Webctl
The state keeps the search intent beside each candidate result, while the same judgment API supports individual calls and batched per-result decisions to limit round trips.
Added to Jevfast
How it uses Jev
Webctl uses Jev as typed judges in its web research/search pipeline. For a result it sends the user query and optional goal together with result title, URL and snippet, then asks either a rubric Score for relevance or a caller-supplied Noul yes/no. Results consume the typed score/probability for qualification and ranking.
What Jev decides
Search-result relevance questionExample answers · not a recorded Jev response · Source ↗
Question 1 · answer
YOUR APP
INSTRUCTION
Answer the user-supplied yes/no relevance question about this result, using only its title, URL and snippet in the context of the query.
STATE
State combines query and optional goal with one search result (title, URL, snippet). Prompt renders the caller-supplied question alongside the query and result.
JEV · NOUL
YesNo
App workflow
- Search the web
Collect candidate results for a query and optional goal.
- Prepare result state and prompt
Pair each result with search intent and render result-specific instructions.
- Ask Jev
Request a rubric Score or caller-defined Noul, singly or in batches.
- Use typed relevance
Consume typed result answers in the qualification pipeline.
Why it is interesting
The state keeps the search intent beside each candidate result, while the same judgment API supports individual calls and batched per-result decisions to limit round trips.