Infrastructure for building AI search into your applications
Trieve combines powerful language models with tools for human fine tuning. Get features like dense vector semantic search, sparse vector full-text search, cross encoder re-ranker models, RAG endpoints, relevance weighting, and more in a single service.Trieve: Production-ready. Use Trieve to show users what they're looking for every time.
What you get with Trieve
Everything you could expect from modern AI infrastructure for search and RAG
Private Managed Embedding models
Bring your own embedding model or use one of our available open source hosted defaults.
SPLADE full-text neural search
BM25 is great, but SPLADE is better. SPLADE is a state-of-the-art retrieval model for full-text search.
Semantic Vector Search
Sometimes you need more than just full-text search. Trieve supports semantic vector search out of the box.
Get the best of both worlds. Trieve supports hybrid search, combining full-text search and semantic vector search with cross-encoder re-ranker models.
Merchandising Relevance tuning
Boost search results based on sales or popularity to show users what you know works.
Date Recency Biasing
Relevance tuning is not just about the query. Trieve supports date recency biasing to ensure the most recent results are shown first.
Build What Matters Most
Trieve handles search and RAG so you can focus on building your unique features
Built on the best foundationsGain a competitive advantage by incorporating AI powered search and RAG
Private Open Source Models
When requested we use open source embedding models and LLMs ran in our own servers, so data is never leaked.
Our API surface is large, the features are many, and the documentation is thorough. We've got you covered.
Sensitive data? Don't worry about vendor agreements, host yourself with terraform templates and no external dependencies.
Get an industry leading search UX setup in hours.
Step 1: Add Existing Data
Send your data to the API. Either upload individual chunks or entire documents that get chunked by our algorithms.
Step 2: Integrate our API
Add calls to our API on your create and update routes to keep your data up to date.
Step 3: Search, Recommend, or Generate
Start testing search quality with our management Search UI and then integrate the search or RAG API calls into your product.
Most used features
Provides frequently used endpoints for rounding out the full spectrum of search and RAG features.
Full-text search with an index built by SPLADE and semantic vector search with your choice of embedding model.
Build RAG experiences with our API that returns generated outputs with citations and manages message history.
Recommendations are a great way to keep users engaged. Trieve supports recommendations based on user history and content similarity.
Duplicate content is a common problem in knowledge management. Trieve has a novel approach to duplicate detection and merging.
Search results can be paragraphs long. Trieve supports sub-sentence highlighting to show users exactly what they are looking for.
Relevance tuning is not just about the query. Trieve supports date recency biasing.