lzy/sci-gui-agent-benchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Resolve conflicts

Browse Source
This commit is contained in:
Timothyxxx
2024-01-13 22:58:20 +08:00
parent a1c3e4c294 f274193265
commit 57a41a279c
18 changed files with 502 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
desktop_env/evaluators/getters/__init__.py
View File

@@ -3,3 +3,5 @@ from .info import get_vm_screen_size, get_vm_window_size, get_vm_wallpaper
from .misc import get_rule, get_accessibility_tree
from .vlc import get_vlc_playing_info, get_vlc_config
from .chrome import get_default_search_engine, get_cookie_data, get_bookmarks, get_open_tabs_info, get_pdf_from_url, get_shortcuts_on_desktop
from .replay import get_replay
from .vscode import get_vscode_config

19
desktop_env/evaluators/getters/replay.py
View File

@@ -0,0 +1,19 @@
from typing import List, Dict, Any
def get_replay(env, trajectory: List[Dict[str, Any]]) -> None:
def parse(action):
if action["type"] == "hotkey":
keys = "', '".join(action["param"])
return f"pyautogui.hotkey('{keys}')"
if action["type"] == "typewrite":
text = action["param"]
return f"pyautogui.typewrite('{text}')"
if action["type"] == "press":
key = action["param"]
return f"pyautogui.press('{key}')"
for action in trajectory:
env.controller.execute_python_command(parse(action))

13
desktop_env/evaluators/getters/vscode.py Normal file
View File

@@ -0,0 +1,13 @@
from typing import Dict
from typing import Any
from replay import get_replay
from file import get_vm_file
def get_vscode_config(env, config: Dict[str, Any]) -> str:
get_replay(env, config["trajectory"])
return get_vm_file(env, {
"path": config["path"],
"dest": config["dest"]
})

5
desktop_env/evaluators/metrics/__init__.py
View File

@@ -10,8 +10,7 @@ from .table import check_sheet_list, check_xlsx_freeze, check_xlsx_zoom
from .table import compare_table
from .vlc import is_vlc_playing, is_vlc_recordings_folder, is_vlc_fullscreen, compare_images, compare_audios, \
compare_videos
from .gimp import increase_saturation, decrease_brightness
from .general import check_csv, check_accessibility_tree, check_list
from .gimp import increase_saturation, decrease_brightness, check_file_exists, compare_triangle_positions
from .general import check_csv, check_accessibility_tree, check_list, run_sqlite3
from .thunderbird import check_thunderbird_prefs, check_thunderbird_filter

7
desktop_env/evaluators/metrics/general.py
View File

@@ -13,6 +13,8 @@ from rapidfuzz import fuzz
from .utils import _match_record
import sqlite3
def exact_match(result, rules) -> float:
expect = rules["expected"]
print(result, expect)
@@ -146,3 +148,8 @@ def check_accessibility_tree(result: str, rules: Dict[str, Any]) -> float:
# def check_existence(result: str, *args) -> float:
# return 1. - (result is None)
def run_sqlite3(result: str, rules: Dict[str, Any]) -> float:
connection: sqlite3.Connection = sqlite3.connect(result)
cursor: sqlite3.Cursor = connection.execute(rules["sql"])
return float(cursor.fetchone()[0] or 0)

