feat: ASRX hybrid pipeline, identity history, worker fixes, checkpoint system

docs/3002_3003_SEPARATION_STATUS.md

@@ -0,0 +1,70 @@
+# 3002/3003 Schema Separation Status
+
+Date: 2026-05-17
+Status: ✅ Pipeline tables created in `public`; schema incompatibilities remain
+
+## Summary
+
+| Schema | Has pipeline tables | Has auth tables | Used by |
+|--------|-------------------|-----------------|---------|
+| `public` | ✅ (newly created) | ✅ (original) | 3002 (production) — currently using `dev` as workaround |
+| `dev` | ✅ (full, working) | ✅ (synced) | 3003 (playground) |
+
+## What Was Done
+
+### Pipeline tables created in `public` schema (11 tables)
+- `videos`, `chunk`, `chunk_vectors`, `cuts`, `frames`
+- `monitor_jobs`, `processor_results`, `processor_versions`
+- `parent_chunks`, `tkg_edges`, `tkg_nodes`
+
+All include proper sequences, indexes, and constraints matching the `dev` schema.
+
+## Remaining Blockers
+
+### Schema incompatibilities between `dev` and `public`
+
+| Table | dev cols | public cols | Status |
+|-------|---------|------------|--------|
+| identities | 17 | 16 | ⚠️ Different columns (e.g. `name` vs `real_name`/`actor_name`) |
+| face_detections | 16 | 17 | ⚠️ Column count mismatch |
+| identity_bindings | 7 | 8 | ⚠️ Column count mismatch |
+| person_identities | 16 | 15 | ⚠️ Column count mismatch |
+| pre_chunks | 19 | 10 | ⚠️ Significantly different |
+| api_keys | 19 | 19 | ✅ Match |
+| resources | 9 | 9 | ✅ Match |
+| users | 8 | 8 | ✅ Match |
+
+### Identities table key differences
+- `public.identities` uses `real_name` + `actor_name` (old schema)
+- `dev.identities` uses `name` (new unified schema)
+- `dev.identities` has `tmdb_poster`, `file_uuid`, `face_embedding`, `voice_embedding`, `identity_embedding`
+- `public.identities` only has `face_embedding`, `voice_embedding` (no `identity_embedding`)
+
+## Options
+
+### Option A: Full data migration (recommended for later)
+1. Dump data from old public tables
+2. Drop old public tables
+3. Recreate from dev schema DDL
+4. Migrate data with column mapping
+5. Switch 3002 to `DATABASE_SCHEMA=public`
+
+### Option B: Keep current workaround (simplest for now)
+- 3002 continues with `DATABASE_SCHEMA=dev`
+- 3003 uses `DATABASE_SCHEMA=dev`
+- Both share the same schema, but have separate Redis key prefixes + ports
+
+### Option C: Rename dev → public (requires downtime)
+1. Stop all services
+2. Rename `dev` schema to something else
+3. Rename `public` to `public_old`
+4. Rename `dev` to `public`
+5. Update references
+
+## Current Status
+
+✅ Pipeline tables exist in both schemas
+✅ auth tables (users, sessions, jwt_blacklist) exist in both
+✅ Redis key prefixes separate (`momentry:` vs `momentry_dev:`)
+⚠️ 3002 still uses `DATABASE_SCHEMA=dev` workaround
+⛔ Shared tables need migration before 3002 can use `public` schema