pg-jev
pg-jev is a Postgres extension for people who already filter rows in SQL. The jev() function judges each row in plain language without embeddings or a vector index. Zachi's launch clip judged 129 rows in about one second for about $0.0009, with cache hits near 6 ms on a second run.
Overview
pg-jev (github.com/realZachi/pg-jev, pgjev.com) is Zachi's (@iam_zachi) PostgreSQL extension exposing jev() for per-row judgments in plain language. Example patterns from the launch clip include WHERE jev(people, 'could work from home') without embeddings. Zachi reported 129 rows judged in about one second for about $0.0009, with cache hits near 6 ms on a second run. This listing is the database extension only, not the separate Chrome ad-blocker demo elsewhere in the showcase. It is also separate from jev4pg. pg-jev is one per-row jev() predicate. jev4pg plans SQL from a natural-language question and ships a larger operator toolkit plus a BIRD comparison.
Problem: Teams want natural-language filters over Postgres rows without standing up embeddings, vector indexes, or a separate inference service.
Built for: Backend engineers and data folks who already trust SQL and want Jev Noul, Choice, or Score predicates inside WHERE clauses.
First indexed on Jev Directory: 2026-09-18
Creator and team
- Name
- Zachi
- Handle
- @iam_zachi
“jev(): a PostgreSQL extension that searches your whole database in natural language. No index, no embeddings, just one function.”
How Jev is used
- Role in the product flow
- Per-row Jev predicates inside SQL filters
- Primitives
- NoulChoiceScore
- State in
- Row JSON or table columns passed into jev() plus a plain-language question string (per launch clip and pgjev.com docs).
- Decision out
- Boolean or typed judgment per row used directly in WHERE clauses.
- Install pg-jev extension in Postgres
- Write SQL with jev(table_or_row, 'plain language criteria')
- Extension calls TypeSafe per row (with caching on repeats)
- Return matching rows to the application
pg-jev pushes Jev to the data plane. Application code stays SQL-native: analysts phrase constraints in English, the extension maps each row through System One, and the planner filters like any other predicate. Caching repeated questions is what drives single-digit millisecond hits on warm rows in the clip. You can mix Noul-style yes or no gates, Choice among small enums, and Score thresholds depending on how you phrase the function call in SQL. The showcase ad-blocker clip for @iam_zachi is a different Chrome extension product; do not conflate it with this Postgres extension.
Sourced performance claims
- 129 rows judged in about one second for about $0.0009; second run about 6 ms from cache.Source: iam_zachi X post 2100679300756435135
- Public GitHub repo realZachi/pg-jev had about two hundred eighty three stars when this listing was drafted.Source: GitHub star count September 2026
Features and stack
Features
- jev() SQL function without vector indexes
- Documented latency and cost figures from launch clip
- pgjev.com site and open-source extension repo
- Distinct from ad-blocker-iam-zachi showcase
Stack
- PostgreSQL extension
- TypeSafe System One
- pgjev.com
Pricing: Open source extension; TypeSafe usage bills per row judged unless cache hits apply.
Demo
Showcase clip with the same lightbox player as the homepage. Click to play.
pg-jev SQL filters in plain language
WHERE jev(people, 'could work from home') and no vector index babysitting. Postgres rows get Noul gates; cache hits land in six milliseconds.
Links
FAQ
- Is this the same as jev4pg?
- No. pg-jev is one jev() function for a single row inside a WHERE clause. jev4pg is a separate project: it plans SQL from a natural-language question, ships 41 Jev operators, and publishes a BIRD comparison. See the jev4pg listing.
- Is this the same as the ad-blocker showcase?
- No. ad-blocker-iam-zachi is a Chrome extension demo. pg-jev is a PostgreSQL extension for SQL filters at pgjev.com.
- Do I need embeddings?
- Zachi's launch clip emphasizes no embeddings and no special indexes for the jev() function path shown.
- What primitives does SQL expose?
- Marketing and docs describe Noul, Choice, and Score style questions depending on how you phrase the plain-language predicate per row.
Tags
Related learn guides
Original Jev guidance that pairs with this product pattern.
- Jev use cases
The patterns builders actually search for: moderation, routing, triage, RAG verify, and agent gates.
Related products
Hand-picked neighbors with rich profiles or overlapping tags.