Configure AI models

A platform administrator configures the AI models used by Insights. The default Agent model handles conversation, reasoning, and answer generation. A default multimodal model is also required when analyzing images, scanned PDFs, or images embedded in documents.

Model role	Requirement	Purpose
Default Agent Model	Required	Conversation, reasoning, tool use, and answer generation
Default Multimodal Model	When needed	Images, scanned PDFs, and images embedded in documents

The first text-only session requires only a default Agent model. If one model satisfies both the conversation and vision requirements, it can be assigned to both default roles; otherwise, configure separate models.

Before you start

Prepare:

• a platform administrator account;
• the AI provider and model ID;
• a custom API base URL, if the provider requires one;
• a valid API key for the model service;
• network access from the HyperFileLens deployment to the model endpoint.

Open AI Models

1. In the Access section of the installation output, find Platform Ops and open its URL in your browser.
2. Sign in with a platform administrator account.
3. Open AI Engine → AI Models.
4. Select Add AI Model.

Add the default Agent model

1. Select the model provider.
2. Select a model or enter its model ID.
3. Enter the API Key. If required, enter the provider's custom API Base URL; otherwise, leave it blank to use the default endpoint.
4. Keep the model Active.
5. Select Test Connection and confirm that the connection succeeds.
6. Select Add Model.
7. In the model list, select the model, then choose More Actions → Set as Default Agent Model.

Configure the default multimodal model

When you need to analyze images, scanned PDFs, or images embedded in documents:

1. Add and test a model with vision capability. You can reuse the default Agent model if it also supports vision.
2. Keep the model Active.
3. In the model list, select the model, then choose More Actions → Set as Default Multimodal Model.
4. Create a small test session with a document that contains an image or scanned page and confirm that its visual content is processed.

Confirm readiness

Before continuing, confirm that:

• the default Agent model is Active and marked Default Agent Model;
• when visual content is required, the multimodal model is Active and marked Default Multimodal Model;
• connection tests succeed for both configured roles;
• New Chat opens without a missing-model warning.

API keys are sensitive. Enter them only in the product configuration and do not include them in screenshots, documentation, or public issues.