How It Works
Understanding Mixpeek’s architecture and capabilities
Mixpeek provides infrastructure that allows developers to construct their own search experiences, aligned to their business. It enables this via three layers:
- Feature Extraction - Extract meaningful information from any content type using state-of-the-art models (OSS, BYO, or Mixpeek proprietary)
- Entity Enrichment - Enhance extracted features with domain knowledge through manual classification, automatic clustering, and relationship mapping
- Search & Discovery - Build powerful search experiences with vector similarity, filtering, grouping, and advanced capabilities like Learning to Rank and Fine-Tuning.
1. Multimodal Feature Extraction
Features are the core primitive in Mixpeek. They represent extracted information from any content type using state-of-the-art AI models.
Example: Video Processing
Learn more about Feature Extraction →
2. Feature Enrichment
Once features are extracted, they can be enriched with domain knowledge through three main mechanisms:
Taxonomies
Manual classification using predefined hierarchies for strict content organization
Clusters
Automatic content grouping using similarity and patterns
Relationships
Define connections between features, nodes, and clusters
3. Search & Discovery
Enriched features enable powerful search capabilities:
Search Configuration Example
Best Practices
Plan Your Feature Extraction
Choose appropriate models and intervals based on your content type and needs
Design Your Domain Model
Create taxonomies and relationships that reflect your business logic
Optimize Search
Use filters, grouping, and reranking to improve result relevance
Monitor & Iterate
Track search interactions and refine your implementation
Need help getting started? Check out our Quick Start Guide or contact our support team.
Was this page helpful?