vision/pc_vision/viewer.py

"""
ESP32 camera viewer — local (on-device) HSV detection only.

Tunes colour / method / perf on the ESP over WebSocket; overlays blob from
X-Vision-Data (or /vision in detect-only mode).

Usage:
  .\\.venv\\Scripts\\python.exe viewer.py --ip 192.168.1.178

Arrow keys / WASD also send directions. Enter sends the text box.
Commands go over ws://IP/ws (HTTP kept as fallback).
"""

from __future__ import annotations

import argparse
import json
import threading
import time
import tkinter as tk
from tkinter import colorchooser, scrolledtext, ttk

import cv2
import numpy as np
import requests
from PIL import Image, ImageTk
from websocket import WebSocketApp

from targets import TargetDetector, create_target
from targets.orange_ball import OrangeBallDetector

ESP_METHODS = ("contour", "rect")
ESP_METHOD_LABELS = {"contour": "Contour", "rect": "Rectangle"}


class RobotApp:
    def __init__(
        self,
        ip: str,
        fps: float,
        timeout: float,
        detector: TargetDetector | None,
    ) -> None:
        self.ip = ip
        self.base = f"http://{ip}"
        self.ws_url = f"ws://{ip}/ws"
        self.period = 1.0 / fps if fps > 0 else 0.0
        self.timeout = timeout
        self.detector = detector
        self.session = requests.Session()
        self.session.headers.update(
            {"Cache-Control": "no-cache", "Pragma": "no-cache"}
        )
        self._stop = threading.Event()
        self._photo: ImageTk.PhotoImage | None = None
        self._ws: WebSocketApp | None = None
        self._ws_ready = threading.Event()
        self._ws_lock = threading.Lock()
        self._last_det_log = 0.0
        self._raw_frame: np.ndarray | None = None
        self._frame_lock = threading.Lock()
        self._preview_scale = 1.0
        self._preview_display_size = (640, 480)
        self._preview_src_size = (640, 480)
        self._vision_sync_after_id: str | None = None
        self._last_vision_payload: str | None = None
        self._pending_pick: tuple[int, int] | None = None
        self._applying_remote = False
        self._vision_settings_loaded = False

        self.root = tk.Tk()
        self.root.title(f"linefollower-eye — {ip}")
        self.root.protocol("WM_DELETE_WINDOW", self.on_close)
        self.show_mask = tk.BooleanVar(master=self.root, value=False)
        self.mask_full = tk.BooleanVar(master=self.root, value=False)
        self.morph_roi = tk.BooleanVar(master=self.root, value=True)
        self.coarse_stride = tk.IntVar(master=self.root, value=4)
        self.detect_only = tk.BooleanVar(master=self.root, value=False)
        self.tol_var = tk.IntVar(master=self.root, value=10)
        self.eyedropper_on = tk.BooleanVar(master=self.root, value=False)

        if isinstance(self.detector, OrangeBallDetector):
            # ESP only supports contour/rect
            self.detector.set_method("contour")

        main = ttk.Frame(self.root, padding=8)
        main.grid(row=0, column=0, sticky="nsew")
        self.root.rowconfigure(0, weight=1)
        self.root.columnconfigure(0, weight=1)
        main.rowconfigure(0, weight=1)
        main.columnconfigure(0, weight=1)

        self.video = ttk.Label(main, cursor="arrow")
        self.video.grid(row=0, column=0, sticky="nsew", padx=(0, 8))
        self.video.bind("<Button-1>", self._on_video_click)

        side = ttk.Frame(main)
        side.grid(row=0, column=1, sticky="ns")

        ttk.Label(side, text="Drive").grid(row=0, column=0, columnspan=3, pady=(0, 4))
        ttk.Button(side, text="↑", width=4, command=lambda: self.send_cmd("up")).grid(
            row=1, column=1, padx=2, pady=2
        )
        ttk.Button(side, text="←", width=4, command=lambda: self.send_cmd("left")).grid(
            row=2, column=0, padx=2, pady=2
        )
        ttk.Button(side, text="●", width=4, command=lambda: self.send_cmd("stop")).grid(
            row=2, column=1, padx=2, pady=2
        )
        ttk.Button(side, text="→", width=4, command=lambda: self.send_cmd("right")).grid(
            row=2, column=2, padx=2, pady=2
        )
        ttk.Button(side, text="↓", width=4, command=lambda: self.send_cmd("down")).grid(
            row=3, column=1, padx=2, pady=2
        )

