agentenv-framework-protocol 0.1.490__tar.gz → 0.1.491__tar.gz

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: agentenv-framework-protocol
-Version: 0.1.490
+Version: 0.1.491
 Summary: Open data-plane protocol and server SDK for agent environments.
 Project-URL: Homepage, https://www.agentenvframework.com
 Project-URL: Documentation, https://www.agentenvframework.com/docs
@@ -757,15 +757,21 @@ driver booted with a control token is probed with it: `--token`, by default
 | Level | Checks | Needs on the computer |
 |---|---|---|
 | `basic` | P0 ready and a screenshot; P1 the card and tool schemas; P2 native frames, regions and JPEG; P5 refusals and their codes; X the declared extensions (`computer-display`, `computer-spaces`, `computer-keyboard`, `computer-exec`, `data/get`, per-role tool access, the pointer) | nothing |
-| `full` | `basic`, then P3 colours land where the mapping says; P4 clicks show their effect; P5 refusals change nothing; P6 typed and pressed keys arrive; P7 scrolls and drags arrive; P9 commands and `data/add` run as one unprivileged user; X pointer and keyboard input, `data/add`, and a recording's segments matching their files with no gap, a repeated `request_id` joining it and `max_concurrent` held; on a page, X the page tools: a prompt answered, a file uploaded, a reload, navigation, back and forward, and tabs | Linux, `python3` with Tk (on a page, `python3` and `computer-page/v1`), a 1280x800 screen or larger, `computer-exec` and `data/add` |
+| `full` | `basic`, then P3 colours land where the mapping says; P4 clicks show their effect; P5 refusals change nothing; P6 typed and pressed keys arrive; P7 scrolls and drags arrive; P9 commands and `data/add` run as one unprivileged user; X pointer and keyboard input, `data/add`, and a recording's segments matching their files with no gap, a repeated `request_id` joining it and `max_concurrent` held; on a page, X the page tools: a prompt answered, a file uploaded, a reload, navigation, back and forward, and tabs | Linux, `python3` with Tk (on Wayland, with PyGObject and GTK 3 instead; on a page, `python3` and `computer-page/v1`), a 1280x800 screen or larger, `computer-exec` and `data/add` |
 
-At `full` the probe writes the Fixture Desk, a Tk window (`conformance/fixture.py`), into
-`~/.agentenv-probe` with `data/add` and starts it with `computer-exec`. It aims at the boxes the
+At `full` the probe writes the Fixture Desk, a Tk window (`conformance/fixture.py`), and its layout
+(`conformance/desk.py`) into `~/.agentenv-probe` with `data/add` and starts it with `computer-exec`. It aims at the boxes the
 window's widgets report, reads back what reached them, and then removes the fixture. On X11 the
 window is maximized and kept above other windows, raising itself again when one covers it, so a
 window an app keeps on top doesn't take the clicks aimed at it; elsewhere it is full screen. When the
 fixture can't be staged, the `full` checks are unverified rather than failed.
 
+On a driver whose platform's display is Wayland, the desk is drawn with GTK 3 as a native Wayland client
+(`conformance/wayland_fixture.py`, staged beside `desk.py` in the Tk window's place), since Tk runs
+only under XWayland. It goes full screen, takes its origin from Hyprland's IPC, keeps the clipboard
+through wl-clipboard. There the probe selects a text box's text with GTK's own Ctrl+A rather than
+Tk's Ctrl+/.
+
 On a driver whose card says `screen: "page"`, the desk is an HTML page with the same widgets at the
 same places (`conformance/page_fixture.py`). The probe starts its server on loopback, which writes
 what the page reports to the same files, and opens the desk with `computer_tabs` in a tab of its
@@ -852,7 +858,8 @@ class MailApp(ComputerAppEnvironment):
   it runs on (`platform.detect()`).
 - **`launch=Launch(...)`** says how it starts in the desktop session: with it (`at="boot"`) or when
   the agent opens it (`"demand"`). `ready=Ready(...)` says when it's ready, within `timeout_s` of its
-  launch: a window matching `launch.window` by its class, instance or title, a port answering on
+  launch: a window matching `launch.window` by its class, instance or title (on Hyprland, a Wayland
+  app ID or an X window's class, its program's file name, and its title), a port answering on
   loopback, or a command exiting 0. `env` sets variables in the app's environment, over its server's. `@ready_check` marks a method for readiness that needs logic (a
   sync one runs in a thread), and `@install_step(scope=...)` one a computer runs as it installs the
   app, after its run scripts, offline and without the class's `__init__`. Each runs on every OS the
@@ -886,13 +893,16 @@ desktop session, and `ready` (GET: an `AppReadiness` of `ready`, `problems`, `la
 and `timeout_s`), which runs the app's readiness checks then, at once, each within 10 s. A server
 that can't start, or whose start in the session fails (adding the launcher, say), writes why to
 `.agentenv/server.failure` in the folder and exits, and the driver reports that in its place. Once
-the X display is up, when `DISPLAY` names one, the server adds the app's launcher to the session user's
+the session is up, the X display when `DISPLAY` names one or on Hyprland its Wayland display and its
+IPC, the server adds the app's launcher to the session user's
 menu, which runs `python -m agentenv_protocol.computer.launch <folder>` to ask the server, so the
 server starts every copy of the app, and launches a `boot` app, once per boot. An app that runs with
 no desktop, a website browser alone, gets no launcher. The driver finds each app by the socket in its
 folder: the computer is ready once every app's server answers and each `boot` app, and each app with
 no launch, says it's ready; a `demand` app counts once something launches it. What the server does
-in the session (`agentenv_protocol.computer.session`) is Linux's.
+in the session (`agentenv_protocol.computer.session`) is Linux's. On Hyprland it lists windows through
+Hyprland's IPC (`hyprctl -j clients`), which sees native Wayland windows and XWayland's alike, where
+X's `xprop` sees only XWayland's.
 
