[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1...¶
[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1) FTS5 AND→OR matching in _fts_search and _fts_search_working. (2) Recall scoring: importance 0.35 (was 0.2), relevance 0.25 (was 0.35), working tier_bonus +0.15. (3) Length penalty: min(1.0, 200/word_count) on substring fallback paths, partial/cross scans capped at 300 words. Without these, large vault pages drown high-importance working entries.
[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1) FTS5 AND→OR matching in _fts_search and _fts_search_working. (2) Recall scoring: importance 0.35 (was 0.2), relevance 0.25 (was 0.35), working tier_bonus +0.15. (3) Length penalty: min(1.0, 200/word_count) on substring fallback paths, partial/cross scans capped at 300 words. Without these, large vault pages drown high-importance working entries.
[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1) FTS5 AND→OR matching in _fts_search and _fts_search_working. (2) Recall scoring: importance 0.35 (was 0.2), relevance 0.25 (was 0.35), working tier_bonus +0.15. (3) Length penalty: min(1.0, 200/word_count) on substring fallback paths, partial/cross scans capped at 300 words. Without these, large vault pages drown high-importance working entries.
[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1) FTS5 AND→OR matching in _fts_search and _fts_search_working. (2) Recall scoring: importance 0.35 (was 0.2), relevance 0.25 (was 0.35), working tier_bonus +0.15. (3) Length penalty: min(1.0, 200/word_count) on substring fallback paths, partial/cross scans capped at 300 words. Without these, large vault pages drown high-importance working entries.
[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1) FTS5 AND→OR matching in _fts_search and _fts_search_working. (2) Recall scoring: importance 0.35 (was 0.2), relevance 0.25 (was 0.35), working tier_bonus +0.15. (3) Length penalty: min(1.0, 200/word_count) on substring fallback paths, partial/cross scans capped at 300 words. Without these, large vault pages drown high-importance working entries.
[scope:core] Mnemosyne beam.py patches (must re-apply after venv/upgrade): (1) FTS5 AND→OR matching in _fts_search and _fts_search_working. (2) Recall scoring: importance 0.35 (was 0.2), relevance 0.25 (was 0.35), working tier_bonus +0.15. (3) Length penalty: min(1.0, 200/word_count) on substring fallback paths, partial/cross scans capped at 300 words. Without these, large vault pages drown high-importance working entries.