        row = 4
        if isinstance(self.detector, OrangeBallDetector):
            row = self._build_tune_panel(side, start_row=row)

        ttk.Label(side, text="Terminal").grid(
            row=row, column=0, columnspan=3, sticky="w", pady=(16, 4)
        )
        self.log = scrolledtext.ScrolledText(
            side, width=36, height=12, state="disabled", wrap="word"
        )
        self.log.grid(row=row + 1, column=0, columnspan=3, sticky="nsew")
        side.rowconfigure(row + 1, weight=1)

        entry_row = ttk.Frame(side)
        entry_row.grid(row=row + 2, column=0, columnspan=3, sticky="ew", pady=(6, 0))
        self.entry = ttk.Entry(entry_row)
        self.entry.pack(side="left", fill="x", expand=True)
        self.entry.bind("<Return>", lambda _e: self.send_text())
        ttk.Button(entry_row, text="Send", command=self.send_text).pack(
            side="left", padx=(6, 0)
        )

        self.root.bind("<Up>", lambda _e: self.send_cmd("up"))
        self.root.bind("<Down>", lambda _e: self.send_cmd("down"))
        self.root.bind("<Left>", lambda _e: self.send_cmd("left"))
        self.root.bind("<Right>", lambda _e: self.send_cmd("right"))
        self.root.bind("<space>", lambda _e: self.send_cmd("stop"))
        self.root.bind("w", lambda _e: self.send_cmd("up"))
        self.root.bind("s", lambda _e: self.send_cmd("down"))
        self.root.bind("a", lambda _e: self.send_cmd("left"))
        self.root.bind("d", lambda _e: self.send_cmd("right"))

        self.append_log(f"HTTP {self.base}  WS {self.ws_url}")
        if self.detector is not None:
            self.append_log(f"Target: {self.detector.name} (ESP local vision)")
            if isinstance(self.detector, OrangeBallDetector):
                self.append_log("Eyedropper / Pick colour → sync HSV to ESP")
        else:
            self.append_log("Target: off (drive + stream only)")
        self._ws_thread = threading.Thread(target=self._ws_loop, daemon=True)
        self._ws_thread.start()
        self._video_thread = threading.Thread(target=self._video_loop, daemon=True)
        self._video_thread.start()

    def _build_tune_panel(self, parent: ttk.Frame, start_row: int) -> int:
        if not isinstance(self.detector, OrangeBallDetector):
            return start_row

        h_tol, _, _ = self.detector.get_tolerance()
        self.tol_var.set(h_tol)

        ttk.Label(parent, text="ESP local vision", font=("", 9, "bold")).grid(
            row=start_row, column=0, columnspan=3, sticky="w", pady=(16, 4)
        )
        start_row += 1

        ttk.Label(parent, text="Target colour").grid(
            row=start_row, column=0, columnspan=3, sticky="w", pady=(8, 4)
        )
        start_row += 1

        row = ttk.Frame(parent)
        row.grid(row=start_row, column=0, columnspan=3, sticky="ew", pady=2)
        self.swatch = tk.Canvas(row, width=36, height=24, highlightthickness=1)
        self.swatch.pack(side="left")
        ttk.Button(row, text="Pick colour…", command=self._pick_colour).pack(
            side="left", padx=(8, 4)
        )
        self.eyedrop_btn = ttk.Checkbutton(
            row,
            text="Eyedropper",
            variable=self.eyedropper_on,
            command=self._on_eyedropper_toggle,
        )
        self.eyedrop_btn.pack(side="left")
        start_row += 1

        method_row = ttk.Frame(parent)
        method_row.grid(row=start_row, column=0, columnspan=3, sticky="ew", pady=2)
        ttk.Label(method_row, text="Method", width=11).pack(side="left")
        self.method_var = tk.StringVar(
            master=self.root,
            value=ESP_METHOD_LABELS["contour"],
        )
        self.method_combo = ttk.Combobox(
            method_row,
            textvariable=self.method_var,
            values=[ESP_METHOD_LABELS[m] for m in ESP_METHODS],
            state="readonly",
            width=16,
        )
        self.method_combo.pack(side="left", fill="x", expand=True, padx=4)
        self.method_combo.bind("<<ComboboxSelected>>", self._on_method_change)
        start_row += 1