47
desktop_env/evaluators/metrics/gimp.py
View File

@@ -20,6 +20,10 @@ def get_gimp_export_path():
# Handle the case where the configuration file is not found
print("GIMP configuration file not found")
return False
def check_file_exists(directory, filename):
file_path = os.path.join(directory, filename)
return 1 if os.path.isfile(file_path) else 0
def increase_saturation(image1_path: str, image2_path: str) -> float:
def calculate_saturation(image):
@@ -61,6 +65,49 @@ def decrease_brightness(image1_path: str, image2_path: str) -> float:
brightness2 = calculate_brightness(image2)
return 1 if brightness1 > brightness2 else 0
import cv2
import numpy as np
def find_yellow_triangle(image):
# Convert the image to RGBA
rgba = cv2.cvtColor(image, cv2.COLOR_BGR2RGBA)
# define range of yellow color in HSV
lower_yellow = np.array([0, 0, 0], dtype=np.uint8)
upper_yellow = np.array([255, 255, 255], dtype=np.uint8)
# expand the dimensions of lower and upper yellow to match the image dimensions
lower_yellow = np.reshape(lower_yellow, (1, 1, 3))
upper_yellow = np.reshape(upper_yellow, (1, 1, 3))
# build a mask for the yellow color
mask = cv2.inRange(rgba, lower_yellow, upper_yellow)
# search for contours in the mask
contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# choose the largest contour
max_contour = max(contours, key=cv2.contourArea)
# calculate the center of the contour
M = cv2.moments(max_contour)
cx = int(M['m10'] / M['m00'])
cy = int(M['m01'] / M['m00'])
return cx, cy
def compare_triangle_positions(image1, image2):
image1 = cv2.imread(image1, cv2.IMREAD_COLOR)
image2 = cv2.imread(image2, cv2.IMREAD_COLOR)
# find the center of the yellow triangle in each image
cx1, cy1 = find_yellow_triangle(image1)
cx2, cy2 = find_yellow_triangle(image2)
# calculate the distance between the center of the triangle and the center of the image
center_distance1 = np.sqrt((cx1 - image1.shape[1] // 2)**2 + (cy1 - image1.shape[0] // 2)**2)
center_distance2 = np.sqrt((cx2 - image2.shape[1] // 2)**2 + (cy2 - image2.shape[0] // 2)**2)
return 1 if center_distance1 > center_distance2 else 0
if __name__ == "__main__":
image1_path = "../Downloads/1.png"

2
desktop_env/evaluators/metrics/thunderbird.py
View File

@@ -85,7 +85,7 @@ def check_thunderbird_prefs(result: str, rule: Dict[str, Dict[str, Dict[str, Any
_value_processor: Callable[[str], str] = lambda val: val.replace("\\\"", "\"").replace("\\\\", "\\")
#_condition_pattern: Pattern[str] = re.compile(r'(?P<type>AND|OR) \((?P<key>[\w ]+),(?P<rel>[\w ' + '\'' + r']+),(?:"(?P<val2>(?:[^"]|\")+)"|(?P<val1>[^)]+))\)')
_condition_pattern: Pattern[str] = re.compile(r'(?:AND|OR) \((?:[\w ]+),(?:[\w ' + '\'' + r']+),(?:"(?:(?:[^"]|\")+)"|(?:[^)]+))\)')
_condition_pattern: Pattern[str] = re.compile(r'\b(?:AND|OR) \((?:[\w ]+),(?:[\w ' + '\'' + r']+),(?:"(?:(?:[^"]|\")+)"|(?:[^)]+))\)|\bALL\b')
def check_thunderbird_filter(result: str, rules: Dict[str, List[Dict[str, str]]]) -> float:
"""
Args:

32
desktop_env/evaluators/metrics/vscode.py
View File

@@ -1,11 +1,8 @@
def compare_text_file(actual: str, expected: str, **options) -> float:
"""
Args:
actual (str): path to result xlsx
expected (str): path to gold xlsx
options (Dict[str, List[str]]): dict like
{
}
actual (str): path to result text file
expected (str): path to gold text file
Return:
float: the score
@@ -20,7 +17,32 @@ def compare_text_file(actual: str, expected: str, **options) -> float:
return 1.0
return 0.0
def compare_config(actual: str, expected: str, **options) -> float:
"""
Args:
actual (str): path to result text file
expected (str): gold string
Return:
float: the score
"""
with open(actual) as f1:
actual_text = f1.read()
if actual_text == expected:
return 1.0
return 0.0
def compare_answer(actual: str, expected: str, **options) -> float:
"""
Args:
actual (str): result string
expected (str): gold string
Return:
float: the score
"""
if actual == expected:
return 1.0

16
evaluation_examples/examples/gimp/554785e9-4523-4e7a-b8e1-8016f565f56a.json
View File

@@ -9,8 +9,8 @@
"parameters": {
"files": [
{
"url": "",
"path": "Desktop/2.png"
"url": "https://drive.usercontent.google.com/download?id=1-3cc3qqjlJ_3d9IpwsuvSuYpyJ2LxrIA&export=download&authuser=0&confirm=t&uuid=6b08e902-ed25-47d1-97d9-7794f9508288&at=APZUnTW74TY2hHtBbWDa0N7-cE58:1705060487948",
"path": "Desktop/woman_sitting_by_the_tree2.png"
}
]
}
@@ -18,7 +18,7 @@
{
"type": "open",
"parameters": {
"path": "Desktop/2.png"
"path": "Desktop/woman_sitting_by_the_tree2.png"
}
}
],
@@ -29,14 +29,14 @@
"evaluator": {
"func": "increase_saturation",
"expected": {
"type": "cloud_file",
"path": "",
"dest": "edited_colorful.png"
"type": "vm_file",
"path": "Desktop/woman_sitting_by_the_tree2.png",
"dest": "woman_sitting_by_the_tree2.png"
},
"result": {
"type": "vm_file",
"path": "Desktop/2.png",
"dest": "2.png"
"path": "Desktop/edited_colorful.png",
"dest": "edited_colorful.png"
}
}
}