 The server serves the app's data plane and its skill. With `paths` alone, `data/add` writes the files
 it's given under them and refuses any other path, or one a link leads out of them; `data/get`

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/README.md

@@ -692,15 +692,21 @@ driver booted with a control token is probed with it: `--token`, by default
 | Level | Checks | Needs on the computer |
 |---|---|---|
 | `basic` | P0 ready and a screenshot; P1 the card and tool schemas; P2 native frames, regions and JPEG; P5 refusals and their codes; X the declared extensions (`computer-display`, `computer-spaces`, `computer-keyboard`, `computer-exec`, `data/get`, per-role tool access, the pointer) | nothing |
-| `full` | `basic`, then P3 colours land where the mapping says; P4 clicks show their effect; P5 refusals change nothing; P6 typed and pressed keys arrive; P7 scrolls and drags arrive; P9 commands and `data/add` run as one unprivileged user; X pointer and keyboard input, `data/add`, and a recording's segments matching their files with no gap, a repeated `request_id` joining it and `max_concurrent` held; on a page, X the page tools: a prompt answered, a file uploaded, a reload, navigation, back and forward, and tabs | Linux, `python3` with Tk (on a page, `python3` and `computer-page/v1`), a 1280x800 screen or larger, `computer-exec` and `data/add` |
+| `full` | `basic`, then P3 colours land where the mapping says; P4 clicks show their effect; P5 refusals change nothing; P6 typed and pressed keys arrive; P7 scrolls and drags arrive; P9 commands and `data/add` run as one unprivileged user; X pointer and keyboard input, `data/add`, and a recording's segments matching their files with no gap, a repeated `request_id` joining it and `max_concurrent` held; on a page, X the page tools: a prompt answered, a file uploaded, a reload, navigation, back and forward, and tabs | Linux, `python3` with Tk (on Wayland, with PyGObject and GTK 3 instead; on a page, `python3` and `computer-page/v1`), a 1280x800 screen or larger, `computer-exec` and `data/add` |
 
-At `full` the probe writes the Fixture Desk, a Tk window (`conformance/fixture.py`), into
-`~/.agentenv-probe` with `data/add` and starts it with `computer-exec`. It aims at the boxes the
+At `full` the probe writes the Fixture Desk, a Tk window (`conformance/fixture.py`), and its layout
+(`conformance/desk.py`) into `~/.agentenv-probe` with `data/add` and starts it with `computer-exec`. It aims at the boxes the
 window's widgets report, reads back what reached them, and then removes the fixture. On X11 the
 window is maximized and kept above other windows, raising itself again when one covers it, so a
 window an app keeps on top doesn't take the clicks aimed at it; elsewhere it is full screen. When the
 fixture can't be staged, the `full` checks are unverified rather than failed.
 