        self.show_mask_btn = ttk.Checkbutton(
            parent,
            text="Show mask",
            variable=self.show_mask,
            command=self._on_show_mask_change,
        )
        self.show_mask_btn.grid(row=start_row, column=0, columnspan=3, sticky="w")
        start_row += 1

        perf = ttk.LabelFrame(parent, text="ESP32 perf / detect", padding=4)
        perf.grid(row=start_row, column=0, columnspan=3, sticky="ew", pady=(8, 4))
        start_row += 1

        stride_row = ttk.Frame(perf)
        stride_row.pack(fill="x", pady=2)
        ttk.Label(stride_row, text="Coarse stride", width=12).pack(side="left")
        self.stride_combo = ttk.Combobox(
            stride_row,
            values=("2", "4"),
            state="readonly",
            width=4,
        )
        self.stride_combo.set(str(self.coarse_stride.get()))
        self.stride_combo.pack(side="left", padx=4)
        self.stride_combo.bind("<<ComboboxSelected>>", self._on_esp_perf_change)

        opt_row = ttk.Frame(perf)
        opt_row.pack(fill="x", pady=2)
        ttk.Checkbutton(
            opt_row,
            text="Full-frame mask (slow)",
            variable=self.mask_full,
            command=self._on_esp_perf_change,
        ).pack(side="left")
        ttk.Checkbutton(
            opt_row,
            text="ROI morph",
            variable=self.morph_roi,
            command=self._on_esp_perf_change,
        ).pack(side="left", padx=(8, 0))

        detect_row = ttk.Frame(perf)
        detect_row.pack(fill="x", pady=2)
        ttk.Checkbutton(
            detect_row,
            text="Detect only (/vision, no video)",
            variable=self.detect_only,
            command=self._on_detect_only_change,
        ).pack(side="left")

        self.esp_stats_label = ttk.Label(
            perf,
            text="timing: —",
            wraplength=240,
            justify="left",
        )
        self.esp_stats_label.pack(fill="x", pady=(4, 0))

        tol_row = ttk.Frame(parent)
        tol_row.grid(row=start_row, column=0, columnspan=3, sticky="ew", pady=2)
        ttk.Label(tol_row, text="Tolerance", width=11).pack(side="left")
        self.tol_label = ttk.Label(tol_row, width=4)
        self.tol_scale = ttk.Scale(
            tol_row,
            from_=2,
            to=40,
            variable=self.tol_var,
            orient="horizontal",
            command=self._on_tol_change,
        )
        self.tol_scale.pack(side="left", fill="x", expand=True, padx=4)
        self.tol_label.pack(side="left")
        self._on_tol_change(str(self.tol_var.get()))
        start_row += 1

        self._update_swatch()
        return start_row

    def _method_key_from_label(self, label: str) -> str | None:
        for key, text in ESP_METHOD_LABELS.items():
            if text == label:
                return key
        return None

    def _on_esp_perf_change(self, _event: object | None = None) -> None:
        if self._applying_remote:
            return
        try:
            self.coarse_stride.set(int(self.stride_combo.get()))
        except ValueError:
            self.coarse_stride.set(4)
            self.stride_combo.set("4")
        self._last_vision_payload = None
        self._sync_vision_to_esp()

    def _on_detect_only_change(self) -> None:
        mode = "detect-only (/vision)" if self.detect_only.get() else "JPEG + overlay"
        self.append_log(f"Stream mode → {mode}")

    def _parse_vision_data_str(self, raw: str) -> dict[str, float | int | bool]:
        out: dict[str, float | int | bool] = {}
        if not raw or "found=" not in raw:
            return out
        if any(ord(ch) < 32 and ch not in "\t\r\n" for ch in raw):
            return out
        for part in raw.split(";"):
            if "=" not in part:
                continue
            key, val = part.split("=", 1)
            try:
                if key == "found":
                    out["found"] = val == "1"
                elif key == "area":
                    out[key] = int(float(val))
                else:
                    out[key] = float(val)
            except ValueError:
                continue
        return out

