jevbooks

← All projects

duckdb-jev

prasanthj/duckdb-jev · Homepage

Native DuckDB extension for Jev classification, scoring, and semantic predicates from SQL.

Classification & Routing52%Runner-up: Infra / SDKs / IntegrationsQuestionnaire

What it is

A native C++ DuckDB extension that applies TypeSafe/Jev judgments directly to structured SQL rows, for analysts and engineers who want classification, scoring, and semantic predicates without Python UDFs or a separate inference server.

How it uses Jev

Jev answers Noul predicates, finite Choice classifications, ordered Score rubrics, and mixed multi-question evaluations. It takes SQL row evidence (text, JSON, STRUCT, LIST, ARRAY) plus instructions and criteria, batches up to 1,000 judgments per request, and returns probabilities, choices, scores, or named answers used directly in SQL.

Primitives:choicescorenoul

Technique worth stealing

Batch up to 1,000 independent judgments per request with bounded concurrent HTTP work and stream across DuckDB chunks.

Try it

duckdb -unsigned; LOAD '/absolute/path/to/duckdb-jev/build/extension/jev/jev.duckdb_extension'; SELECT jev(...);

View on GitHub

judged by Jevjev-1.13.0

Evidence

Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.

  • Jev-centricyes0.94
  • Shows a System One patternyes0.83
  • Handles uncertaintyno0.06
  • Measuredyes0.96
  • Runnableyes0.96
  • Worth recommendingno0.31
  • Model replicano0.33
  • Problem scopescore on a 0–2 scale1.06
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 21 Sept 2026.