Cosmos DB Embeddings Generator

This sample shows how to use a Cosmos DB Trigger and Output Binding in Azure Functions to automatically generate vector embeddings on new or updated data.

Details (more to come)

• This sample demonstrates in both C# and Python
• It uses the native Azure OpenAI SDK to generate embeddings
• To prevent the Cosmos DB Trigger from endless loops it generates a hash of the document and stores it with the generated embeddings. Then checks it when the trigger is called when initially generating and storing the vector.

To Do:

• azd deployment. Modify the parameters.bicep to create the account, database and container with a vector index and container policy. The output is also injected into the local.settings.json for both the C# and Python samples allowing you to F5 the sample immediately.

Getting Started:

Deployment

1. Open a terminal and navigate to where you would like to clone this solution

2. From the terminal, navigate to either the csharp or python directory in this solution.

3. Provision the Azure services, build your local solution container, and deploy the application.

   azd up

Take note of the value of OPENAI_ENDPOINT which can be found in ./.azure/<env name from azd provision>/.env. It will look something like:

OPENAI_ENDPOINT: "https://{my-open-ai-account}.openai.azure.com/"

1. Copy the sample.settings.json file into a new file local.settings.json in the same folder (csharp or python).

2. Replace the placeholder values like OPENAI_ENDPOINT with values from the previous step found in the .azure folder. By default the sample uses Entra identity (user identity and mananaged identity) so it is secretless.

Python:

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "UseDevelopmentStorage=true",
    "FUNCTIONS_WORKER_RUNTIME": "python",
    "COSMOS_CONNECTION__accountEndpoint": "https://{my-cosmos-account}.documents.azure.com:443/",
    "COSMOS_DATABASE_NAME": "{database-name}",
    "COSMOS_CONTAINER_NAME": "{container-name}",
    "COSMOS_OUTPUT_CONTAINER_NAME": "{container-name}",  
    "COSMOS_VECTOR_PROPERTY": "{vector-property-name}",
    "COSMOS_HASH_PROPERTY": "{hash-property-name}",
    "COSMOS_PROPERTY_TO_EMBED": "{property-to-generate-vectors-for}",
    "OPENAI_ENDPOINT": "https://{my-open-ai-account}.openai.azure.com/",
    "OPENAI_DEPLOYMENT_NAME": "{my-embeddings-deployment-name}",
    "OPENAI_DIMENSIONS": "1536"
  }
}

C# (.NET):

{
  "IsEncrypted": false,
  "Values": {
    "AzureWebJobsStorage": "UseDevelopmentStorage=true",
    "FUNCTIONS_WORKER_RUNTIME": "dotnet-isolated",
    "COSMOS_CONNECTION__accountEndpoint": "https://{my-cosmos-account}.documents.azure.com:443/",
    "COSMOS_DATABASE_NAME": "{database-name}",
    "COSMOS_CONTAINER_NAME": "{container-name}",
    "COSMOS_OUTPUT_CONTAINER_NAME": "{container-name}",  
    "COSMOS_VECTOR_PROPERTY": "{vector-property-name}",
    "COSMOS_HASH_PROPERTY": "{hash-property-name}",
    "COSMOS_PROPERTY_TO_EMBED": "{property-to-generate-vectors-for}",
    "OPENAI_ENDPOINT": "https://{my-open-ai-account}.openai.azure.com/",
    "OPENAI_DEPLOYMENT_NAME": "{my-embeddings-deployment-name}",
    "OPENAI_DIMENSIONS": "1536"
  }
}

• F5
• Open a browser and navigate to the container in Cosmos Data Explorer
• Create a new document and save. """

The expected document/item has at least these 3 properties, and note that 'text' is the property that gets embedded.

Example document:

{
   "id": "cosmosdb_overview_1",
   "customerId": "1",
   "text": "Azure Cosmos DB is a fully managed NoSQL, relational, and vector database. It offers single-digit millisecond response times, automatic and instant scalability, along with guaranteed speed at any scale. Business continuity is assured with SLA-backed availability and enterprise-grade security."
}