    def _parse_vision_json(self, data: dict) -> dict[str, float | int | bool]:
        return {
            "found": bool(data.get("found")),
            "cx": float(data.get("cx", 0)),
            "cy": float(data.get("cy", 0)),
            "r": float(data.get("radius", 0)),
            "area": int(data.get("area", 0)),
            "total": float(data.get("total_ms", 0)),
            "detect": float(data.get("detect_ms", 0)),
            "coarse": float(data.get("coarse_ms", 0)),
            "refine": float(data.get("refine_ms", 0)),
            "morph": float(data.get("morph_ms", 0)),
            "mask": float(data.get("mask_ms", 0)),
            "jpeg": float(data.get("jpeg_ms", 0)),
        }

    def _format_esp_stats_line(self, stats: dict[str, float | int | bool]) -> str:
        parts: list[str] = []
        for key, label in (
            ("total", "total"),
            ("detect", "det"),
            ("coarse", "coarse"),
            ("refine", "refine"),
            ("morph", "morph"),
            ("mask", "mask"),
            ("jpeg", "jpeg"),
        ):
            val = stats.get(key)
            if isinstance(val, (int, float)):
                if key in ("morph", "mask") and val <= 0.05:
                    continue
                if key == "total":
                    parts.append(f"{label} {val:.0f}ms")
                else:
                    parts.append(f"{label} {val:.0f}")

        blob = ""
        if stats.get("found"):
            cx = stats.get("cx", "?")
            cy = stats.get("cy", "?")
            r = stats.get("r", stats.get("radius", "?"))
            blob = f" | blob @({cx},{cy}) r={r}"
        return "timing: " + (" ".join(parts) if parts else "—") + blob

    def _annotate_esp_target(
        self, frame_bgr: np.ndarray, stats: dict[str, float | int | bool]
    ) -> np.ndarray:
        if not stats.get("found"):
            return frame_bgr
        out = frame_bgr.copy()
        cx = int(float(stats.get("cx", 0)))
        cy = int(float(stats.get("cy", 0)))
        radius = stats.get("r", stats.get("radius", 0))
        r = max(4, int(float(radius)))
        cv2.circle(out, (cx, cy), r, (0, 255, 0), 2)
        cv2.drawMarker(out, (cx, cy), (0, 255, 0), cv2.MARKER_CROSS, 12, 2)
        return out

    def _seed_display_frame(self, capture_url: str) -> bool:
        """Fetch one JPEG for overlay when detect-only has no cached frame yet."""
        try:
            resp = self.session.get(capture_url, timeout=self._capture_timeout())
            resp.raise_for_status()
            buf = np.frombuffer(resp.content, dtype=np.uint8)
            frame = cv2.imdecode(buf, cv2.IMREAD_COLOR)
            if frame is None:
                return False
            with self._frame_lock:
                self._raw_frame = frame.copy()
            return True
        except requests.RequestException:
            return False

    def _refresh_detect_overlay(self, esp_stats: dict[str, float | int | bool]) -> None:
        with self._frame_lock:
            frame = None if self._raw_frame is None else self._raw_frame.copy()
        if frame is None:
            return
        display = self._annotate_esp_target(frame, esp_stats)
        self.root.after(0, self._show_frame, display)

    def _apply_esp_stats(self, stats: dict[str, float | int | bool]) -> None:
        line = self._format_esp_stats_line(stats)
        self.root.after(0, self._update_esp_stats_label, line)

        now = time.perf_counter()
        if stats.get("found") and now - self._last_det_log > 1.0:
            self.append_log(
                f"ESP blob @ ({stats.get('cx')},{stats.get('cy')}) "
                f"r={stats.get('r', stats.get('radius'))} area={stats.get('area')}"
            )
            self._last_det_log = now

    def _apply_vision_headers(self, headers: object) -> dict[str, float | int | bool] | None:
        getter = getattr(headers, "get", None)
        if getter is None:
            return None
        raw = getter("X-Vision-Data") or getter("x-vision-data")
        if not raw:
            return None
        stats = self._parse_vision_data_str(str(raw))
        if not stats:
            return None
        self._apply_esp_stats(stats)
        return stats

    def _update_esp_stats_label(self, text: str) -> None:
        if hasattr(self, "esp_stats_label"):
            self.esp_stats_label.configure(text=text)

    def _request_vision_settings(self) -> None:
        """Pull current ESP vision params without writing (HTTP, WS fallback)."""

