Overview

Semantic Search (also known as Similarity Search) allows you to find records based on their meaning rather than just matching specific keywords. Instead of looking for the exact characters you typed, the system understands the context and intent behind your search.

For example, searching for “athletic footwear” can find products labeled as “running shoes” or “sneakers,” even if the word “athletic” isn’t in their description.

Semantic search is integrated into several parts of the ProductSync platform:

  • Filter Bar: Use the “similar to” operator to find records across the entire grid.
  • Searchable Selects: When choosing a value from a dropdown (like a Category or Brand), you can search by meaning.
  • Add Related Dialogs: Find the most relevant items to link to your current record.

Using the “similar to” Operator

You can trigger a semantic search manually in any filter bar or the Advanced Filter Editor.

[navigate:/products|Open Products|shopping-cart]

1. Click the **Advanced Editor** button (slider icon) in the search bar.
2. Select **Add Condition** to start a new rule.
3. Choose a field that supports embeddings (usually **Name** or **Description**).
4. Pick the **similar to** operator from the dropdown.
5. Type a descriptive phrase or natural language query.
6. Hit **Apply** to see the results.

Adjusting Similarity (Threshold)

By default, the system uses a balanced threshold to find matches. You can fine-tune how “strict” the search is by adding a threshold value in Direct Filter Mode.

name similar to "eco-friendly packaging" threshold: 0.3
  • Lower values (e.g., 0.2): Stricter matches. Only items very similar to your query will appear.
  • Higher values (e.g., 0.8): Broader matches. The system will include items with a looser conceptual connection.

Background Processing

Embeddings are calculated in the background by a dedicated worker process.

**Note:** When you import new data or update a product's description, it may take a few moments for the semantic search index to update. If a new item doesn't appear in similarity results immediately, wait a minute and try again.
  • Be Descriptive: Instead of searching for “red,” try “vibrant crimson summer dress.” The more context you provide, the better the AI can match the meaning.
  • Use Natural Language: You don’t need to use technical keywords. Search the way you would describe the item to a colleague.
  • Combine with Filters: You can combine semantic search with traditional filters. For example: categories.category.name = "shoes" AND name similar to "waterproof hiking".

Troubleshooting

I don’t see the “similar to” operator

The “similar to” operator only appears for entities and fields where embeddings have been configured in the system metamodel. If you need this enabled for a specific field, contact your administrator.

Results are not relevant

Try lowering the threshold (e.g., to 0.3) to force the system to find closer matches. Also, ensure your search query is specific enough to distinguish between different types of products.

New products aren’t showing up

The background worker might still be processing the new data. Check the system status or wait a short period for the embeddings to be generated.