34
evaluation_examples/examples/gimp/77b8ab4d-994f-43ac-8930-8ca087d7c4b4 Normal file
View File

@@ -0,0 +1,34 @@
{
"id": "77b8ab4d-994f-43ac-8930-8ca087d7c4b4",
"snapshot": "gimp",
"instruction": "Could you help me keep my photo on the desktop and rename it export.jpg?",
"source": "https://superuser.com/questions/1636113/how-to-get-gimp-to-recognize-images-or-pictures-folder-as-the-default-folder-for",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=13ECFsPxznuoCANto21ijj9OzP0APukIH&export=download&authuser=0&confirm=t&uuid=d8f2dd03-8992-4646-be62-3a3cf89583f2&at=APZUnTVsR0xmbXvpFIpXLzCcLrMa:1705062951627",
"path": "Desktop/The_Lost_River_Of_Dreams.jpg"
}
]
}
},
{
"type": "open",
"parameters": {
"path": "Desktop/The_Lost_River_Of_Dreams.jpg"
}
}
],
"trajectory": "trajectories/",
"related_apps": [
"gimp"
],
"evaluator": {
"func": "check_file_exists",
"file_name": "export.png",
"directory": "/home/user/Desktop/"
}
}

14
evaluation_examples/examples/gimp/7a4deb26-d57d-4ea9-9a73-630f66a7b568.json
View File

@@ -10,7 +10,7 @@
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1SIvX9Wimyw6i2UvnoLTNDHIObvDLAsIM&export=download&authuser=0&confirm=t&uuid=a48447ab-13a2-421f-9662-6ffff8f6f6d5&at=APZUnTVRxofs822XxgEv33WwYCkb:1705046264363",
"path": "Desktop/1.png"
"path": "Desktop/woman_sitting_by_the_tree.png"
}
]
}
@@ -18,7 +18,7 @@
{
"type": "open",
"parameters": {
"path": "Desktop/1.png"
"path": "Desktop/woman_sitting_by_the_tree.png"
}
}
],
@@ -29,14 +29,14 @@
"evaluator": {
"func": "decrease_brightness",
"expected": {
"type": "cloud_file",
"path": "https://drive.usercontent.google.com/download?id=1M7T5iPv0kWvmZBbdTcRdG45zIpHMtrBn&export=download&authuser=0&confirm=t&uuid=d95db1f8-27dc-4532-9976-ea99c293f53e&at=APZUnTXfpgNCtXDtlqd55LgmrIAj:1705048270986",
"dest": "edited_darker.png"
"type": "vm_file",
"path": "Desktop/woman_sitting_by_the_tree.png",
"dest": "woman_sitting_by_the_tree.png"
},
"result": {
"type": "vm_file",
"path": "Desktop/1.png",
"dest": "1.png"
"path": "Desktop/edited_darker.png",
"dest": "edited_darker.png"
}
}
}

42
evaluation_examples/examples/gimp/f4aec372-4fb0-4df5-a52b-79e0e2a5d6ce Normal file
View File

@@ -0,0 +1,42 @@
{
"id": "f4aec372-4fb0-4df5-a52b-79e0e2a5d6ce",
"snapshot": "gimp",
"instruction": "Help me select the yellow triangle and move it to the center of my photo.",
"source": "https://superuser.com/questions/612338/how-do-i-select-and-move-an-object-in-gimp",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1IMFTLy3lLSP8ee5b-940JjaoOd5-oGwn&export=download&authuser=0&confirm=t&uuid=0e06dffa-b602-42c6-8841-cf6eb3524e2b&at=APZUnTVpN4xHqAIAYmHZgk-UXvY4:1705081430356",
"path": "Desktop/Triangle_On_The_Side.png"
}
]
}
},
{
"type": "open",
"parameters": {
"path": "Desktop/Triangle_On_The_Side.png"
}
}
],
"trajectory": "trajectories/",
"related_apps": [
"gimp"
],
"evaluator": {
"func": "compare_triangle_positions",
"expected": {
"type": "vm_file",
"path": "Desktop/Triangle_On_The_Side.png",
"dest": "Triangle_On_The_Side.png"
},
"result": {
"type": "vm_file",
"path": "Desktop/Triangle_In_The_Middle.png",
"dest": "Triangle_In_The_Middle.png"
}
}
}