+On a driver whose platform's display is Wayland, the desk is drawn with GTK 3 as a native Wayland client
+(`conformance/wayland_fixture.py`, staged beside `desk.py` in the Tk window's place), since Tk runs
+only under XWayland. It goes full screen, takes its origin from Hyprland's IPC, keeps the clipboard
+through wl-clipboard. There the probe selects a text box's text with GTK's own Ctrl+A rather than
+Tk's Ctrl+/.
+
 On a driver whose card says `screen: "page"`, the desk is an HTML page with the same widgets at the
 same places (`conformance/page_fixture.py`). The probe starts its server on loopback, which writes
 what the page reports to the same files, and opens the desk with `computer_tabs` in a tab of its
@@ -787,7 +793,8 @@ class MailApp(ComputerAppEnvironment):
   it runs on (`platform.detect()`).
 - **`launch=Launch(...)`** says how it starts in the desktop session: with it (`at="boot"`) or when
   the agent opens it (`"demand"`). `ready=Ready(...)` says when it's ready, within `timeout_s` of its
-  launch: a window matching `launch.window` by its class, instance or title, a port answering on
+  launch: a window matching `launch.window` by its class, instance or title (on Hyprland, a Wayland
+  app ID or an X window's class, its program's file name, and its title), a port answering on
   loopback, or a command exiting 0. `env` sets variables in the app's environment, over its server's. `@ready_check` marks a method for readiness that needs logic (a
   sync one runs in a thread), and `@install_step(scope=...)` one a computer runs as it installs the
   app, after its run scripts, offline and without the class's `__init__`. Each runs on every OS the
@@ -821,13 +828,16 @@ desktop session, and `ready` (GET: an `AppReadiness` of `ready`, `problems`, `la
 and `timeout_s`), which runs the app's readiness checks then, at once, each within 10 s. A server
 that can't start, or whose start in the session fails (adding the launcher, say), writes why to
 `.agentenv/server.failure` in the folder and exits, and the driver reports that in its place. Once
-the X display is up, when `DISPLAY` names one, the server adds the app's launcher to the session user's
+the session is up, the X display when `DISPLAY` names one or on Hyprland its Wayland display and its
+IPC, the server adds the app's launcher to the session user's
 menu, which runs `python -m agentenv_protocol.computer.launch <folder>` to ask the server, so the
 server starts every copy of the app, and launches a `boot` app, once per boot. An app that runs with
 no desktop, a website browser alone, gets no launcher. The driver finds each app by the socket in its
 folder: the computer is ready once every app's server answers and each `boot` app, and each app with
 no launch, says it's ready; a `demand` app counts once something launches it. What the server does
-in the session (`agentenv_protocol.computer.session`) is Linux's.
+in the session (`agentenv_protocol.computer.session`) is Linux's. On Hyprland it lists windows through
+Hyprland's IPC (`hyprctl -j clients`), which sees native Wayland windows and XWayland's alike, where
+X's `xprop` sees only XWayland's.
 
 The server serves the app's data plane and its skill. With `paths` alone, `data/add` writes the files
 it's given under them and refuses any other path, or one a link leads out of them; `data/get`

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "agentenv-framework-protocol"
-version = "0.1.490"
+version = "0.1.491"
 description = "Open data-plane protocol and server SDK for agent environments."
 readme = "README.md"
 license = "Apache-2.0"

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/src/agentenv_protocol/client.py

@@ -80,7 +80,7 @@ def granted_headers(url: str, grants: dict[str | None, dict[str, str]]) -> dict[
 async def _rpc(base_url: str, method: str, params: dict, timeout: int, verify: bool,
                headers: dict[str, str] | None) -> dict:
     request = {"jsonrpc": "2.0", "id": 1, "method": method, "params": params}
-    url = f"{base_url}{RPC_PATH}"
+    url = f"{base_url.rstrip('/')}{RPC_PATH}"
     async with httpx.AsyncClient(verify=verify) as client:
         response = await client.post(url, json=request, timeout=timeout, headers=request_headers(url, headers))
         response.raise_for_status()
@@ -117,7 +117,7 @@ async def get_data(
 
 async def get_card(base_url: str, timeout: int = 10, verify: bool = True) -> dict:
     async with httpx.AsyncClient(verify=verify) as client:
-        response = await client.get(f"{base_url}{WELL_KNOWN_PATH}", timeout=timeout)
+        response = await client.get(f"{base_url.rstrip('/')}{WELL_KNOWN_PATH}", timeout=timeout)
         response.raise_for_status()
         return response.json()
 
@@ -341,7 +341,7 @@ async def invoke_extension(base_url: str, card: dict, uri: str, params: dict | N
         raise ValueError(f"extension {uri} does not advertise method {method!r} (advertises: {offered})")
     if not op["endpoint"]:
         raise ValueError(f"extension {uri} has no endpoint in params")
-    url = f"{base_url}{op['endpoint']}"
+    url = f"{base_url.rstrip('/')}{op['endpoint']}"
     sent = request_headers(url, headers)
     async with httpx.AsyncClient(verify=verify) as client:
         if op["method"] == "GET":

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/src/agentenv_protocol/computer/_session_posix.py

