What it does
Its development build supports several search stores, with Pinecone still experimental. Failed Jev stages mark results as degraded but do not all restore the original search results; injection screening is not a security boundary.
Integrations
jev-rag-retrieval
A Python library uses Jev to screen and tag document passages, decide which search results contain evidence, and check whether they can answer a question.
Bookmark: jev-rag-retrieval
Keep this in your collection.
Leave a noteWhat would you try with this?
: jev-rag-retrieval
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Your developer can borrow how this project separates helpful facts from useless text in a search app. The program asks an outside AI service called TypeSafe simple questions about each paragraph. Ordinary code then decides what text to keep before a separate AI writes the final answer.
You can test how the system sorts a few of your own documents. Have your developer check what happens if the outside connection fails during this review process. A broken connection leaves a warning mark on the search results, but it cannot restore text that was already discarded.