Jev for RAG? Page Search API Announced
PageIndex announces the Page Search API, claiming it is faster and more accurate than SOTA embedding models, and discusses the Jev decision model's retrieval performance for RAG. Link
THE JEV GUIDE · CURATED EDITION
Launch news, explanations, demos, and community projects around TypeSafe Jev, organized by topic. Every entry links back to its X source.
Curated selection · Not exhaustive · Updated manually
Sponsors
These products pay for the directory. Want yours up here?
PageIndex announces the Page Search API, claiming it is faster and more accurate than SOTA embedding models, and discusses the Jev decision model's retrieval performance for RAG. Link
The post recommends the jev-seo tool: give it a homepage URL, it crawls the entire site in about a minute and checks robots.txt, sitemap, dead links, redirects, canonical, structured data, and Core Web Vitals against Google's official 52 search guidelines; every page also goes through the TypeSafe Jev model.
Analyzed 36 retrieval queries across 3 knowledge collections, comparing original ranking with TypeSafe AI's Jev on identical candidate sets. Evidence coverage@10 improved from 83.3% to 91.7%.
Discusses how hybrid search is not always best for RAG queries; simple questions may skip reranking for lower cost, and fast decision models like Jev can dynamically choose the optimal retrieval strategy.
A people search app built on Jev's judgment API. Give it a name and any details, and it picks the intended person with probabilities, including top-two output when close. Runs on your own keys. Link
One card in the feed, every sixteen, on every page of it. $50 a week, cancel any week.
The author notes open-weight models are falling behind for general LLMs, but the situation differs for smaller decision models (Jev-class models).
The author shares that with Parallel's web search and AI classifier, they can quickly answer questions like 'Is Fable 5.5 available now?', praising Jev for making classifier AI use cases cheaper. Link
Introduces the paper 'Retriever-Free Retrieval,' which proposes using Jev or similar models for document retrieval instead of RAG, along with its mechanism and results.
A people-search tool powered by TypeSafe AI's Jev model, priced at $0.0089/lead, 85% cheaper than Clay, plugin-ready for any agent, and fully open source. Link
On hevmind's three-corpus reranker leaderboard, Jev batch and Perplexity Choice both score 0.501 mean nDCG@10, while Mixedbread v3.1 listwise leads at 0.511. Link
On B.AI, JEV-1.13.0 can filter retrieved passages before they reach your expensive model, improving RAG at lower cost. Link
The user asks whether similar questions are generated and wonders if it displays high-scoring items in Jev.
Star the project on GitHub to support continued updates and curation.