Scope
Produce Turkish (tr) LLM-assisted candidate translations for the existing Babelon profile.
Current audit
- workflow:
tr-gap-fill
- existing profile:
babelon/hp-tr.babelon.tsv
- current
NOT_TRANSLATED rows: 6,128
- current completion: 80.3%
Workflow
Follow docs/llm_translation_workflows.md and translation_workflows/candidate_workflows.json.
- generate candidate JSONL with Codex CLI /
gpt-5
- preserve exact HPO keys from exported work items
- validate with
agent-apply-dry-run
- triangulate with free review lanes:
- OpenCode /
opencode/deepseek-v4-flash-free
- OpenCode /
opencode/mimo-v2.5-free
- apply only as
CANDIDATE
- keep
translation_type=LLM_ASSISTED_DRAFT
- open/update a dedicated upstream PR when Turkish reaches zero
NOT_TRANSLATED rows
Conflict handling
Prefer established Turkish terminology already present in hp-tr.babelon.tsv. If model reviewers disagree, keep unresolved rows as CANDIDATE and document the issue in the PR body for human review.
Scope
Produce Turkish (
tr) LLM-assisted candidate translations for the existing Babelon profile.Current audit
tr-gap-fillbabelon/hp-tr.babelon.tsvNOT_TRANSLATEDrows: 6,128Workflow
Follow
docs/llm_translation_workflows.mdandtranslation_workflows/candidate_workflows.json.gpt-5agent-apply-dry-runopencode/deepseek-v4-flash-freeopencode/mimo-v2.5-freeCANDIDATEtranslation_type=LLM_ASSISTED_DRAFTNOT_TRANSLATEDrowsConflict handling
Prefer established Turkish terminology already present in
hp-tr.babelon.tsv. If model reviewers disagree, keep unresolved rows asCANDIDATEand document the issue in the PR body for human review.