        def _worker() -> None:
            data: dict | None = None
            try:
                resp = self.session.get(
                    f"{self.base}/vision/settings",
                    timeout=self.timeout,
                )
                resp.raise_for_status()
                parsed = resp.json()
                if isinstance(parsed, dict) and parsed.get("ok"):
                    data = parsed
            except (requests.RequestException, ValueError) as exc:
                self.append_log(f"! vision settings HTTP: {exc}")

            if data is None:
                if not self._ws_send({"type": "vision", "get": True}):
                    self.append_log("! vision get failed (WS not connected)")
                return

            self.root.after(0, self._apply_vision_from_esp, data)

        threading.Thread(target=_worker, daemon=True).start()

    def _build_vision_payload(self, *, consume_pick: bool = True) -> dict | None:
        if not isinstance(self.detector, OrangeBallDetector):
            return None
        if not hasattr(self, "method_var"):
            return None
        h, s, v = self.detector.get_hsv_center()
        h_tol, s_tol, v_tol = self.detector.get_tolerance()
        key = self._method_key_from_label(self.method_var.get()) or "contour"
        if key not in ESP_METHODS:
            key = "contour"
        payload: dict = {
            "type": "vision",
            "local": True,
            "show_mask": self.show_mask.get(),
            "method": key,
            "h": h,
            "s": s,
            "v": v,
            "h_tol": h_tol,
            "s_tol": s_tol,
            "v_tol": v_tol,
            "coarse_stride": int(self.coarse_stride.get()),
            "min_area": int(self.detector.min_area),
            "mask_full": self.mask_full.get(),
            "morph_roi": self.morph_roi.get(),
        }
        if consume_pick and self._pending_pick is not None:
            payload["pick_x"], payload["pick_y"] = self._pending_pick
            self._pending_pick = None
        return payload

    def _set_swatch_bgr(self, b: int, g: int, r: int) -> None:
        if not hasattr(self, "swatch"):
            return
        color = f"#{r:02x}{g:02x}{b:02x}"
        self.swatch.configure(bg=color)
        self.swatch.delete("all")
        self.swatch.create_rectangle(0, 0, 36, 24, fill=color, outline="")

    def _apply_vision_from_esp(self, data: dict) -> None:
        """Populate UI + detector from a vision ack JSON."""
        if not isinstance(self.detector, OrangeBallDetector):
            return
        if not data.get("ok"):
            err = data.get("error", "unknown")
            self.append_log(f"! vision settings rejected: {err}")
            return

        self._applying_remote = True
        try:
            if "h" in data and "s" in data and "v" in data:
                self.detector.set_hsv_center(
                    int(data["h"]), int(data["s"]), int(data["v"])
                )
            if "h_tol" in data or "s_tol" in data or "v_tol" in data:
                self.detector.set_tolerance(
                    h_tol=int(data["h_tol"]) if "h_tol" in data else None,
                    s_tol=int(data["s_tol"]) if "s_tol" in data else None,
                    v_tol=int(data["v_tol"]) if "v_tol" in data else None,
                )
                if "h_tol" in data and hasattr(self, "tol_label"):
                    ht = int(data["h_tol"])
                    # Avoid Scale command overwriting s_tol/v_tol with h*6
                    if hasattr(self, "tol_scale"):
                        self.tol_scale.configure(command="")
                    self.tol_var.set(ht)
                    self.tol_label.configure(text=str(ht))
                    if hasattr(self, "tol_scale"):
                        self.tol_scale.configure(command=self._on_tol_change)
            if "min_area" in data:
                self.detector.min_area = float(int(data["min_area"]))
            method = str(data.get("method", "contour"))
            if method not in ESP_METHODS:
                method = "contour"
            self.detector.set_method(method)
            if hasattr(self, "method_var"):
                self.method_var.set(ESP_METHOD_LABELS[method])
            if "show_mask" in data:
                self.show_mask.set(bool(data["show_mask"]))
            if "mask_full" in data:
                self.mask_full.set(bool(data["mask_full"]))
            if "morph_roi" in data:
                self.morph_roi.set(bool(data["morph_roi"]))
            if "coarse_stride" in data and hasattr(self, "stride_combo"):
                stride = int(data["coarse_stride"])
                if stride not in (2, 4):
                    stride = 4
                self.coarse_stride.set(stride)
                self.stride_combo.set(str(stride))
            # Prefer ESP display BGR (eyedrop pixel / stored swatch) over HSV→BGR
            if all(k in data for k in ("b", "g", "r")):
                self._set_swatch_bgr(int(data["b"]), int(data["g"]), int(data["r"]))
            elif hasattr(self, "swatch"):
                self._update_swatch()
        finally:
            self._applying_remote = False