@@ -1,10 +1,12 @@
 """What a computer app's own server does in the desktop session on Linux and macOS (``session``): wait for the X
-session, launch the app into a session of its own and stop or pause it by its process group, add its freedesktop menu
-launcher and list the X session's windows."""
+session, or on Hyprland for its Wayland and IPC sockets, launch the app into a session of its own and stop or pause it
+by its process group, add its freedesktop menu launcher and list the session's windows, X's with ``xprop`` or
+Hyprland's with ``hyprctl``."""
 from __future__ import annotations
 
 import asyncio
 import contextlib
+import json
 import os
 import re
 import signal
@@ -21,9 +23,13 @@ WITH_FDS = Path(__file__).with_name("_with_fds.py")
 
 
 async def wait_until_up() -> None:
-    """Until the X display's socket is there."""
-    display = os.environ["DISPLAY"].lstrip(":").split(".")[0]
-    while not Path(f"/tmp/.X11-unix/X{display}").exists():
+    """Until the session's sockets are there: on Hyprland its Wayland display's and its IPC's, else the X display's."""
+    if signature := _hyprland():
+        runtime = Path(os.environ["XDG_RUNTIME_DIR"])
+        sockets = [runtime / os.environ["WAYLAND_DISPLAY"], runtime / "hypr" / signature / ".socket.sock"]
+    else:
+        sockets = [Path(f"/tmp/.X11-unix/X{os.environ['DISPLAY'].lstrip(':').split('.')[0]}")]
+    while not all(socket.exists() for socket in sockets):
         await asyncio.sleep(POLL_S)
 
 
@@ -101,7 +107,11 @@ def add_launcher(name: str, folder: Path) -> Path:
 async def windows() -> list[dict[str, str]]:
     """The window manager's top-level windows, each as its WM_CLASS ``instance`` and ``class`` and its ``title``, read
     with ``xprop``: ``_NET_WM_NAME``, or ``WM_NAME`` for a window that sets only that; a window that closes while it's
-    read is left out."""
+    read is left out. On Hyprland, its mapped clients as its IPC lists them, native Wayland and XWayland alike: its
+    ``class``, a Wayland app ID or an X window's WM_CLASS class, and its ``title``, and as its ``instance``, which a
+    Wayland window hasn't, its program's file name, as on Windows."""
+    if _hyprland():
+        return await _clients()
     listed = await _xprop("-root", "_NET_CLIENT_LIST")
     found = []
     for window in re.findall(r"0x[0-9a-f]+", listed.partition("#")[2]):
@@ -119,8 +129,37 @@ def _strings(value: str) -> list[str]:
     return [re.sub(r"\\(.)", r"\1", text) for text in re.findall(r'"((?:[^"\\]|\\.)*)"', value)]
 
 
+def _hyprland() -> str | None:
+    """The Hyprland instance whose session this is, by its signature; None outside one."""
+    return os.environ.get("HYPRLAND_INSTANCE_SIGNATURE") or None
+
+
+async def _clients() -> list[dict[str, str]]:
+    """Hyprland's mapped clients, by their program, class and title; none when its IPC doesn't answer."""
+    try:
+        clients = json.loads(await _output("hyprctl", "-j", "clients") or "[]")
+    except ValueError:
+        return []
+    return [{"instance": _program(client.get("pid")), "class": client.get("class") or "",
+             "title": client.get("title") or ""}
+            for client in clients if isinstance(client, dict) and client.get("mapped")]
+
+
+def _program(pid: object) -> str:
+    """The file name of the program process ``pid`` runs, or "" when it can't be read."""
+    try:
+        return Path(os.readlink(f"/proc/{int(pid)}/exe")).name
+    except (OSError, TypeError, ValueError):
+        return ""
+
+
 async def _xprop(*args: str) -> str:
-    process = await asyncio.create_subprocess_exec("xprop", *args, stdout=asyncio.subprocess.PIPE,
+    return await _output("xprop", *args)
+
+
+async def _output(*argv: str) -> str:
+    """What ``argv`` prints, or nothing when it fails; killed when the listing it's part of is cut short."""
+    process = await asyncio.create_subprocess_exec(*argv, stdout=asyncio.subprocess.PIPE,
                                                    stderr=asyncio.subprocess.DEVNULL)
     try:
         out, _ = await process.communicate()

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/src/agentenv_protocol/computer/app.py

@@ -187,7 +187,8 @@ class Launch(BaseModel):
     at: Literal["boot", "demand"] = "boot"
     command: list[str]
     #: Matches the app's window by ``class``, ``instance`` and ``title``, as given: on X11 its WM_CLASS class and
-    #: instance; on Windows its window class and its program's file name, lowercased, such as ``notepad.exe``.
+    #: instance; on Hyprland its class, a Wayland app ID or an X window's WM_CLASS class, and its program's file name;
+    #: on Windows its window class and its program's file name, lowercased, such as ``notepad.exe``.
     window: Optional[dict[str, str]] = None
     ready: Optional[Ready] = None
     #: Variables set in the app's environment, over its server's.