72
evaluation_examples/examples/thunderbird/2ad9387a-65d8-4e33-ad5b-7580065a27ca.json Normal file
View File

@@ -0,0 +1,72 @@
{
"id": "2ad9387a-65d8-4e33-ad5b-7580065a27ca",
"snapshot": "thunderbird",
"instruction": "Create two local folders for me: COMPANY and UNIVERSITY.",
"source": "https://support.mozilla.org/bm/questions/1027435",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1EHLRWzBCOsyERkSMUnTF2pnsR0n6ZvtR&export=download&authuser=0&confirm=t&uuid=de09bd5e-bef8-499a-b599-c642af190e10&at=APZUnTXqOsQkxl0zMSX6R1Sgp_v3:1704362491712",
"path": "/home/user/thunderbird-profile.tar.gz"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": [
"tar",
"-xzv",
"--recursive-unlink",
"-f",
"/home/user/thunderbird-profile.tar.gz",
"-C",
"/home/user/"
]
}
},
{
"type": "launch",
"parameters": {
"command": [
"/usr/bin/thunderbird"
]
}
}
],
"trajectory": "trajectories/2ad9387a-65d8-4e33-ad5b-7580065a27ca",
"related_apps": [
"thunderbird"
],
"evaluator": {
"postconfig": [
{
"type": "execute",
"parameters": {
"command": ["ls", "-R", "/home/user/.thunderbird/t5q2a5hp.default-release/Mail/Local Folders"],
"stdout": "thunder-local-folder.ls"
}
}
],
"func": "check_list",
"result": {
"type": "cache_file",
"path": "thunder-local-folder.ls"
},
"expected": {
"type": "rule",
"rules": {
"expect": [
"\\bCOMPANY\\.msf\\b",
"\\bCOMPANY/?(?!\\.msf)",
"\\bUNIVERSITY\\.msf\\b",
"\\bUNIVERSITY/?(?!\\.msf)"
]
}
}
}
}

77
evaluation_examples/examples/thunderbird/35253b65-1c19-4304-8aa4-6884b8218fc0.json Normal file
View File

@@ -0,0 +1,77 @@
{
"id": "35253b65-1c19-4304-8aa4-6884b8218fc0",
"snapshot": "thunderbird",
"instruction": "Set up to forward every email received by anonym-x2024@outlook.com in the future to anonym-x2024@gmail.com. Please don't touch the online account. Just locally in the Thunderbird!",
"source": "https://support.mozilla.org/en-US/questions/1259354",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1EHLRWzBCOsyERkSMUnTF2pnsR0n6ZvtR&export=download&authuser=0&confirm=t&uuid=de09bd5e-bef8-499a-b599-c642af190e10&at=APZUnTXqOsQkxl0zMSX6R1Sgp_v3:1704362491712",
"path": "/home/user/thunderbird-profile.tar.gz"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": [
"tar",
"-xzv",
"--recursive-unlink",
"-f",
"/home/user/thunderbird-profile.tar.gz",
"-C",
"/home/user/"
]
}
},
{
"type": "launch",
"parameters": {
"command": [
"/usr/bin/thunderbird"
]
}
}
],
"trajectory": "trajectories/35253b65-1c19-4304-8aa4-6884b8218fc0",
"related_apps": [
"thunderbird"
],
"evaluator": {
"postconfig": [
{
"type": "command",
"parameters": {
"command": ["wmctrl", "-Fc", "Message Filters"],
"until": {
"returncode": 1
}
}
}
],
"func": "check_thunderbird_filter",
"result": {
"type": "vm_file",
"path": "/home/user/.thunderbird/t5q2a5hp.default-release/ImapMail/outlook.office365.com/msgFilterRules.dat",
"dest": "msgFilterRules.dat"
},
"expected": {
"type": "rule",
"rules": {
"expect": [
{
"enabled": "yes",
"action": "Forward",
"actionValue": "anonym-x2024@gmail.com",
"condition": ["ALL"]
}
]
}
}
}
}

74
evaluation_examples/examples/thunderbird/3d1682a7-0fb0-49ae-a4dc-a73afd2d06d5.json Normal file
View File