        payload = self._build_vision_payload(consume_pick=False)
        if payload is not None:
            self._last_vision_payload = json.dumps(
                payload, separators=(",", ":"), sort_keys=True
            )

        first = not self._vision_settings_loaded
        self._vision_settings_loaded = True
        if first:
            h, s, v = self.detector.get_hsv_center()
            ht, st, vt = self.detector.get_tolerance()
            bgr = ""
            if all(k in data for k in ("b", "g", "r")):
                bgr = f" BGR=({int(data['b'])},{int(data['g'])},{int(data['r'])})"
            self.append_log(
                f"ESP settings loaded: HSV=({h},{s},{v}){bgr} "
                f"tol=({ht},{st},{vt}) method={self.detector.get_method()} "
                f"stride={self.coarse_stride.get()} mask={self.show_mask.get()} "
                f"local={bool(data.get('local'))}"
            )

    def _sync_vision_to_esp(self, delay_ms: int = 150) -> None:
        if self._applying_remote:
            return
        if not self._vision_settings_loaded:
            return
        if not isinstance(self.detector, OrangeBallDetector):
            return
        if self._vision_sync_after_id is not None:
            self.root.after_cancel(self._vision_sync_after_id)
            self._vision_sync_after_id = None

        def _do() -> None:
            self._vision_sync_after_id = None
            if self._applying_remote or not self._vision_settings_loaded:
                return
            payload = self._build_vision_payload()
            if payload is None:
                return
            blob = json.dumps(payload, separators=(",", ":"), sort_keys=True)
            if blob == self._last_vision_payload:
                return
            self._last_vision_payload = blob

            def _worker() -> None:
                if not self._ws_send(payload):
                    self.append_log("! vision sync failed (WS not connected)")

            threading.Thread(target=_worker, daemon=True).start()

        self._vision_sync_after_id = self.root.after(delay_ms, _do)

    def _capture_timeout(self) -> float:
        if self.show_mask.get():
            return max(self.timeout, 15.0)
        return max(self.timeout, 6.0)

    def _capture_period(self) -> float:
        if self.show_mask.get():
            return max(self.period, 0.5)
        return max(self.period, 0.2)

    def _on_show_mask_change(self) -> None:
        if self._applying_remote:
            return
        self._last_vision_payload = None
        self._sync_vision_to_esp()

    def _on_method_change(self, _event: object | None = None) -> None:
        if self._applying_remote:
            return
        if not isinstance(self.detector, OrangeBallDetector):
            return
        label = self.method_var.get()
        key = self._method_key_from_label(label)
        if key is None or key not in ESP_METHODS:
            return
        self.detector.set_method(key)
        self.append_log(f"method → {label}")
        self._sync_vision_to_esp()

    def _on_eyedropper_toggle(self) -> None:
        if self.eyedropper_on.get():
            self.video.configure(cursor="crosshair")
            self.append_log("Eyedropper on — click the video")
        else:
            self.video.configure(cursor="arrow")

    def _update_swatch(self) -> None:
        if not isinstance(self.detector, OrangeBallDetector):
            return
        b, g, r = self.detector.get_bgr_center()
        color = f"#{r:02x}{g:02x}{b:02x}"
        self.swatch.configure(bg=color)
        self.swatch.delete("all")
        self.swatch.create_rectangle(0, 0, 36, 24, fill=color, outline="")

    def _pick_colour(self) -> None:
        if not isinstance(self.detector, OrangeBallDetector):
            return
        self.eyedropper_on.set(False)
        self._on_eyedropper_toggle()
        b, g, r = self.detector.get_bgr_center()
        result = colorchooser.askcolor(
            color=(r, g, b), title="Target ball colour", parent=self.root
        )
        if not result or result[0] is None:
            return
        r2, g2, b2 = (int(v) for v in result[0])
        self.detector.set_from_bgr(b2, g2, r2)
        self._update_swatch()
        self.append_log(f"colour set RGB=({r2},{g2},{b2})")
        self._sync_vision_to_esp()

    def _on_tol_change(self, _value: str) -> None:
        if not isinstance(self.detector, OrangeBallDetector):
            return
        tol = int(round(float(self.tol_var.get())))
        if hasattr(self, "tol_label"):
            self.tol_label.configure(text=str(tol))
        if self._applying_remote:
            return
        self.detector.set_tolerance(h_tol=tol, s_tol=tol * 6, v_tol=tol * 6)
        self._sync_vision_to_esp()

