@elizaos/sweagent-root 2.0.0-alpha

package/python/tools/web_browser/bin/close_site

@@ -0,0 +1,28 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import _print_response_with_metadata, send_request
+
+config = ClientConfig()
+
+
+def close():
+    """Close the currently open window."""
+    response = send_request(config.port, "close", "POST")
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    args = parser.parse_args()
+    close()

package/python/tools/web_browser/bin/double_click_mouse

@@ -0,0 +1,37 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def double_click(x, y):
+    """Double-click at the specified coordinates (x, y)."""
+    response = send_request(
+        config.port, "double_click", "POST", {"x": x, "y": y, "return_screenshot": config.autoscreenshot}
+    )
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("x", type=int)
+    parser.add_argument("y", type=int)
+    args = parser.parse_args()
+    double_click(args.x, args.y)

package/python/tools/web_browser/bin/drag_mouse

@@ -0,0 +1,46 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_error,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def drag(path):
+    """Drag mouse along a path. Path should be a JSON list of x, y lists: '[[0, 0], [100, 100]]'."""
+    import json
+
+    try:
+        path_data = json.loads(path)
+    except json.JSONDecodeError:
+        _print_error("Path must be valid JSON")
+        return
+    response = send_request(
+        config.port, "drag", "POST", {"path": path_data, "return_screenshot": config.autoscreenshot}
+    )
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument(
+        "path", type=str, help="The path to drag the mouse along (JSON list of x, y lists) e.g. '[[0, 0], [100, 100]]'"
+    )
+    args = parser.parse_args()
+    drag(args.path)

package/python/tools/web_browser/bin/execute_script_on_page

@@ -0,0 +1,39 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def execute_script(script):
+    """Execute a custom JavaScript code snippet on the current page."""
+    response = send_request(
+        config.port,
+        "execute_script",
+        "POST",
+        {"script": script, "return_screenshot": config.autoscreenshot},
+    )
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("script", type=str, help="The JavaScript code snippet to execute")
+    args = parser.parse_args()
+    execute_script(args.script)

package/python/tools/web_browser/bin/get_console_output

@@ -0,0 +1,48 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def get_console_output():
+    """Get console output from the browser."""
+    response = send_request(config.port, "console", "GET", {})
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    if "console_messages" in response:
+        console_messages = response["console_messages"]
+        if console_messages:
+            print("\n--- Console Messages ---")
+            for ix, msg in enumerate(console_messages):
+                # timestamp = msg.get("timestamp", 0)
+                msg_type = msg.get("type", "log")
+                text = msg.get("text", "")
+                location = msg.get("location", {})
+                print(f"[{ix + 1}] {msg_type.upper()}: {text}")
+                if location:
+                    url = location.get("url", "")
+                    line_number = location.get("lineNumber", "")
+                    column_number = location.get("columnNumber", "")
+                    if url:
+                        print(f"    Location: {url}:{line_number}:{column_number}")
+            print("Console buffer cleared.")
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    args = parser.parse_args()
+    get_console_output()

package/python/tools/web_browser/bin/move_mouse

@@ -0,0 +1,35 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def move(x, y):
+    """Move mouse to the specified coordinates (x, y)."""
+    response = send_request(config.port, "move", "POST", {"x": x, "y": y, "return_screenshot": config.autoscreenshot})
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("x", type=int)
+    parser.add_argument("y", type=int)
+    args = parser.parse_args()
+    move(args.x, args.y)

package/python/tools/web_browser/bin/navigate_back

@@ -0,0 +1,33 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def back():
+    """Navigate back in the browser history."""
+    response = send_request(config.port, "back", "POST", {"return_screenshot": config.autoscreenshot})
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    args = parser.parse_args()
+    back()

package/python/tools/web_browser/bin/navigate_forward

@@ -0,0 +1,33 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def forward():
+    """Navigate forward in the browser history."""
+    response = send_request(config.port, "forward", "POST", {"return_screenshot": config.autoscreenshot})
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    args = parser.parse_args()
+    forward()

package/python/tools/web_browser/bin/open_site

@@ -0,0 +1,36 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def open(url):
+    """Open the specified website URL."""
+    if Path(url).is_file():
+        url = f"file://{Path(url).resolve()}"
+    response = send_request(config.port, "goto", "POST", {"url": url, "return_screenshot": config.autoscreenshot})
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument("url", type=str, help="The URL of the website to open")
+    args = parser.parse_args()
+    open(args.url)

package/python/tools/web_browser/bin/press_keys_on_page

@@ -0,0 +1,51 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_error,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def keypress(keys):
+    """Press the specified keys. Keys should be a JSON string like '["ctrl", "c"]'."""
+    import json
+
+    try:
+        keys_data = json.loads(keys)
+    except json.JSONDecodeError:
+        _print_error("Keys must be valid JSON")
+        return
+    response = send_request(
+        config.port,
+        "keypress",
+        "POST",
+        {"keys": keys_data, "return_screenshot": config.autoscreenshot},
+    )
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    parser.add_argument(
+        "keys",
+        type=str,
+        help='The keys to press (JSON string like \'["ctrl", "c"]\')',
+    )
+    args = parser.parse_args()
+    keypress(args.keys)

package/python/tools/web_browser/bin/reload_page

@@ -0,0 +1,33 @@
+#!/root/python3.11/bin/python3
+from __future__ import annotations
+
+import sys
+from argparse import ArgumentParser
+from pathlib import Path
+
+lib_path = str(Path(__file__).resolve().parent.parent / "lib")
+sys.path.insert(0, lib_path)
+
+from web_browser_config import ClientConfig
+from web_browser_utils import (
+    _autosave_screenshot_from_response,
+    _print_response_with_metadata,
+    send_request,
+)
+
+config = ClientConfig()
+
+
+def reload():
+    """Reload the current webpage."""
+    response = send_request(config.port, "reload", "POST", {"return_screenshot": config.autoscreenshot})
+    if response is None:
+        return
+    _print_response_with_metadata(response)
+    _autosave_screenshot_from_response(response, config.screenshot_mode)
+
+
+if __name__ == "__main__":
+    parser = ArgumentParser()
+    args = parser.parse_args()
+    reload()