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/src/agentenv_protocol/computer/app_runtime.py

@@ -80,14 +80,14 @@ class AppRuntime:
         self.checking: dict[str, asyncio.Future] = {}
 
     async def start(self) -> None:
-        """Once its X display is up, when it has one, add the app's menu launcher on a desktop and launch a boot app,
-        unless the computer has launched it since it booted, at boot or on demand, or its boot launch is held
-        (``boot_launch_held``): a server that restarts leaves the app be, or with ``relaunch_on_start``, for an app
-        only the server that launched it can talk to, stops it and launches it again, whether it was launched at boot
-        or on demand."""
+        """Once its session is up, its X display or Hyprland, when it has one, add the app's menu launcher on a desktop
+        and launch a boot app, unless the computer has launched it since it booted, at boot or on demand, or its boot
+        launch is held (``boot_launch_held``): a server that restarts leaves the app be, or with ``relaunch_on_start``,
+        for an app only the server that launched it can talk to, stops it and launches it again, whether it was
+        launched at boot or on demand."""
         if self.launch is None:
             return
-        if os.environ.get("DISPLAY"):
+        if os.environ.get("DISPLAY") or os.environ.get("HYPRLAND_INSTANCE_SIGNATURE"):
             await session.wait_until_up()
         if self.desktop:
             session.add_launcher(os.environ["ENVIRONMENT_NAME"], Path.cwd())

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/src/agentenv_protocol/computer/conformance/checks.py

@@ -116,8 +116,15 @@ class Probe:
         self.layout: dict = {}
         self.fixture_pid: str | None = None
         self.page = self.features.get("screen") == "page"
-        self.fixture = "page_fixture.py" if self.page else "fixture.py"
-        self.windows = (self.features.get("platform") or {}).get("os") == "windows"
+        platform = self.features.get("platform") or {}
+        #: Whether the desk is the GTK one: on Wayland, where Tk runs only under XWayland, which a Wayland driver may
+        #: not reach.
+        self.gtk = not self.page and platform.get("display") == "wayland"
+        self.fixture = "page_fixture.py" if self.page else "wayland_fixture.py" if self.gtk else "fixture.py"
+        self.windows = platform.get("os") == "windows"
+        #: The chord that selects a text box's text: Tk's Ctrl+/, which the Tk desk binds everywhere, or on the GTK
+        #: desk GTK's own Ctrl+A.
+        self.select_all = ["Control", "a"] if self.gtk else ["Control", "/"]
         self.desk = ""
         self.tab: str | None = None
         self.opened: list[str] = []
@@ -562,12 +569,13 @@ class Probe:
     # full: the fixture
 
     async def _stage(self) -> str | None:
-        """The Fixture Desk started, or what stopped it: the Tk desk, or on a page, the HTML desk's server, with the
-        desk opened in a tab of the probe's own."""
-        source = resources.files(__package__).joinpath(self.fixture).read_bytes()
-        added = await self.driver.add_file(f"{STAGE}/{self.fixture}", source)
-        if "error" in added:
-            return f"data/add refused it: {added['error'].get('message')}"
+        """The Fixture Desk started, or what stopped it: the Tk desk or on Wayland the GTK desk, either beside the layout
+        they share (desk.py), or on a page, the HTML desk's server, with the desk opened in a tab of the probe's own."""
+        for file in [self.fixture] if self.page else [self.fixture, "desk.py"]:
+            source = resources.files(__package__).joinpath(file).read_bytes()
+            added = await self.driver.add_file(f"{STAGE}/{file}", source)
+            if "error" in added:
+                return f"data/add refused it: {added['error'].get('message')}"
         if self.windows:
             await self.driver.run(*WIN_REMOVE, f"{STAGE}/state")
             started = await self.driver.run(*WIN_START, f"{STAGE}/{self.fixture}", f"{STAGE}/state", f"{STAGE}/fixture.log")
@@ -781,7 +789,7 @@ class Probe:
 
         async def clear() -> None:
             await focus("name")
-            await self.act("computer_key", {"keys": ["Control", "/"]})
+            await self.act("computer_key", {"keys": self.select_all})
             await self.act("computer_key", {"keys": ["Backspace"]})
 
         async def unicode() -> str | None:
@@ -796,7 +804,7 @@ class Probe:
 
         async def chord() -> str | None:
             await focus("name")
-            await self.act("computer_key", {"keys": ["Control", "/"]})
+            await self.act("computer_key", {"keys": self.select_all})
             await self.act("computer_key", {"keys": ["Backspace"]})
             return await self.settled(lambda s: s["name"] == "", lambda s: s["name"])
 

agentenv_framework_protocol-0.1.491/src/agentenv_protocol/computer/conformance/desk.py