    def _video_click_to_frame(self, event: tk.Event) -> tuple[int, int] | None:  # type: ignore[type-arg]
        """Map a click on the video label to pixel coords in the source frame."""
        scale = self._preview_scale
        disp_w, disp_h = self._preview_display_size
        src_w, src_h = self._preview_src_size
        if scale <= 0 or disp_w <= 0 or disp_h <= 0:
            return None

        widget_w = max(1, self.video.winfo_width())
        widget_h = max(1, self.video.winfo_height())
        offset_x = max(0, (widget_w - disp_w) // 2)
        offset_y = max(0, (widget_h - disp_h) // 2)

        ix = event.x - offset_x
        iy = event.y - offset_y
        if not (0 <= ix < disp_w and 0 <= iy < disp_h):
            return None

        x = int(round(ix / scale))
        y = int(round(iy / scale))
        x = max(0, min(x, src_w - 1))
        y = max(0, min(y, src_h - 1))
        return x, y

    def _on_video_click(self, event: tk.Event) -> None:  # type: ignore[type-arg]
        if not isinstance(self.detector, OrangeBallDetector):
            return
        if not self.eyedropper_on.get():
            return
        if self.show_mask.get():
            self.append_log("! turn off Show mask to pick colour from live image")
            return
        with self._frame_lock:
            frame = None if self._raw_frame is None else self._raw_frame.copy()
        if frame is None:
            return
        pt = self._video_click_to_frame(event)
        if pt is None:
            self.append_log("! eyedropper: click on the video image")
            return
        x, y = pt
        b, g, r = (int(v) for v in frame[y, x])
        self.detector.set_from_bgr(b, g, r)
        self._update_swatch()
        self.append_log(f"eyedrop @ ({x},{y}) RGB=({r},{g},{b})")
        self._pending_pick = (x, y)
        self._last_vision_payload = None
        self._sync_vision_to_esp(delay_ms=0)
        self.eyedropper_on.set(False)
        self._on_eyedropper_toggle()

    def append_log(self, line: str) -> None:
        def _do() -> None:
            self.log.configure(state="normal")
            self.log.insert("end", line + "\n")
            self.log.see("end")
            self.log.configure(state="disabled")

        self.root.after(0, _do)

    def _ws_loop(self) -> None:
        while not self._stop.is_set():
            self._ws_ready.clear()

            def on_open(_ws: WebSocketApp) -> None:
                self._ws_ready.set()
                self.append_log("WS connected — fetching ESP vision settings")
                self._vision_settings_loaded = False
                self.root.after(200, self._request_vision_settings)

            def on_message(_ws: WebSocketApp, message: str) -> None:
                self.append_log(f"< {message}")
                try:
                    data = json.loads(message)
                except (TypeError, json.JSONDecodeError):
                    return
                if isinstance(data, dict) and data.get("type") == "vision":
                    self.root.after(0, self._apply_vision_from_esp, data)

            def on_error(_ws: WebSocketApp, error: object) -> None:
                self.append_log(f"! WS error: {error}")

            def on_close(_ws: WebSocketApp, *_args: object) -> None:
                self._ws_ready.clear()
                self.append_log("WS closed")

            ws = WebSocketApp(
                self.ws_url,
                on_open=on_open,
                on_message=on_message,
                on_error=on_error,
                on_close=on_close,
            )
            with self._ws_lock:
                self._ws = ws
            try:
                ws.run_forever(ping_interval=20, ping_timeout=10)
            except Exception as exc:  # noqa: BLE001
                self.append_log(f"! WS run failed: {exc}")
            with self._ws_lock:
                self._ws = None
            if not self._stop.is_set():
                time.sleep(1.0)

    def _ws_send(self, payload: dict) -> bool:
        data = json.dumps(payload, separators=(",", ":"))
        with self._ws_lock:
            ws = self._ws
            ready = self._ws_ready.is_set()
        if not ws or not ready:
            return False
        try:
            ws.send(data)
            return True
        except Exception as exc:  # noqa: BLE001
            self.append_log(f"! WS send failed: {exc}")
            return False

