Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

jev-kiyafet-bul

Turkish natural-language outerwear search where Jev decides in two steps, code filters and ranks, and a panel shows each step.

ServiceprototypeCustom license
View on GitHub ↗

Overview

jev-kiyafet-bul is a Turkish-language demo for natural-language clothing search. A user writes what they want, for example a long beige belted trench in size M under 3,000 lira, and Jev decides in two steps while code filters and ranks the catalogue, returning result cards. A dark side panel shows each search's step-by-step 'what Jev is doing' flow with real data.

The README explains Jev's primitives in Turkish and states a rule of thumb: Jev for decisions and percentages, an LLM for sentences, and code for numbers, arithmetic and dates. Code modules cover filtering, regex, narrowing, sorting, the search pipeline, the Jev client and quotas. Keys can come from TypeSafe or Vercel AI Gateway, and without a key the project runs with a deterministic fake Jev. A list of related projects by the same author is included.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich attribute values does the query ask for (category, colour, length, size)?(inferred)—
ScoreHow well does this item match the query?Irrelevant · Partial · Good · Exact

How it works

  1. User types a natural-language query
  2. Jev decides query attributes, then match quality
  3. Code filters by price and size and ranks
  4. Show cards and the step-by-step panel

Strengths

  • Clear split of Jev, LLM and code roles
  • Works with a fake Jev without a key
  • Transparent step panel

Caveats

  • Turkish-language demo
  • Custom licence

Getting started

`cp .env.example .env`, add a TypeSafe or Vercel AI Gateway key (or run with the fake Jev), then start the Next.js app.

At a glance

Type
Service
Status
prototype
Language
TypeScript
Runtime
Next.js web app
Has code
yes
License
Custom license

Use cases and integrations

How curated lists describe it

  • Kod yerleri: \`lib/filtre.ts\` (Adım 1), \`lib/regex.ts\`, \`lib/daralt.ts\`, \`lib/sirala.ts\` (Adım 2 + sıralama), \`lib/ara.ts\` (hat), \`lib/jev.ts\` (istemci), \`lib/kota.ts\`, \`app/api/ara/route.ts\`, \`components/\` (arayüz).

    logicrw/awesome-jev-projects

Frequently asked questions

What does Jev decide?

What the query asks for and how well each item matches.

Who does the price filtering?

Code, following the rule that numbers belong in code.

Can I run it without a key?

Yes, with a deterministic fake Jev.

Related projects

  • Jev reranking is not a free win

    X post reporting a measured catalog-search run where Jev reranking alone did not beat vector retrieval.

    Discussiondiscussion
  • Jev Search★ 534

    Plain-language web search where a decision model like Jev picks query, sources and time range, then ranks Search1API results.

    AppproductionTypeScript
  • apple-rag-mcp

    MCP server over Apple developer docs and WWDC transcripts that uses hybrid search, then Jev relevance scores to rank results.

    MCP serverproductionTypeScript
  • hippo-memory

    Local long-term memory for coding agents with an optional Jev reranker that orders recalled memories.

    ProjectproductionTypeScript
  • OpenViking

    Volcengine's context database for agents, with a Jev rerank client that scores candidate documents with jev-latest as relevance.

    LibraryproductionPython
  • jev-rerank★ 17

    Jev as a calibrated reranker: query plus up to 30 documents in one call, one relevance Noul per document, benchmarked on BEIR.

    ProjectbetaPython