# Kurmancî v1 Scope Decisions

> Status: v1 decision doc for #42. This document captures the agreed direction from issue discussion and Plan PR #48. It does not replace native-speaker review.

## Decision

- v1 starts with **Kurmancî only**.
- Current v1 scope is **Upper Central Kurdish**.
- First MVP path is **Badinî / Bahdini**.
- Candidate related branches are **Torikî / Cizîrî / Şingalî**.
- Contributors should **self-identify** their zone or sub-dialect.
- City/country is recorded as **context or a soft prior**, not as a hard dialect label.
- The first MVP prompt set is capped at **max 50 prompts** across words, phrases, and sentences.

## Candidate Places

These are review anchors from #42, not a final city-to-dialect inference table.

| Branch | Candidate places |
|---|---|
| Badinî / Bahdini | Zakho, Duhok, Amedi, Hakkari, Urmia |
| Torikî / Cizîrî / Şingalî | Omeriyan, Midyat, Qamishlo, Hasakah, Shingal |

## Prompt Direction

Do **not** create the concrete 50 prompt questions in this ticket.

For v1, only the prompt direction is captured:

- include a mix of single words, short phrases, and complete sentences;
- keep the prompt set small enough for the MVP, with a hard cap of 50;
- use #42 native-speaker review before finalizing concrete prompt text.

Candidate variable families from #42:

- "so / like this / like that"
- "to speak"
- simple sentences, such as "I am going home"
- days of the week
- negation forms
- kinship terms

## Downstream Use

- `kk-web` should ask contributors to self-identify their zone/sub-dialect and collect city/country as context.
- `kk-api` may store city/country as metadata or a soft prior, but must not treat it as stronger than user self-identification.
- Prompt catalogue work may link prompts to these variable families later, after concrete prompt questions are reviewed.

## Still Open

- Which concrete max-50 prompt variables/questions should the MVP start with?
- How should the final branch labels and spellings be shown to contributors?
- How granular should later city/place metadata become for v1 versus future corpus analysis?