Trieve

What is Trieve?

Trieve is an AI search infrastructure platform designed to deliver relevant and fast search results by combining efficient language models with tools for fine-tuning ranking and relevance. It supports semantic and full-text searches using dense and sparse vectors, along with features like a cross-encoder re-ranker model for refining search results. Trieve allows the integration of open-source or custom embedding models and offers hybrid search functionalities that combine full-text and semantic vector searches with cross-encoder re-ranker models. Users can benefit from private managed embedding models, semantic vector search, recency biasing, and other features to enhance search results. Additionally, Trieve provides solutions for specific needs such as duplicate detection, content similarity-based recommendation, and more. It is self-hostable, giving users control over data privacy with easy API integration and thorough documentation, making it a powerful tool for incorporating AI-based search into applications.

Who created Trieve?

Trieve was created by DevFlow Inc. It was launched on March 11, 2024, and is a comprehensive AI search infrastructure platform designed to deliver fast and relevant search results. The platform offers features such as semantic and full-text searches, hybrid search capabilities, private open-source models, and tools for fine-tuning ranking and relevance. Users can bring their own embedding models or utilize the ones provided by Trieve. The platform also supports AI-native end-to-end API and self-hosting options for maximum control over data privacy and usage. Overall, Trieve is a powerful and customizable tool for implementing AI-based search functionalities into applications.

What is Trieve used for?

• Choose between state-of-the-art retrieval models for full-text search
• Support hybrid search combining full-text and semantic vector search
• Tune and boost search results through merchandising and relevance tuning
• Enable sub-sentence highlighting in search results
• Manage ingestion, embeddings, and analytics with ease
• Build unfair competitive advantages through search, discovery, and RAG experiences
• Utilize private open-source models and LLMs running on Trieve servers
• Cover various functions including chunking, ingestion, search, recommendations, RAG, and front-end
• Self-host Trieve for maximum performance and data privacy
• Set up industry-leading search in 30 minutes

Who is Trieve for?

• Developers
• Researchers
• Data scientists
• Content creators

How to use Trieve?

To use Trieve, follow these steps:

1. Add Existing Data: Upload individual chunks or entire documents via the API or the no-code dashboard; the data will be chunked by Trieve's algorithms.
2. Integrate the API: Add calls to the Trieve API on your create and update routes to keep your data current.
3. Search, Recommend, or Generate: Test and fine-tune search, recommendations, and chat quality using the search playground; then, integrate the API calls into your product.

Trieve is an all-in-one infrastructure for creating search, discovery, and RAG experiences, offering semantic vector search, full-text search, hybrid search, customizable embedding models, and more. It emphasizes transparency, data portability, and easy communication with the developers. Trieve is self-hostable, allowing users to maximize data privacy and performance. It provides comprehensive tools for tuning search quality and relevance, making it a versatile and powerful tool for incorporating AI-driven search into applications.

For more information, visit Trieve on GitHub at devflowinc/trieve.