NewObsidian LLM Hub
An Obsidian plugin for multi-provider chat, workflows, and local vault search, with optional Jev filtering for retrieved results.
Added to Jevfast
NewHow it uses Jev
After local retrieval returns chunks for a query, the plugin creates one Choice question per result with keep/exclude criteria. Jev's selections filter the result list before the chat model receives context.
What Jev decides
Keep or exclude a retrieved chunkExample answers · not a recorded Jev response · Source ↗
Question 1 · result 0
YOUR APP
INSTRUCTION
Decide whether this retrieved chunk answers the query or materially supports an answer; word overlap alone is insufficient.
STATE
Representative context: the user's search query and one retrieved file path and text chunk. One question is created per result, so the batch size varies.
JEV · CHOICE
- keep
- exclude
App workflow
- Search the vault
The user's query retrieves local RAG chunks from the Obsidian vault.
- Judge retrieved chunks
When Jev filtering is enabled, the query and each chunk are submitted as typed Choice questions.
- Pass relevant context onward
Only chunks selected as keep remain for the chat flow; Jev does not generate the response.
Why it is interesting
It uses a compact relevance judgment to refine local retrieval results before synthesis, and keeps the relevance decision separate from response generation.