Add transitional learning detect with in-dialog mask review.
Deploy Development / deploy (push) Successful in 57s
Test Suite / pytest-backend (push) Failing after 2m57s
Test Suite / smoke-dev (push) Successful in 0s
Test Suite / frontend-build (push) Successful in 16s

Senses grow from confirmations instead of a word list, so a later local pipeline can decide homonyms on a short passage. Default stays semantic. Local detect waits longer, and the reverse-proxy timeout is documented so Dev does not 504 first.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
LarsandCursor committed 2026-09-09 08:34:43 +02:00
1 parent 59921fc5b5
commit fa1c11c33e
29 files changed
+1135 -45

No files matched your search

@@ -0,0 +1,23 @@
-- Transitional learning detect: user-confirmed label senses, pending mask review.
-- Detect output still does not auto-activate identities.
CREATE TABLE IF NOT EXISTS label_senses (
profile_id TEXT NOT NULL REFERENCES profiles(id) ON DELETE CASCADE,
normalized_label TEXT NOT NULL,
identity_hits INTEGER NOT NULL DEFAULT 0,
non_identity_hits INTEGER NOT NULL DEFAULT 0,
updated TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP::text,
PRIMARY KEY (profile_id, normalized_label)
);
CREATE TABLE IF NOT EXISTS pending_mask_reviews (
id TEXT PRIMARY KEY,
profile_id TEXT NOT NULL REFERENCES profiles(id) ON DELETE CASCADE,
conversation_id TEXT NOT NULL REFERENCES conversations(id) ON DELETE CASCADE,
user_message_id TEXT NOT NULL,
payload_json TEXT NOT NULL DEFAULT '{}',
created TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP::text
);
CREATE INDEX IF NOT EXISTS idx_pending_mask_reviews_conv
ON pending_mask_reviews (profile_id, conversation_id);