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

Merge branch 'zdy'

Browse Source
This commit is contained in:
David Chang
2024-03-06 15:22:20 +08:00
parent ef7e8942af f72b788cdc
commit 2fad66f187
11 changed files with 447 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
desktop_env/evaluators/getters/misc.py
View File

@@ -1,5 +1,5 @@
import logging
from typing import TypeVar
from typing import TypeVar, Dict
from datetime import datetime, timedelta
logger = logging.getLogger("desktopenv.getters.misc")
@@ -74,13 +74,13 @@ relativeTime_to_IntDay = {
"first monday four months later": "special"
}
def get_rule(env, config: R) -> R:
def get_rule(env, config: Dict[str, R]) -> R:
"""
Returns the rule as-is.
"""
return config["rules"]
def get_rule_relativeTime(env, config: R) -> R:
def get_rule_relativeTime(env, config: Dict[str, R]) -> R:
"""
According to the rule definded in funciton "apply_rules_to_timeFormat", convert the relative time to absolute time.
config:

58
desktop_env/evaluators/metrics/general.py
View File

@@ -1,6 +1,7 @@
import csv
import functools
import json
import yaml
import operator
import re
import sqlite3
@@ -132,11 +133,11 @@ _accessibility_ns_map = {"st": "uri:deskat:state.at-spi.gnome.org"
}
def check_accessibility_tree(result: str, rules: Dict[str, Any]) -> float:
def check_accessibility_tree(result: str, rules: List[Dict[str, Any]]) -> float:
"""
Args:
result (str): XML of GNOME Accessibility Tree
rules (Dict[str, Any]): dict like
rules (List[Dict[str, Any]]): list of dict like
{
"selectors": list of str as CSS selectors, will be connected by ", "
to form a composite selector. Only one from `selectors` and
@@ -154,30 +155,33 @@ def check_accessibility_tree(result: str, rules: Dict[str, Any]) -> float:
"""
at: _Element = lxml.etree.fromstring(result)
if "xpath" in rules:
elements: List[_Element] = at.xpath(rules["xpath"], namespaces=_accessibility_ns_map)
elif "selectors" in rules:
selector = CSSSelector(", ".join(rules["selectors"]), namespaces=_accessibility_ns_map)
elements: List[_Element] = selector(at)
else:
raise ValueError("At least one of xpath and selectors is required")
total_match_score = 1.
for r in rules:
if "xpath" in r:
elements: List[_Element] = at.xpath(r["xpath"], namespaces=_accessibility_ns_map)
elif "selectors" in r:
selector = CSSSelector(", ".join(r["selectors"]), namespaces=_accessibility_ns_map)
elements: List[_Element] = selector(at)
else:
raise ValueError("At least one of xpath and selectors is required")
if len(elements) == 0:
print("no elements")
return 0.
if len(elements) == 0:
print("no elements")
return 0.
if "text" in rules:
match_func: Callable[[str], Number] = functools.partial(operator.eq if rules["exact"] \
else (lambda a, b: fuzz.ratio(a, b) / 100.)
, rules["text"]
)
match_score: Number = 0
for elm in elements:
match_score = max(match_score, match_func(elm.text or None))
else:
match_score = 1.
if "text" in r:
match_func: Callable[[str], Number] = functools.partial( operator.eq if r["exact"] \
else (lambda a, b: fuzz.ratio(a, b) / 100.)
, r["text"]
)
match_score: Number = 0
for elm in elements:
match_score = max(match_score, match_func(elm.text or None))
else:
match_score = 1.
total_match_score *= match_score
return float(match_score)
return float(total_match_score)
# def check_existence(result: str, *args) -> float:
@@ -189,7 +193,7 @@ def run_sqlite3(result: str, rules: Dict[str, Any]) -> float:
return float(cursor.fetchone()[0] or 0)
def check_json(result: str, rules: Dict[str, List[Dict[str, Union[List[str], str]]]]) -> float:
def check_json(result: str, rules: Dict[str, List[Dict[str, Union[List[str], str]]]], is_yaml: bool = False) -> float:
"""
Args:
result (str): path to json file
@@ -204,6 +208,7 @@ def check_json(result: str, rules: Dict[str, List[Dict[str, Union[List[str], str
],
"unexpect": <the same as `expect`
}
is_yaml (bool): yaml rather than json
Returns:
float
@@ -212,7 +217,10 @@ def check_json(result: str, rules: Dict[str, List[Dict[str, Union[List[str], str
if result is None:
return 0.
with open(result) as f:
result: Dict[str, Any] = json.load(f)
if is_yaml:
result: Dict[str, Any] = yaml.load(f, Loader=yaml.Loader)
else:
result: Dict[str, Any] = json.load(f)
expect_rules = rules.get("expect", {})
unexpect_rules = rules.get("unexpect", {})

