Native semantic image search built entirely inside ODC. Upload images, embed them with Cohere Embed v4, and store the vectors in ODC entities. No external vector database needed.

Two screens. The ingestion screen lets users upload images and add them to the searchable index. A live gallery shows everything currently indexed.

The search screen supports two query modes. Text-to-image: describe what you are looking for in natural language and get back ranked results. Image-to-image: upload a reference image and get back visually similar matches. Both modes return results with similarity scores.

Prerequisites

• An active ODC environment.

ImageVectorLibrary Forge component installed and configured in your ODC tenant.

MultimodalEmbeddingLibrary Forge component installed. It is a dependency of ImageVectorLibrary.

• A valid Cohere API key with access to the embed-v4.0 multimodal model.

ImageVectorLibrary Site Properties.

• ODC Portal access to assign roles to users.

Installation
Installing this Forge component will automatically install the dependency libraries: ImageVectorLibrary and MultimodalEmbeddingLibrary

Configure ImageVectorLibrary

ImageVectorLibrary, then Settings.

• Set the following Site Properties:
• APIEndpoint — the Cohere API base URL (e.g. https://api.cohere.com)
• ModelProvider — the provider name (e.g. Cohere)
• ModelVersion — the model version to use (e.g. embed-v4.0)
• APIKey — your Cohere API key (secret property)
• Save the settings.

Usage Guide Ingesting Images

• Navigate to the Image Ingestion screen.
• Click Select image asset to index and choose an image file from your device. JPEG, PNG, and other common formats are supported.
• A preview of the staged image appears on the right with the file name and a thumbnail.

Generate and Store Vector. A loading spinner shows while the embedding is being generated.

• On success, a confirmation message appears and the image is added to the Current Vector Index gallery below.
• Repeat for each image you want to index.
• To remove an image, click the trash icon on its card. A confirmation prompt appears before deletion.
• Scroll down to load more images if your index exceeds 200 entries.

Images are processed one at a time. There is no bulk upload.

Text-to-Image Search

• Type a description in the search box, for example: a cat wearing sunglasses.

Max Results (Top K) and Minimum Score Threshold. Higher threshold values return only strong matches.

Run Text Search. Results appear as a grid of image cards with file name, image ID, and a Score badge.

Image-to-Image Search

Use this mode to find images visually similar to a reference image you provide.

Image Match Mode.

Select Query Image and choose a reference image from your device. A thumbnail preview appears with the status Image Staged.

Max Results (Top K) and Minimum Score Threshold.

Run Image Match. Results appear ranked by visual similarity with a Score badge on each card.

If no results appear, try lowering the Minimum Score Threshold to 0.2 or increasing Max Results.

Configuration Reference

Data Model

Deleting an Image record cascades to its ImageVector record automatically.