Merge pull request 'tune(learning): סף-אשכול סינתזת-לקחים 0.82 → 0.68 (#158 )' (#349 ) from worktree-synth-threshold-068 into main

tune(learning): lesson-synthesis cluster threshold 0.82 → 0.68 (#158 )
כיול אמפירי (dry-run על 81 לקחים): הלקחים מגוונים (חציון-דמיון ~0.48, מקס' ~0.72), כך ש-0.82 מיזג 0 אשכולות. 0.68 תופס 5 קונסולידציות אמיתיות (רישוי+השבחה+מבנה), כולן עברו את שער-ה-drift 0.80 (0.848–0.916) → נאמנות נשמרת. שער-ה-drift נשאר רשת-הביטחון מתחת לסף. Invariants: INV-LRN8 (כיול-סף בלבד; המנגנון/השערים ללא שינוי). depends-on #158. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-28 22:45:11 +00:00 · 2026-06-28 22:44:55 +00:00
1 changed files with 6 additions and 1 deletions
--- a/mcp-server/src/legal_mcp/config.py
+++ b/mcp-server/src/legal_mcp/config.py
@@ -286,7 +286,12 @@ LESSON_SYNTH_MODEL = os.environ.get("LESSON_SYNTH_MODEL", HALACHA_EXTRACT_MODEL)
 LESSON_SYNTH_EFFORT = os.environ.get("LESSON_SYNTH_EFFORT", "high")
 LESSON_SYNTH_DRIFT_FLOOR = float(os.environ.get("LESSON_SYNTH_DRIFT_FLOOR", "0.80"))
 # Cosine floor for two lessons to land in the same cluster (greedy, within shard).
-LESSON_SYNTH_CLUSTER_THRESHOLD = float(os.environ.get("LESSON_SYNTH_CLUSTER_THRESHOLD", "0.82"))
+# Tuned empirically (#158 dry-run, 81 lessons): the lessons are diverse (median
+# pairwise cosine ~0.48, max ~0.72), so 0.82 merged nothing. 0.68 captures the real
+# near-duplicates (5 faithful clusters across rishuy+betterment, all clearing the
+# 0.80 drift gate) without merging weakly-related lessons; the drift gate is the
+# quality backstop below this.
+LESSON_SYNTH_CLUSTER_THRESHOLD = float(os.environ.get("LESSON_SYNTH_CLUSTER_THRESHOLD", "0.68"))

 # Mistral OCR (fallback for scanned PDFs — replaces Google Cloud Vision)
 MISTRAL_API_KEY = os.environ.get("MISTRAL_API_KEY", "")
Author	SHA1	Message	Date
chaim	5e7097399f	Merge pull request 'tune(learning): סף-אשכול סינתזת-לקחים 0.82 → 0.68 (#158 )' (#349 ) from worktree-synth-threshold-068 into main All checks were successful Build & Deploy / build-and-deploy (push) Successful in 1m30s Details G12 Leak-Guard / leak-guard (push) Successful in 6s Details Lint — undefined names / undefined-names (push) Successful in 11s Details	2026-06-28 22:45:11 +00:00
Chaim	6c3c6a16c2	tune(learning): lesson-synthesis cluster threshold 0.82 → 0.68 (#158 ) All checks were successful G12 Leak-Guard / leak-guard (pull_request) Successful in 4s Details Lint — undefined names / undefined-names (pull_request) Successful in 10s Details כיול אמפירי (dry-run על 81 לקחים): הלקחים מגוונים (חציון-דמיון ~0.48, מקס' ~0.72), כך ש-0.82 מיזג 0 אשכולות. 0.68 תופס 5 קונסולידציות אמיתיות (רישוי+השבחה+מבנה), כולן עברו את שער-ה-drift 0.80 (0.848–0.916) → נאמנות נשמרת. שער-ה-drift נשאר רשת-הביטחון מתחת לסף. Invariants: INV-LRN8 (כיול-סף בלבד; המנגנון/השערים ללא שינוי). depends-on #158. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>	2026-06-28 22:44:55 +00:00