13
desktop_env/evaluators/metrics/table.py
View File

@@ -12,6 +12,7 @@ import pandas as pd
from openpyxl import Workbook
from openpyxl.cell.cell import Cell
from openpyxl.worksheet.cell_range import MultiCellRange
from openpyxl.utils import get_column_letter
from openpyxl.worksheet.datavalidation import DataValidation
from openpyxl.worksheet.worksheet import Worksheet
@@ -208,8 +209,10 @@ def compare_table(result: str, expected: str = None, **options) -> float:
for rl in r["rules"]:
for rng in MultiCellRange(rl["range"]):
for cdn in rng.cells:
value1: str = str(read_cell_value(*sheet1, cdn))
value2: str = str(read_cell_value(*sheet2, cdn))
coordinate: str = "{:}{:d}".format(get_column_letter(cdn[1]), cdn[0])
value1: str = str(read_cell_value(*sheet1, coordinate))
value2: str = str(read_cell_value(*sheet2, coordinate))
logger.debug("%s: %s vs %s", cdn, value1, value2)
for rplc in rl.get("normalization", []):
value1 = value1.replace(rplc[0], rplc[1])
@@ -230,11 +233,11 @@ def compare_table(result: str, expected: str = None, **options) -> float:
if rl["type"]=="includes":
metric: bool = value1 in value2
if rl["type"]=="includes_by":
elif rl["type"]=="includes_by":
metric: bool = value2 in value1
if rl["type"]=="fuzzy_match":
elif rl["type"]=="fuzzy_match":
metric: bool = fuzz.ratio(value1, value2) >= rl.get("threshold", 85.)
if rl["type"]=="exact_match":
elif rl["type"]=="exact_match":
metric: bool = value1==value2
total_metric = total_metric and metric

147
evaluation_examples/examples/multi_apps/415ef462-bed3-493a-ac36-ca8c6d23bf1b.json Normal file
View File

