AI Search - Semantic – Papirfly Help Center

What is it?

Semantic search helps you find assets by describing what you are looking for in your own words.

Important: This is not a keyword or metadata search. Semantic search analyzes the visual content of the image to find what you describe.

For example: if you search for "a mountain at sunset," the system analyzes the images themselves (colors, shapes, composition) to find images with those visual characteristics. It does not look for these words in tags or metadata.

Appropriate examples

This approach works well for:

"White SUV, side view, beige leather interior, photographed in studio"
"Navy blue silk dress on mannequin, zipper detail"
"18-carat gold ring with central sapphire, neutral gray background, studio lighting"
"White sandy beach, turquoise water, palm and coconut trees"
"Table set with white tablecloth, glass of red wine, gourmet plate"

→ The more detailed and visual the description, the better.

What does not work

Semantic search does not look into metadata. Do not use it for:

Product codes ("SKU564FRED")
Exact dates ("published on March 15, 2024")
Version numbers ("file 152A")
Business statuses ("active," "archived")

→ For these: Use classic search with filters.

How to use it?

Open the search bar
Click the "Semantic Search" button (AI icon)
Describe in detail what you are looking for (colors, mood, context)
Use classic filters to refine if necessary

How does it work?

The AI analyzes the images and the preconfigured metadata to create a digital representation of the content. It then compares this representation with your search text to find the visually closest assets.

It analyzes: Shapes, colors, objects, contexts, moods

It does not look for: Exact keywords, business filters, technical references

How to activate it?

Semantic search is an optional feature. It is not enabled by default on your account.

To use it:

Contact your CSM (Customer Success Manager) or the support team
Request activation of semantic search
Once activated, you will see the "Semantic Search" button (AI icon) in your search bar

This feature requires specific configuration for your DAM instance.

Model used

Semantic search uses a multi-modal AI model based on AWS Bedrock. This model understands both images and natural language to perform comparisons.

Data security and privacy

Your data remains confidential. Here’s how:

The process: When you upload media, an embedding calculation is performed. This process occurs in two steps: one by Keepeek, one by AWS Bedrock. This process is strictly limited to similarity calculation and does not involve storage or further use by AWS.

Data transmitted to AWS:

Thumbnail of your image (max 1200x1200 pixels)
Selected metadata (title, keywords, categories, custom fields)

Storage and use:

AWS does not store your images or metadata
Data is used only for embedding calculation
Once calculation is complete, AWS no longer has access
Your data does not train AI models
No distribution to third parties

Data flow:

Embeddings are calculated by AWS Bedrock and immediately returned to the search microservice (Keepeek)
Embeddings are indexed in Elasticsearch at Keepeek (not at AWS)
Original images and metadata remain only at Keepeek

Security in transit:

All data flows are encrypted with TLS 1.2. All requests to the Amazon Bedrock API are sent over a secure connection (SSL).

Multi-tenant isolation:

Amazon Bedrock has one account per region/provider. These accounts are managed exclusively by the Bedrock team. Providers have no access to accounts, logs, or customer data.

Infrastructure in Europe:

Our Bedrock infrastructure is deployed in eu-north-1 (Stockholm). IAM policies limit access to this region. By default, all data stays in Stockholm. No interregional routing.

GDPR:

The AWS Global DPA applies automatically. It is incorporated in the AWS Service Terms.

You can also:

Disable AI indexing on certain statuses/folders.
Limit semantic search to a selection of media.
Define which fields to vectorize (4-5 max).