LayersLayer 6 of 15
Retrieval
Retrieval is how an AI system finds the right information at the moment it needs it: web search APIs, vector databases, full-text search engines, crawlers, and the frameworks that tie them into an answer.
The categories
11 categories and 326 products in Retrieval.
CategoryWhat it coversProducts
Chunking and embedding pipelinesChunking and embedding pipelines covers libraries and services that split, enrich and embed content into indexes.38
Enterprise AI searchEnterprise AI search covers permission-aware search and answers across workplace apps and documents.28
Hybrid and full-text search enginesHybrid and full-text search engines covers keyword, BM25 and hybrid lexical-plus-vector search engines and libraries.26
Knowledge graphsKnowledge graphs covers databases that store facts as connected entities, and the retrieval methods built on them.28
RAG frameworksRAG frameworks covers libraries for retrieval-augmented generation, where a model answers from documents it looks up.28
Recommendation & personalization enginesRecommendation & personalization engines covers APIs and open-source engines that rank items per user (recommendations, personalized feeds, search ranking) from interaction data.29
Text-to-SQL, semantic layers & data agentsText-to-SQL, semantic layers & data agents covers governed semantic/metric layers plus text-to-SQL engines and data-analyst agents that let people and agents query warehouses and databases correctly (structured retrieval, often exposed over MCP).34
Vector databasesVector databases covers databases that store embeddings and find the closest matches.38
Video understanding & video search APIsVideo understanding & video search APIs covers APIs that index video/audio libraries into embeddings, timelines and metadata so apps and agents can search, summarize and ask questions about video.20
Web data / scraping / crawlingWeb data / scraping / crawling covers tools that fetch and crawl web pages for AI systems.34
Questions
Retrieval, answered.
What is the Retrieval layer of AI infrastructure?Retrieval is how an AI system finds the right information at the moment it needs it: web search APIs, vector databases, full-text search engines, crawlers, and the frameworks that tie them into an answer.
What does the Retrieval layer include?It has 11 categories: Chunking and embedding pipelines, Enterprise AI search, Hybrid and full-text search engines, Knowledge graphs, RAG frameworks, Recommendation & personalization engines, Search APIs, Text-to-SQL, semantic layers & data agents, Vector databases, Video understanding & video search APIs and Web data / scraping / crawling.
How many Retrieval products does Unfragile track?326 products across 11 categories, each category with its own page.
What works here
The three products on this layer.
SearchLive
RecommendNot live yet.
RouteNot live yet. This is the layer Route starts with, through search APIs.