@@ -0,0 +1,147 @@
{
"id": "415ef462-bed3-493a-ac36-ca8c6d23bf1b",
"snapshot": "thunderbird",
"instruction": "Save the AWS invoice of December from the email. I have moved that email to local \"Bills\" folder. Save it to the my receipts folder. Keep the file name pattern and update a record to my tally book.",
"source": "authors",
"config": [
{
"type": "execute",
"parameters": {
"command": ["mkdir", "-p", "/home/user/Documents/Finance/receipts", "/home/user/Documents/Projects"]
}
},
{
"type": "download",
"parameters": {
"files": [
{"path": "/home/user/Documents/Finance/receipts/aws-invoice-2308.pdf", "url": "https://drive.google.com/uc?id=1azRFXf4A7fvW0S7r9upHvleMEi-92hHM&export=download"},
{"path": "/home/user/Documents/Finance/receipts/aws-invoice-2309.pdf", "url": "https://drive.google.com/uc?id=1x-lpHm8U4U7uRPZ74-9wq9KzW2R55ln1&export=download"},
{"path": "/home/user/Documents/Finance/receipts/aws-invoice-2310.pdf", "url": "https://drive.google.com/uc?id=1pcrgV9G6NO4ekMEQBiupwXtq6mmke7b_&export=download"},
{"path": "/home/user/Documents/Finance/receipts/aws-invoice-2311.pdf", "url": "https://drive.google.com/uc?id=1JzbCK_nIY8X_3QZjnkzTtb-cRoq9zNT-&export=download"},
{"path": "/home/user/Documents/Finance/receipts/X-receipt-2312.pdf", "url": "https://drive.google.com/uc?id=1QzWjNzvNosG_yQr7VVonvYb3cUYF5f3u&export=download"},
{"path": "/home/user/Documents/Finance/tally_book.xlsx", "url": "https://drive.google.com/uc?id=13yuLhBPmouoWR-DybfgaIbWUOxbY_jhL&export=download"},
{"path": "/home/user/.projects.tar.xz", "url": "https://drive.google.com/uc?id=1oJcxpjqF474Wm16i1aZc8DlCEfAvc4t_&export=download"},
{
"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", "-xJvf", "/home/user/.projects.tar.xz", "-C", "/home/user/Documents/Projects"]
}
},
{
"type": "execute",
"parameters": {
"command": [
"tar",
"-xzv",
"--recursive-unlink",
"-f",
"/home/user/thunderbird-profile.tar.gz",
"-C",
"/home/user/"
]
}
},
{
"type": "launch",
"parameters": {
"command": ["thunderbird"]
}
}
],
"trajectory": "trajectories/415ef462-bed3-493a-ac36-ca8c6d23bf1b",
"related_apps": ["thunderbird", "libreoffice_calc", "os"],
"evaluator": {
"postconfig": [
{
"type": "activate_window",
"parameters": {
"window_name": "tally_book.xlsx - LibreOffice Calc",
"strict": true
}
},
{
"type": "sleep",
"parameters": {
"seconds": 0.5
}
},
{
"type": "execute",
"parameters": {
"command": [
"python",
"-c",
"import pyautogui; pyautogui.hotkey(\"ctrl\", \"s\");"
]
}
},
{
"type": "sleep",
"parameters": {
"seconds": 0.5
}
},
{
"type": "download",
"parameters": {
"files": [
{"path": "/home/user/.aws-invoice-2312.pdf", "url": "https://drive.google.com/uc?id=1RqbulzKG_HeYb1GZmLABOzlohlFg02UU&export=download"}
]
}
},
{
"type": "execute",
"parameters": {
"command": ["diff", ".aws-invoice-2312.pdf", "/home/user/Documents/Finance/receipts/aws-invoice-2312.pdf"],
"stdout": "diff.out"
}
}
],
"func": ["compare_table", "check_list"],
"result": [
{
"type": "vm_file",
"path": "/home/user/Documents/Finance/tally_book.xlsx",
"dest": "tally_book.xlsx"
},
{
"type": "cache_file",
"path": "diff.out"
}
],
"expected": [
{
"type": "cloud_file",
"path": "https://drive.google.com/uc?id=1x8m-korGI1PhJm8PAQVTlWYKneK4WKvn&export=download",
"dest": "tally_book_gt.xlsx"
},
{
"type": "rule",
"rules": {
"unexpect": [
".+"
]
}
}
],
"options": [
{
"rules": [
{
"type": "sheet_data",
"sheet_idx0": 0,
"sheet_idx1": "EI0"
}
]
},
{}
]
}
}

2
evaluation_examples/examples/multi_apps/b5062e3e-641c-4e3a-907b-ac864d2e7652.json
View File

@@ -1,7 +1,7 @@
{
"id": "b5062e3e-641c-4e3a-907b-ac864d2e7652",
"snapshot": "libreoffice_calc",
"instruction": "I've got a mass of different categories of paper in PDF under folder \"~/Documents/Papers\". Please help me to extract the name, mail, and affiliation of the first author of all the papers and note them in an Excel table. The corresponding headers should be added in the table. Simply sort the authors by their full names ascendingly. The summary file should be saved as \"~/authors.xlsx\".",
"instruction": "I've got a mass of papers in PDF under folder \"~/Documents/Papers\". Please help me to extract the name, mail, and affiliation of the first author of all the papers and note them in an Excel table. The corresponding headers should be added in the table. Simply sort the authors by their full names ascendingly. The summary file should be saved as \"~/authors.xlsx\".",
"source": "authors",
"config": [
{

59
evaluation_examples/examples/multi_apps/e1fc0df3-c8b9-4ee7-864c-d0b590d3aa56.json Normal file
View File

