[KageRyo]

Summary

Closes #66.

Live CLI analysis and Discord /predict now use exactly the requested number of completed candles. For a four-hour candle starting at 12:00 UTC, a 14:00 snapshot excludes it and a 16:00 snapshot includes it. One UTC snapshot governs both pages and all retries; a missing window raises an error with requested and available closed counts.

The first page requests one extra row where possible, with at most one older page to retain support for 1,000 completed candles. Shared interval arithmetic handles fixed minutes, daily, weekly, and calendar months while preserving backtest import paths and error types. Historical retrieval and persisted datasets retain their existing behavior. Both READMEs and the Unreleased changelog describe the defaults; package version remains 4.1.5 pending a separate release.

Validation

• ruff check .
• ruff format --check .
• pyright
• pytest — 165 passed, 85.58% coverage
• python -m build and python -m twine check --strict — wheel and sdist passed
• Boundary, calendar month, 42/180/1000-window, duplicate/empty/nonadvancing page, and once-per-call clock/retry regressions
• Real adapter → service → CLI/Discord tests using raw fake API data; negative control with old filtering fails all five new integration cases

Checklist

• Branch follows feature/<issue>-<description>.
• No credentials, .env files, or generated market data were committed.
• English and 正體中文 documentation are aligned when relevant.
• Outputs remain clearly labelled as informational, not financial advice.

[sonarqubecloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud