Find emails from a plain-language description
Narrow an inbox with keywords or embeddings, then have Jev judge which emails match a plain-language request such as 'unpaid invoices from clients'.
The use case
Founders, assistants and account managers waste time hunting for an email they cannot find by keyword, such as the message where a client agreed to new terms. Here a plain-language request goes in, a quick search or embedding step pulls candidate emails, and Jev judges each candidate against the request, returning the matches in order. The author advises the pre-filter for very large inboxes so the per-email calls stay cheap. The person reads the emails themselves before acting on anything found, and email content sent to an API needs a privacy check.
Example workflow
- Take a plain-language request from the user
- Pull candidates with Gmail search or an embedding index
- Ask Jev a Score question per candidate against the request
- Show ranked matches with links in a sidebar
- The person opens and reads the emails before acting
Tools
- Jev (TypeSafe AI)
- Gmail API
- pgvector
Risks and guardrails
- Email content leaves your account when sent to an API
- Pre-filters can drop the email you wanted
- Results can miss threads where meaning is spread across replies
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Intent-based search in Gmail by Nader Dabit on Made with Jev: madewithjev.com/builds/gmail-intent-search
- Intent-based search in Gmail: original post or repo by Nader Dabit: x.com/dabit3/status/2100960281769738433