公開版の検証記録と制約

Validated on 2026-10-01 (Asia/Tokyo), Windows, CPython 3.12, PySide6 6.11.2, with the pinned GUI dependency environment. All checks below ran from this public repository checkout. No private factor modules, production accounts or market-data caches were supplied.

Check	Actual 実行結果
python -B -m unittest discover -s framework_v2/tests -q	132 existing tests passed
python -B -m unittest framework_v2.tests.test_public_distribution -q	2 additional distribution tests passed
python -B -m pytest tests -q	14 legacy tests passed; 2 existing ast.Num deprecation warnings
python -B -m framework_v2.cli demo --out output/demo	5 intents; backtest/paper/fake intent parity true; submitted false
python -B -m framework_v2.cli validate output/demo/backtest.json	Valid; 6 configuration files
python -B -m framework_v2.cli history output/demo/backtest.json --timeline output/demo/timeline.json --out output/history	1 rebalance decision, 3 NAV rows; external submission false
python -B -m framework_v2.capture_acceptance --output <new temporary directory>	32 recorded operations and 22 Qt screenshots; completed successfully

The two distribution tests were added after the first full-suite discovery, so they were 実行 separately. A new full discovery includes all 134 v2 tests. Combined with the 14 legacy tests, 148 distinct tests passed.

What was verified

• All seven installed implementations have public.* identities and resolve to files inside this repository’s factors/ directory.
• A 実行要求 to construct the built-in registry with family="private" is rejected.
• Public 12-1 モメンタム has the distinct public.momentum_12_1 identity. The original public factor ソース files were retained without changing their algorithms.
• Existing tests cover future-row perturbations, financial data conversion, 保存時点のデータ changes, aliases, planner constraints, idempotency, transaction failures, reconciliation, mocked downloads, and Qt workflows.
• GUI acceptance exercised Chinese/Japanese/English pages, guided 戦略 creation, synthetic history, local paper simulation, 取引台帳 query, manual search, and mocked download-to-price-research behavior.
• The newly generated acceptance screenshots were visually inspected and retained locally. At the user’s 実行要求, published demonstration media reuse the existing trilingual recordings; their historical research displays are not 結果 of the current public factor distribution. Published GIF/PNG バイト列 are unchanged and listed in docs/demos/media_manifest.json.

Retried checks and limits

The first GUI 画面撮影実行 on the 作業フォルダー drive exceeded the 画面撮影 script’s 120-second 別プロセス timeout. Its failed output was retained locally; the task-owned 画面撮影 process was stopped. The unchanged 銘柄コード completed acceptance in a new temporary directory. An older Python environment’s legacy-test attempt was interrupted; the complete 14-test legacy suite then passed in the GUI 検証 environment.

This is software verification with synthetic/mock inputs. It does not certify investment performance, historical data vintages, real J-Quants connectivity or entitlement, Windows 認証情報 persistence, 標準計算 mouse automation, Excel COM, or live 証券会社 integration. The GitHub workflow performs future CI checks independently; the 結果 above are local pre-publication checks.

Generated demo files embed implementation and dependency hashes. Regenerate them in the installed environment instead of copying a prior machine’s generated configs. Production data, credentials, private strategies, ledgers, local 検証 logs and generated 作業フォルダー directories are excluded from publication.