@@ -0,0 +1,74 @@
{
"id": "3d1682a7-0fb0-49ae-a4dc-a73afd2d06d5",
"snapshot": "thunderbird",
"instruction": "Add a star to every email in local Bills folder",
"source": "https://support.mozilla.org/en-US/kb/organize-your-messages-using-filters",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1EHLRWzBCOsyERkSMUnTF2pnsR0n6ZvtR&export=download&authuser=0&confirm=t&uuid=de09bd5e-bef8-499a-b599-c642af190e10&at=APZUnTXqOsQkxl0zMSX6R1Sgp_v3:1704362491712",
"path": "/home/user/thunderbird-profile.tar.gz"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": [
"tar",
"-xzv",
"--recursive-unlink",
"-f",
"/home/user/thunderbird-profile.tar.gz",
"-C",
"/home/user/"
]
}
},
{
"type": "launch",
"parameters": {
"command": [
"/usr/bin/thunderbird"
]
}
}
],
"trajectory": "trajectories/3d1682a7-0fb0-49ae-a4dc-a73afd2d06d5",
"related_apps": [
"thunderbird"
],
"evaluator": {
"postconfig": [
{
"type": "execute",
"parameters": {
"command": [
"wmctrl",
"-xFc",
"Mail.thunderbird"
],
"until": {
"returncode": 1
}
}
}
],
"result": {
"type": "vm_file",
"path": "/home/user/.thunderbird/t5q2a5hp.default-release/global-messages-db.sqlite",
"dest": "global-messages-db.sqlite"
},
"func": "run_sqlite3",
"expected": {
"type": "rule",
"rules": {
"sql": "select sum(1) > 0 from messageAttributes where attributeID = 58 and value = 1 and messageID in (select id from messages where folderID = 13);"
}
}
}
}

52
evaluation_examples/examples/thunderbird/3d1682a7-0fb0-49ae-a4dc-a73afd2d06d5.json.nosetup Normal file
View File

@@ -0,0 +1,52 @@
{
"id": "3d1682a7-0fb0-49ae-a4dc-a73afd2d06d5",
"snapshot": "thunderbird",
"instruction": "Add a star to every email in local Bills folder",
"source": "https://support.mozilla.org/en-US/kb/organize-your-messages-using-filters",
"config": [
{
"type": "download",
"parameters": {
"files": [
{
"url": "https://drive.usercontent.google.com/download?id=1EHLRWzBCOsyERkSMUnTF2pnsR0n6ZvtR&export=download&authuser=0&confirm=t&uuid=de09bd5e-bef8-499a-b599-c642af190e10&at=APZUnTXqOsQkxl0zMSX6R1Sgp_v3:1704362491712",
"path": "/home/user/thunderbird-profile.tar.gz"
}
]
}
}
],
"trajectory": "trajectories/3d1682a7-0fb0-49ae-a4dc-a73afd2d06d5",
"related_apps": [
"thunderbird"
],
"evaluator": {
"postconfig": [
{
"type": "execute",
"parameters": {
"command": [
"wmctrl",
"-xFc",
"Mail.thunderbird"
],
"until": {
"returncode": 1
}
}
}
],
"result": {
"type": "vm_file",
"path": "/home/user/.thunderbird/t5q2a5hp.default-release/global-messages-db.sqlite",
"dest": "global-messages-db.sqlite"
},
"func": "run_sqlite3",
"expected": {
"type": "rule",
"rules": {
"sql": "select sum(1) > 0 from messageAttributes where attributeID = 58 and value = 1 and messageID in (select id from messages where folderID = 13);"
}
}
}
}

18
evaluation_examples/examples/vs_code/53ad5833-3455-407b-bbc6-45b4c79ab8fb.json
View File

@@ -15,5 +15,21 @@
"related_apps": [
"vscode"
],
"evaluator": "evaluation_dir"
"evaluator": {
"func": "compare_config",
"expected": {
"type": "string",
"string": "project"
},
"result": {
"type": "vscode_config",
"trajectory": [
{"type": "hotkey", "param": ["command", "shift", "p"]},
{"type": "typewrite", "param": "OpenProject"},
{"type": "press", "param": "enter"}
],
"path": "OpenProject.txt",
"dest": "OpenProject.txt"
}
}
}

1
requirements.txt
View File

@@ -28,4 +28,5 @@ pyacoustid
opencv-python
ImageHash
scikit-image
librosa
pymupdf