refactor: single predict.py entrypoint (production + CLI), archive experiments

docs/architecture.md

@@ -112,14 +112,20 @@ stats panel (Loading / Unloading / Net / last-3-batch history), and a bottom sta
   (`batch_id,start,end,duration,loading,unloading,net`) and `sack_events.csv`
   (`timestamp,batch_id,sack_crossed,T-<id>,direction`).
 
-## Legacy pipelines (same concepts, different implementation)
+## Production pipeline (`predict.py`)
+
+Same stage order as v3 but implemented with shapely polygons (`zones.json`:
+pallet/truck/counting), visual-similarity + Re-ID registries, static-frame debounce,
+SQLite persistence, and live-frame publishing to `/dev/shm` for the dashboard. Uses the
+repo `src/` modules (imports at `predict.py`). Zero CLI flags = systemd behaviour;
+dev flags (`--source`, `--no-dashboard`, …) are additive overrides — see
+`predict.py --help` and `docs/scripts.md`. (`init_db()` and the `http.server` imports
+are dead code — persistence goes through `finalize_batch`/`save_active_batch_state`,
+dashboard integration is file-based.)
+
+## Retired (`archive/`, not imported)
 
-- **`predict.py`** — production Jetson pipeline. Same stage order but implemented with
-  shapely polygons (`zones.json`: pallet/truck/counting), visual-similarity + Re-ID
-  registries, static-frame debounce, SQLite persistence, live-frame publishing to
-  `/dev/shm`, and an embedded HTTP status server. Uses the repo `src/` modules
-  (imports at `predict.py:19-25`).
 - **`predict_new.py`** — duo-model state machine (`WAITING_FOR_TRUCK → COUNTING_SACKS →
-  TRUCK_LEAVING`) with hardcoded default zones, perspective-area filtering and HUD.
+  TRUCK_LEAVING`) with hardcoded default zones, perspective filtering and HUD.
 - **`rpo_iki/`** — alternate geometric engine (`count.py`: `LineCounter` with approach
   depth, burst/cooldown dedup, blind-truck overlay) + `predict_rpo_iki.py` runner.

docs/scripts.md

@@ -1,17 +1,17 @@
 # Scripts reference
 
-## Entry points (pipelines)
-- `predict.py` — **production** Jetson pipeline: combined sack+truck model + repo `src/`
-  modules, shapely zones, visual-similarity/Re-ID debounce, SQLite, live-frame publish,
-  embedded HTTP status. Run by `karung-counter.service`.
-- `predict_new.py` — duo-model state machine (`WAITING_FOR_TRUCK → COUNTING_SACKS →
-  TRUCK_LEAVING`), hardcoded zones + `zones.json` override, perspective filtering, HUD.
-  Standalone experiment path.
-- `src/main.py` — clean v3 pipeline (`python -m src.main --source VID --env .env`).
-- `rpo_iki/predict_rpo_iki.py` — runner for the alternate `rpo_iki/count.py` geometric
-  engine (approach-depth + burst/cooldown dedup, blind-truck overlay).
-- `simple_predict.py` — minimal CPU demo: `best.pt`, stride 3, `imgsz=320`, OpenCV window.
-- `ini.py` — tiny import/smoke check.
+## Entry points
+- `predict.py` — **single entrypoint: production service (zero flags) + dev CLI.**
+  Combined sack+truck model + repo `src/` modules, shapely zones,
+  visual-similarity/Re-ID debounce, SQLite, live-frame publish. Run by
+  `karung-counter.service`. Flags: `--source/--env/--model/--output-dir/
+  --output-json/--sack-conf/--truck-conf/--box-conf(placeholder)/
+  --batch-timeout/--max-frames/--no-dashboard/--no-db` (see `predict.py --help`).
+- `src/main.py` — deprecated shim: prints a pointer to `predict.py`, still runs the
+  old v3 loop for backward compat.
+- `archive/` — retired: `predict_new.py` (duo-model state machine),
+  `rpo_iki/` (alternate geometric engine + `predict_rpo_iki.py`), `simple_predict.py`
+  (minimal CPU demo), `ini.py`, check/merge/test scripts. Not imported anywhere.
 
 ## Calibration & data tools
 - `get_coordinates.py` / `get_calib_frame.py` — zone calibration (`calib_frame.jpg`,