    def send_cmd(self, direction: str) -> None:
        def _worker() -> None:
            if self._ws_send({"type": "cmd", "dir": direction}):
                return
            try:
                r = self.session.get(
                    f"{self.base}/cmd",
                    params={"dir": direction},
                    timeout=self.timeout,
                )
                self.append_log(f"< {r.text.strip()}")
            except requests.RequestException as exc:
                self.append_log(f"! cmd {direction} failed: {exc}")

        threading.Thread(target=_worker, daemon=True).start()

    def send_text(self) -> None:
        text = self.entry.get().strip()
        if not text:
            return
        self.entry.delete(0, "end")

        def _worker() -> None:
            if self._ws_send({"type": "msg", "text": text}):
                return
            try:
                r = self.session.post(
                    f"{self.base}/msg",
                    data=text.encode("utf-8"),
                    headers={"Content-Type": "text/plain; charset=utf-8"},
                    timeout=self.timeout,
                )
                self.append_log(f"< {r.text.strip()}")
            except requests.RequestException as exc:
                self.append_log(f"! send failed: {exc}")

        threading.Thread(target=_worker, daemon=True).start()

    def _show_frame(self, frame_bgr: np.ndarray) -> None:
        preview_w = 640
        h, w = frame_bgr.shape[:2]
        scale = preview_w / float(w)
        preview_h = max(1, int(round(h * scale)))
        self._preview_scale = scale
        self._preview_display_size = (preview_w, preview_h)
        self._preview_src_size = (w, h)
        resized = cv2.resize(frame_bgr, (preview_w, preview_h), interpolation=cv2.INTER_AREA)
        rgb = cv2.cvtColor(resized, cv2.COLOR_BGR2RGB)
        image = Image.fromarray(rgb)
        photo = ImageTk.PhotoImage(image=image)
        self._photo = photo
        self.video.configure(image=photo)

    def _video_loop(self) -> None:
        capture_url = f"{self.base}/capture"
        vision_url = f"{self.base}/vision"
        while not self._stop.is_set():
            t0 = time.perf_counter()
            try:
                esp_stats: dict[str, float | int | bool] | None = None
                if self.detect_only.get():
                    with self._frame_lock:
                        need_seed = self._raw_frame is None
                    if need_seed:
                        self._seed_display_frame(capture_url)
                    resp = self.session.get(vision_url, timeout=self._capture_timeout())
                    resp.raise_for_status()
                    esp_stats = self._parse_vision_json(resp.json())
                    self._apply_esp_stats(esp_stats)
                    self._refresh_detect_overlay(esp_stats)
                else:
                    resp = self.session.get(capture_url, timeout=self._capture_timeout())
                    resp.raise_for_status()
                    esp_stats = self._apply_vision_headers(resp.headers)
                    buf = np.frombuffer(resp.content, dtype=np.uint8)
                    frame = cv2.imdecode(buf, cv2.IMREAD_COLOR)
                    if frame is not None:
                        with self._frame_lock:
                            self._raw_frame = frame.copy()
                        display = frame
                        if esp_stats is not None:
                            display = self._annotate_esp_target(frame, esp_stats)
                        self.root.after(0, self._show_frame, display)
            except requests.RequestException as exc:
                self.append_log(f"! video: {exc}")
                time.sleep(1.0)

            period = self._capture_period()
            if period > 0:
                elapsed = time.perf_counter() - t0
                sleep_s = period - elapsed
                if sleep_s > 0:
                    time.sleep(sleep_s)

    def on_close(self) -> None:
        self._stop.set()
        with self._ws_lock:
            if self._ws is not None:
                try:
                    self._ws.close()
                except Exception:  # noqa: BLE001
                    pass
        self.session.close()
        self.root.destroy()

    def run(self) -> None:
        self.root.mainloop()


def main() -> int:
    parser = argparse.ArgumentParser(
        description="ESP32 camera + local (on-device) vision UI"
    )
    parser.add_argument("--ip", default="192.168.1.178")
    parser.add_argument("--fps", type=float, default=10.0)
    parser.add_argument("--timeout", type=float, default=2.0)
    parser.add_argument(
        "--target",
        default="orange_ball",
        help="Vision target module: orange_ball | none",
    )
    args = parser.parse_args()

    detector = create_target(args.target)
    app = RobotApp(
        ip=args.ip,
        fps=args.fps,
        timeout=args.timeout,
        detector=detector,
    )
    app.run()
    return 0


if __name__ == "__main__":
    raise SystemExit(main())