Configure the Azure OpenAI Topic Splitter tuning stage

Configure the Azure OpenAI Topic Splitter query tuning stage to enable SmartHub to detect, split, and process multi-topic user queries in parallel. These settings control how sub-queries are generated, executed, logged, and recombined into a single coherent response.

Prerequisites

Note the following prerequisites before configuring the AI Visual Interpreter component:

Configure the tuning stage

  1. In the AutoClassifier administration portal, Add a new component to a new or existing pipeline.

  2. When adding your component, select Azure OpenAI Knowledge Graph Entity Extractor from the New Component list and provide a name for your component.

  3. Click Add Component.

  4. Select your Azure OpenAI Topic Splitter component from the list.

  5. In the Endpoint URL field, enter the deployment url of your Azure OpenAI service resource. To find your Deployment URL, in Azure AI Studio, click Deployments from the left panel, then click on your deployed resource. On your deployed resource page, copy the Target URI value in the Endpoint section.

    You must include the API version in the endpoint URL. Upland BA Insight recommends copying the URL from Azure AI Studio as the API version is included.

  6. In the Api Key field, enter the Api key for your Azure OpenAI service resource. To find your Api key, in the Azure Portal, click on your Azure OpenAI service resource. In the left panel, click Keys and Endpoint and copy the value in the KEY field.

  7. In the Prompt field, define the instructions that guide the AI model on how to split the user's multi-topic input into separate sub-qeuries. For example "Given a natural language customer search input, split it into multiple independent and self-contained search queries suitable for downstream search or retrieval tasks without using verbs".

  8. In the Test Search Query field, enter a sample query to test your tuning stage configuration.

  9. Click Test Request. The tuning stage will send the test search query to the OpenAI model and will populate a response in the Response field.

  10. When you are happy with your test results, click Accept.