Merge c5/fast: parallel look + remote-lane ladder (C5)

src/computer.rs

@@ -479,12 +479,23 @@ pub fn fit(w: u32, h: u32) -> (u32, u32, bool) {
     (o, a, o != w || a != h)
 }
 
+/// The remote lane (C5): when this computer is serving another AnyBot's brain through the
+/// relay, every picture crosses a network first — a tighter ladder (small models read 60-quality
+/// screens fine) buys the remote tasks a quarter less to carry. Local pictures keep 85.
+thread_local! {
+    pub static REMOTE_LANE: std::cell::Cell<bool> = const { std::cell::Cell::new(false) };
+}
+
+/// The remote lane's ladder: it starts where the local one ends.
+pub const REMOTE_QUALITIES: [u8; 3] = [70, 60, 50];
+
 /// j3r: JPEG at 85, 70, 55: the first within the budget, else the smallest if it's under the cap.
 pub fn encode_jpeg(img: &image::RgbImage) -> Option<Vec<u8>> {
+    let ladder = if REMOTE_LANE.with(|c| c.get()) { &REMOTE_QUALITIES[..] } else { &JPEG_QUALITIES[..] };
     let mut best: Option<Vec<u8>> = None;
-    for q in JPEG_QUALITIES {
+    for q in ladder {
         let mut out = vec![];
-        let enc = image::codecs::jpeg::JpegEncoder::new_with_quality(&mut out, q);
+        let enc = image::codecs::jpeg::JpegEncoder::new_with_quality(&mut out, *q);
         if img.write_with_encoder(enc).is_err() {
             continue;
         }
@@ -1157,27 +1168,47 @@ pub fn act(name: &str, v: &Value, scale: (f64, f64)) -> Result<String, String> {
 /// After a step (`after`), it also says when nothing changed, so the bot doesn't click the same
 /// empty place again and again.
 fn look(after: Option<String>, pointer: bool) -> Result<(String, Option<String>), String> {
-    let picture = if mac::screen_capture_preflight() { Some(capture_screen().map_err(|e| format!("看不到畫面：{e}"))?) } else { None };
+    // The two slow reads of every look (C5): the picture and the control tree. They ask two
-    // The app in front: on Linux the window system's own answer when it gives one (agent-desktop
+    // different doors of the system and take about the same time, so they are asked at once —
-    // alone goes by AT-SPI's "active" window, which some desktops never mark: no numbers, 「」).
+    // on a Mac that is a third of a second back on every step.
-    // agent-desktop knows an app by its program's name (a Snap Firefox's window says
+    let front_now = {
-    // "firefox_firefox", its program is "firefox").
         #[cfg(target_os = "linux")]
-    let front = mac::front_app().map(|(n, pid)| crate::desk::program_of(pid).unwrap_or(n)).unwrap_or_default();
+        {
+            mac::front_app().map(|(n, pid)| crate::desk::program_of(pid).unwrap_or(n)).unwrap_or_default()
+        }
         #[cfg(not(target_os = "linux"))]
-    let front = String::new();
+        {
-    let (found, why) = if controls_allowed() {
+            String::new()
+        }
+    };
+    let want_ax = controls_allowed();
+    let (picture, ax) = {
+        let (cap, els) = std::thread::scope(|s| {
+            let cap = s.spawn(|| if mac::screen_capture_preflight() { capture_screen().map(Some) } else { Ok(None) });
+            let fe = front_now.clone();
+            let els = s.spawn(move || if want_ax { crate::desktop::elements(&fe).ok() } else { None });
+            (cap.join(), els.join())
+        });
+        let picture = cap.unwrap_or_else(|_| Err("看不到畫面：截圖執行緒掉了".to_string())).map_err(|e| format!("看不到畫面：{e}"))?;
+        (picture, els.unwrap_or(None))
+    };
+    let front = front_now;
+    let (found, why) = match ax {
+        Some(f) => (Some(f), String::new()),
+        None => {
             let read = || crate::desktop::elements(&front).or_else(|e| if front.is_empty() { Err(e) } else { crate::desktop::elements("") });
             // Right after a click some apps have no focused window for a moment (macOS Calculator
             // after 「乘」): the list came back with only the menu bar. One more look shortly after.
-        let first = read();
+            if controls_allowed() {
-        let again = first.is_err() || first.as_ref().is_ok_and(|f| f.0.is_empty());
+                std::thread::sleep(std::time::Duration::from_millis(300));
-        match if again { std::thread::sleep(std::time::Duration::from_millis(300)); read().or(first) } else { first } {
+                match read() {
                     Ok(f) => (Some(f), String::new()),
                     Err(e) => (None, e),
                 }
             } else {
                 (None, String::new())
+            }
+        }
     };
     if picture.is_none() && found.is_none() {
         ask_permission("Privacy_ScreenCapture");
@@ -1918,6 +1949,16 @@ mod mac {
 
 #[cfg(test)]
 mod tests {
+    #[test]
+    fn the_remote_lane_carries_a_smaller_picture() {
+        // C5: the same screen, tighter ladder, less to carry over the relay.
+        let img = image::RgbImage::from_fn(800, 600, |x, y| image::Rgb([(x % 251) as u8, (y % 241) as u8, ((x + y) % 199) as u8]));
+        let local = encode_jpeg(&img).unwrap().len();
+        REMOTE_LANE.with(|c| c.set(true));
+        let remote = encode_jpeg(&img).unwrap().len();
+        REMOTE_LANE.with(|c| c.set(false));
+        assert!(remote * 12 < local * 13, "remote {remote} should be clearly under local {local}");
+    }
     use super::*;
 
     #[test]

src/remote.rs

@@ -740,6 +740,8 @@ fn device_run(v: &Value, s: &Settings, via_relay: bool, from: &str) -> Resp {
     if !s.hands_for.is_empty() && from != s.hands_for {
         return Resp::error(403, &format!("這台電腦只讓大腦「{}」使用", s.hands_for_name));
     }
+    // Pictures this computer sends now cross a network: WebP, smaller by a quarter (C5).
+    crate::computer::REMOTE_LANE.with(|c| c.set(true));
     let tool = v["tool"].as_str().unwrap_or("");
     let args = v["args"].as_str().unwrap_or("{}");
     let who = v["who"].as_str().unwrap_or("遠端 AnyBot");