Управление проектом Streamify
Работа по Streamify ведется через задачи размером под одного агента. В каждой задаче должны быть продуктовый результат, ответственное направление и конкретные доказательства готовности.
Направления агентов
| Направление | Зона ответственности | Проверка по умолчанию |
|---|---|---|
| Repo/Build | Local setup, Docker Compose, Makefile, CI, release automation | make test, make compose-smoke-local |
| Yandex Ingestion | Token-safe metadata ingestion, raw contracts, API edge cases | make preflight, make ingest, make raw-contract |
| Analytics/dbt | DuckDB targets, typed staging, marts, tests, lineage | make dbt-build, make doctor |
| Product/Dashboard | Streamlit UX, exports, product docs, answer quality | make dashboard-smoke, make product-answers-smoke |
| QA/Integration | End-to-end acceptance, privacy gates, release readiness | make acceptance-local, make acceptance-real |
Правила работы
- Do not paste tokens, raw account data, DuckDB files, or screenshots with private listening data into issues.
- Use sample metadata for CI, GitHub Pages, and public release artifacts.
- Use
make acceptance-realonly on a trusted local machine with.env. - Every PR should update tests or explain why the existing gates prove the change.
- Release candidates should link the GitHub issue set they close and include known API limitations.
План релизов
v0.1.0-local-mvp: real-account metadata ingestion, DuckDB marts, dashboard, safety gates.
v0.2.0-product-answers: richer action queues, more dashboard filters, better empty/error states.
v0.3.0-official-import: optional official Yandex data archive importer if usable account export data is available.