@@ -0,0 +1,65 @@
+"""The Fixture Desk's layout and what it records, shared by the Tk desk (fixture.py) and the GTK desk
+(wayland_fixture.py), so the two can't drift apart. The probe stages it beside whichever desk it starts. It imports no
+toolkit, since a Wayland computer may have no Tk."""
+from __future__ import annotations
+
+import json
+import os
+from pathlib import Path
+
+#: The colours a grid cell cycles through, one more on each click.
+COLOURS = ("#ffffff", "#ff0000", "#00aa00", "#0000ff")
+CELL = 50
+GRID = 8
+#: Single-pixel markers on a black field, at odd and even pixels: (x, y) within the field, then the colour.
+MARKERS = ((11, 11, "#ff0000"), (12, 11, "#00ff00"), (11, 12, "#0000ff"), (12, 12, "#ffff00"), (37, 20, "#ff00ff"),
+           (20, 37, "#00ffff"))
+CLIPBOARD = "agentenv-probe clipboard"
+MIN_SIZE = (1280, 800)
+#: Where each widget's top left goes, the name box's label at "label" and the list's frame at "list".
+AT = {"label": (40, 40), "name": (140, 40), "save": (560, 34), "subscribe": (40, 100), "notes": (760, 40),
+      "list": (40, 160), "grid": (400, 160), "pad": (840, 160), "wheel": (840, 500), "field": (1080, 500)}
+#: The widgets sized in pixels: the list's frame, the drawing pad, the wheel and the marker field.
+SIZE = {"list": {"width": 300, "height": 400}, "pad": {"width": 400, "height": 300},
+        "wheel": {"width": 200, "height": 120}, "field": {"width": 48, "height": 48}}
+#: The list's rows, Item 1 to Item ITEMS.
+ITEMS = 100
+#: How many of the latest keys and drags state.json keeps, and about how many points of a drag's path.
+KEYS_KEPT, DRAGS_KEPT, PATH_POINTS = 100, 20, 50
+#: How often the desk reads inbox.txt.
+INBOX_EVERY_MS = 200
+
+
+def new_state(width: int, height: int, windowing: str) -> dict:
+    """What state.json holds before anything reaches the desk."""
+    return {"name": "", "notes": "", "keys": [], "saved_count": 0, "subscribed": False, "selected": None,
+            "scroll_top": 0.0, "cells": {}, "right_clicks": 0, "double_clicks": 0, "drags": [],
+            "wheel": {"up": 0, "down": 0, "left": 0, "right": 0}, "clipboard": None, "loaded": None,
+            "screen": [width, height], "windowing": windowing}
+
+
+def write_state(state_dir: Path, state: dict) -> None:
+    """state.json replaced whole, so the probe never reads half of one."""
+    part = state_dir / ".state.json.part"
+    part.write_text(json.dumps(state))
+    os.replace(part, state_dir / "state.json")
+
+
+def write_layout(state_dir: Path, boxes: dict, row_height: int, field: tuple[int, int]) -> None:
+    """layout.json: each widget's box on the screen, a list row's height, the markers' pixels from the field's top left,
+    and the cell size and clipboard text the probe checks against."""
+    layout = {**boxes, "list_row_height": row_height,
+              "markers": [[field[0] + x, field[1] + y, colour] for x, y, colour in MARKERS], "cell": CELL,
+              "clipboard": CLIPBOARD}
+    (state_dir / "layout.json").write_text(json.dumps(layout))
+
+
+def record_key(state: dict, key: str) -> None:
+    state["keys"] = (state["keys"] + [key])[-KEYS_KEPT:]
+
+
+def record_drag(state: dict, stroke: list) -> None:
+    """A finished drag: its ends, its number of points and its path thinned to about PATH_POINTS points."""
+    path = stroke[::max(1, len(stroke) // PATH_POINTS)] + stroke[-1:]
+    drag = {"start": stroke[0], "end": stroke[-1], "points": len(stroke), "path": path}
+    state["drags"] = (state["drags"] + [drag])[-DRAGS_KEPT:]

{agentenv_framework_protocol-0.1.490 → agentenv_framework_protocol-0.1.491}/src/agentenv_protocol/computer/conformance/fixture.py

@@ -3,26 +3,19 @@ Desk, whose widgets' screen boxes are written to <state-dir>/layout.json and eve
 whatever lands in <state-dir>/inbox.txt, to <state-dir>/state.json, so the probe can aim at it and read back what
 arrived. It records the same effects under X11, macOS's Aqua and Windows, whose Tk names keys, binds select-all and
 reports the wheel differently. On X11 it stays above an app's windows (``keep_on_top``); elsewhere it's full screen.
-Run as ``python3 fixture.py <state-dir>``; it needs a 1280x800 screen or larger."""
+Its layout and what it records are desk.py's, which the GTK desk shares. Run as ``python3 fixture.py <state-dir>``
+beside desk.py; it needs a 1280x800 screen or larger."""
 from __future__ import annotations
 
 import ctypes
-import json
-import os
 import sys
 import time
 import tkinter as tk
 from pathlib import Path
 
