AI Search resource in your AI Foundry project. See Use an existing AI Search index with the Azure AI Search tool.
If you'd like to use the Bing Grounding functionality in the demo, you'll find details about the setup in the Bing Grounding tooling document.
After going through the QuickStart steps (and optionally the AI Search and Bing Grounding tooling setup), perform the following steps:
-
Rename
.env.template
to.env
. -
Assign your Azure AI Foundry endpoint, your AI Search connection ID, the AI Search index name, and the Bing Grounding connection ID from Azure AI Foundry to the appropriate env keys. You can optionally change the model deployment name.
AI_FOUNDRY_PROJECT_ENDPOINT=https://<your-endpoint>.services.ai.azure.com/api/projects/<your-project> AI_SEARCH_CONNECTION_ID=/subscriptions/<subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.CognitiveServices/accounts/<account-name>/projects/<project-name>/connections/<connection-name> AI_SEARCH_INDEX_NAME=<your-index-name> BING_GROUNDING_CONNECTION_ID=/subscriptions/<subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.CognitiveServices/accounts/<account-name>/projects/<project-name>/connections/<connection-name>
-
Install the project dependencies:
npm install
-
Start the demo:
npm start
-
Select an agent demo to run.