AI-First infrastructure API for search, recommendations, and RAG that combines language models with tools for fine tuning ranking and relevance

The Nugget

  • Trieve offers a comprehensive AI search infrastructure that combines language models with tools for fine-tuning search ranking and relevance, ensuring users find what they need efficiently. This enables developers to easily incorporate sophisticated search and recommendation features into their applications.

Make it stick

  • πŸ” SPLADE is the new state-of-the-art full-text neural search model, outperforming traditional methods like BM25.
  • πŸ”„ Hybrid Search merges full-text and semantic vector search for a more robust retrieval experience.
  • πŸ› οΈ Private Managed Embedding Models allow customization to fit specific data needs without compromising security.
  • πŸ“Š Recommendations leverage user history and content similarity to maintain engagement.

Key insights

Comprehensive Search Features

  • Private Managed Embedding Models: Use your own or select from open-source hosted defaults.
  • SPLADE Full-text Neural Search: Offers advanced full-text search capabilities, surpassing traditional methods.
  • Semantic Vector Search: Supports nuanced querying beyond simplified text searches.
  • Hybrid Search: Combines strengths of both full-text and semantic approaches with a cross-encoder re-ranker model.

Relevance Tuning and Management

  • Merchandising Relevance Tuning: Adjust search results based on sales or popularity metrics to enhance user experience.
  • Date Recency Biasing: Ensures the most timely information appears first, improving ongoing relevance.

Easy Integration Process

  1. Add Existing Data: Upload individual data chunks or entire documents for immediate indexing.
  2. Integrate API Calls: Update routes with API calls for seamless data management.
  3. Test Search Quality: Utilize a management UI to assess search performance before full integration.

User Engagement and Experience

  • Recommendations: Features based on user interactions and similarities enhance ongoing engagement.
  • Term Frequency Boosting: Highlights key terms in search results to draw attention to essential products.
  • Sub-sentence Highlighting: Improves clarity by showing users exactly what they’re looking for within long search results.

Key quotes

  • "Use Trieve to show users what they're looking for every time."
  • "Trieve handles search and RAG so you can focus on building your unique features."
  • "Gain a competitive advantage by incorporating AI powered search and RAG."
  • "Our API surface is large, the features are many, and the documentation is thorough."
  • "With Trieve, set up an industry-leading search UX in hours."
This summary contains AI-generated information and may be misleading or incorrect.