-#: The colours a grid cell cycles through, one more on each click.
-COLOURS = ("#ffffff", "#ff0000", "#00aa00", "#0000ff")
-CELL = 50
-GRID = 8
-#: Single-pixel markers on a black field, at odd and even pixels: (x, y) within the field, then the colour.
-MARKERS = ((11, 11, "#ff0000"), (12, 11, "#00ff00"), (11, 12, "#0000ff"), (12, 12, "#ffff00"), (37, 20, "#ff00ff"),
-           (20, 37, "#00ffff"))
-CLIPBOARD = "agentenv-probe clipboard"
-MIN_SIZE = (1280, 800)
+from desk import (AT, CELL, CLIPBOARD, COLOURS, GRID, INBOX_EVERY_MS, ITEMS, MARKERS, MIN_SIZE, SIZE, new_state,
+                  record_drag, record_key, write_layout, write_state)
+
 #: X11's keysyms for the keys Aqua's Tk names by their character.
 KEYSYMS = {",": "comma", " ": "space"}
 #: A wheel notch's <MouseWheel> delta where Tk scales the wheel: Windows', and Tk 9's everywhere.
@@ -46,15 +39,10 @@ def main(state_dir: Path) -> None:
     else:
         root.attributes("-fullscreen", True)
     root.configure(bg="#f0f0f0")
-    state: dict = {"name": "", "notes": "", "keys": [], "saved_count": 0, "subscribed": False, "selected": None,
-                   "scroll_top": 0.0, "cells": {}, "right_clicks": 0, "double_clicks": 0, "drags": [],
-                   "wheel": {"up": 0, "down": 0, "left": 0, "right": 0}, "clipboard": None, "loaded": None,
-                   "screen": [width, height], "windowing": root.tk.call("tk", "windowingsystem")}
+    state = new_state(width, height, root.tk.call("tk", "windowingsystem"))
 
     def save() -> None:
-        part = state_dir / ".state.json.part"
-        part.write_text(json.dumps(state))
-        os.replace(part, state_dir / "state.json")
+        write_state(state_dir, state)
 
     if width < MIN_SIZE[0] or height < MIN_SIZE[1]:
         state["error"] = f"the screen is {width}x{height}; the fixture needs {MIN_SIZE[0]}x{MIN_SIZE[1]}"
@@ -63,13 +51,13 @@ def main(state_dir: Path) -> None:
         return
 
     font = ("DejaVu Sans", 16)
-    tk.Label(root, text="Name:", font=font, bg="#f0f0f0").place(x=40, y=40)
+    _place(tk.Label(root, text="Name:", font=font, bg="#f0f0f0"), "label")
     name = tk.Entry(root, font=font, width=30)
-    name.place(x=140, y=40)
+    _place(name, "name")
 
     def key(event) -> None:
         state["name"] = name.get()
-        state["keys"] = (state["keys"] + [KEYSYMS.get(event.keysym, event.keysym)])[-100:]
+        record_key(state, KEYSYMS.get(event.keysym, event.keysym))
         state["clipboard"] = _clipboard(root)
         save()
 
@@ -88,7 +76,7 @@ def main(state_dir: Path) -> None:
         save()
 
     button = tk.Button(root, text="Save", font=font, command=on_save, width=8)
-    button.place(x=560, y=34)
+    _place(button, "save")
 
     subscribed = tk.BooleanVar(value=False)
 
@@ -98,10 +86,10 @@ def main(state_dir: Path) -> None:
 
     subscribe = tk.Checkbutton(root, text="Subscribe", font=font, variable=subscribed, command=on_subscribe,
                                bg="#f0f0f0")
-    subscribe.place(x=40, y=100)
+    _place(subscribe, "subscribe")
 
     notes = tk.Text(root, font=("DejaVu Sans", 14), width=28, height=3)
-    notes.place(x=760, y=40)
+    _place(notes, "notes")
 
     def note(event) -> None:
         state["notes"] = notes.get("1.0", "end-1c")
@@ -110,7 +98,7 @@ def main(state_dir: Path) -> None:
     notes.bind("<KeyRelease>", note)
 
     frame = tk.Frame(root)
-    frame.place(x=40, y=160, width=300, height=400)
+    _place(frame, "list", **SIZE["list"])
     listbox = tk.Listbox(frame, font=("DejaVu Sans", 14), exportselection=False)
     listbox.pack(side="left", fill="both", expand=True)
     bar = tk.Scrollbar(frame, command=listbox.yview)
@@ -122,7 +110,7 @@ def main(state_dir: Path) -> None:
         save()
 
     listbox.configure(yscrollcommand=scrolled)
-    for i in range(1, 101):
+    for i in range(1, ITEMS + 1):
         listbox.insert("end", f"Item {i}")
 
     def select(event) -> None:
@@ -133,7 +121,7 @@ def main(state_dir: Path) -> None:
     listbox.bind("<<ListboxSelect>>", select)
 
     grid = tk.Canvas(root, width=GRID * CELL, height=GRID * CELL, bg="#ffffff", highlightthickness=0)
-    grid.place(x=400, y=160)
+    _place(grid, "grid")
     for row in range(GRID):
         for col in range(GRID):
             grid.create_rectangle(col * CELL, row * CELL, (col + 1) * CELL, (row + 1) * CELL, fill=COLOURS[0],
@@ -155,8 +143,8 @@ def main(state_dir: Path) -> None:
     grid.bind("<Button-3>", lambda e: cell(e, True))
     grid.bind("<Double-Button-1>", double)
 
-    pad = tk.Canvas(root, width=400, height=300, bg="#fffbe6", highlightthickness=0)
-    pad.place(x=840, y=160)
+    pad = tk.Canvas(root, **SIZE["pad"], bg="#fffbe6", highlightthickness=0)
+    _place(pad, "pad")
     stroke: list = []
 
     def press(event) -> None:
@@ -170,8 +158,7 @@ def main(state_dir: Path) -> None:
     def release(event) -> None:
         if stroke:
             stroke.append((event.x, event.y))
-            state["drags"] = (state["drags"] + [{"start": stroke[0], "end": stroke[-1], "points": len(stroke),
-                                                 "path": stroke[::max(1, len(stroke) // 50)] + stroke[-1:]}])[-20:]
+            record_drag(state, stroke)
             save()
             stroke.clear()
 
@@ -179,8 +166,8 @@ def main(state_dir: Path) -> None:
     pad.bind("<B1-Motion>", motion)
     pad.bind("<ButtonRelease-1>", release)
 
-    wheel = tk.Canvas(root, width=200, height=120, bg="#e6f0ff", highlightthickness=0)
-    wheel.place(x=840, y=500)
+    wheel = tk.Canvas(root, **SIZE["wheel"], bg="#e6f0ff", highlightthickness=0)
+    _place(wheel, "wheel")
 
     def wheeled(horizontal: bool, steps: int) -> None:
         way = ("left" if horizontal else "up") if steps > 0 else ("right" if horizontal else "down")
@@ -202,8 +189,8 @@ def main(state_dir: Path) -> None:
     if tk.TkVersion >= 8.7:
         wheel.bind("<TouchpadScroll>", lambda e: scrolled(e.delta))
 
-    field = tk.Canvas(root, width=48, height=48, bg="#000000", highlightthickness=0, bd=0)
-    field.place(x=1080, y=500)
+    field = tk.Canvas(root, **SIZE["field"], bg="#000000", highlightthickness=0, bd=0)
+    _place(field, "field")
     for x, y, colour in MARKERS:
         field.create_rectangle(x, y, x + 1, y + 1, fill=colour, outline="", width=0)
 
@@ -212,28 +199,29 @@ def main(state_dir: Path) -> None:
         if message.exists() and (text := message.read_text()) != state["loaded"]:
             state["loaded"] = text
             save()
-        root.after(200, poll_inbox)
+        root.after(INBOX_EVERY_MS, poll_inbox)
 
-    def write_layout() -> None:
+    def layout() -> None:
         root.update_idletasks()
         boxes = {key: [w.winfo_rootx(), w.winfo_rooty(), w.winfo_width(), w.winfo_height()]
                  for key, w in (("name", name), ("save", button), ("subscribe", subscribe), ("notes", notes),
                                 ("list", listbox), ("grid", grid), ("pad", pad), ("wheel", wheel), ("field", field))}
-        boxes["list_row_height"] = listbox.bbox(0)[3] if listbox.bbox(0) else 0
-        boxes["markers"] = [[field.winfo_rootx() + x, field.winfo_rooty() + y, colour] for x, y, colour in MARKERS]
-        boxes["cell"] = CELL
-        boxes["clipboard"] = CLIPBOARD
         save()
-        (state_dir / "layout.json").write_text(json.dumps(boxes))
+        write_layout(state_dir, boxes, listbox.bbox(0)[3] if listbox.bbox(0) else 0,
+                     (field.winfo_rootx(), field.winfo_rooty()))
 
     root.clipboard_clear()
     root.clipboard_append(CLIPBOARD)
     state["clipboard"] = CLIPBOARD
-    root.after(300, write_layout)
+    root.after(300, layout)
     root.after(500, poll_inbox)
     root.mainloop()
 
 
+def _place(widget: tk.Widget, at: str, **size: int) -> None:
+    widget.place(x=AT[at][0], y=AT[at][1], **size)
+
+
 def keep_on_top(root: tk.Tk) -> None:
     """Keep the fixture over the windows an app opens, which would take the probe's clicks: maximized and above other
     windows rather than full screen, since xfwm4 keeps a window an app puts above over a full-screen one, and raised