@@ -0,0 +1,59 @@
{
"id": "e1fc0df3-c8b9-4ee7-864c-d0b590d3aa56",
"snapshot": "libreoffice_writer",
"instruction": "Install LanguageTool extension for my LibreOffice",
"source": "authors",
"config": [
{
"type": "launch",
"parameters": {
"command": ["libreoffice", "--writer"]
}
}
],
"trajectory": "trajectories/e1fc0df3-c8b9-4ee7-864c-d0b590d3aa56",
"related_apps": ["chrome", "libreoffice", "os"],
"evaluator": {
"postconfig": [
{
"type": "command",
"parameters": {
"command": ["grep", "-nHr", "languagetool", "/home/user/.config/libreoffice/4/user/uno_packages/cache/uno_packages/"],
"stdout": "grep.out"
}
},
{
"type": "command",
"parameters": {
"command": ["apt", "list", "--installed"],
"stdout": "apt.out"
}
}
],
"func": ["check_list", "check_list"],
"result": [
{
"type": "cache_file",
"path": "grep.out"
},
{
"type": "cache_file",
"path": "apt.out"
}
],
"expected": [
{
"type": "rule",
"rules": {
"expect": ["org\\.openoffice\\.languagetool\\.oxt"]
}
},
{
"type": "rule",
"rules": {
"expect": ["openjdk-\\d+-(jre|jdk)"]
}
}
]
}
}

86
evaluation_examples/examples/multi_apps/e2392362-125e-4f76-a2ee-524b183a3412.json Normal file
View File

@@ -0,0 +1,86 @@
{
"id": "e2392362-125e-4f76-a2ee-524b183a3412",
"snapshot": "chrome",
"instruction": "I recently started using the famous personal academic homepage template from academicpages.github.io to build my own personal homepage, and I have cloned it to my local ~/Code/Website folder. According to an online tutorial, I can configure my name and contact information in the _config.yaml file. However, I am not familiar with the YAML file format. Please help me find the sections related to the name and contact information in this file and change them to “Test Account” and “Test@gmail.com”.",
"source": "authors",
"config": [
{
"type": "command",
"parameters": {
"command": ["mkdir", "-p", "/home/user/Code/Website"]
}
},
{
"type": "download",
"parameters": {
"files": [
{
"path": "/home/user/.tmp.tar.xz",
"url": "https://drive.google.com/uc?id=1ordb5kRSPDKgRi7nYQchn8hGt-INELML&export=download"
}
]
}
},
{
"type": "execute",
"parameters": {
"command": ["tar", "-xJvf", ".tmp.tar.xz", "-C", "/home/user/Code/Website/"]
}
},
{
"type": "launch",
"parameters": {
"command": ["google-chrome", "--remote-debugging-port=1337"]
}
},
{
"type": "launch",
"parameters": {
"command": [
"socat",
"tcp-listen:9222,fork",
"tcp:localhost:1337"
]
}
},
{
"type": "chrome_open_tabs",
"parameters": {
"urls_to_open": ["https://academicpages.github.io/"]
}
}
],
"trajectory": "trajectories/e2392362-125e-4f76-a2ee-524b183a3412",
"related_apps": ["chrome", "os", "vscode"],
"evaluator": {
"func": "check_json",
"options": {"is_yaml": true},
"expected": {
"type": "rule",
"rules": {
"expect": [
{
"key": ["name"],
"method": "eq",
"ref": "Test Account"
},
{
"key": ["author", "name"],
"method": "eq",
"ref": "Test Account"
},
{
"key": ["author", "email"],
"method": "eq",
"ref": "Test@gmail.com"
}
]
}
},
"result": {
"type": "vm_file",
"path": "/home/user/Code/Website/academicpages.github.io/_config.yml",
"dest": "_config.yaml"
}
}
}

71
evaluation_examples/examples/multi_apps/f5c13cdd-205c-4719-a562-348ae5cd1d91.json Normal file
View File

@@ -0,0 +1,71 @@
{
"id": "f5c13cdd-205c-4719-a562-348ae5cd1d91",
"snapshot": "thunderbird",
"instruction": "Here is a table recording tuition payments in ~/Documents/Departments/finance. I have already composed an e-mail to remind people who haven't finished payment yet. Please help me to add their emails to the receiver field.",
"source": "authors",
"config": [
{
"type": "execute",
"parameters": {
"command": ["mkdir", "-p", "/home/user/Documents/Departments/finance"]
}
},
{
"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"
},
{
"url": "https://drive.google.com/uc?id=1k0_69RKAx18bEX00EJXtWmSwdfNVP9NA&export=download",
"path": "/home/user/.payment-reminder-mail-body.html"
},
{
"url": "https://drive.google.com/uc?id=1nNc0NoOuP3Of0eGsKY-1kctg63vIjXl5&export=download",
"path": "/home/user/Documents/Departments/finance/tuition.xlsx"
}
]
}
},
{
"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 -compose \"from='Anonym Tester <anonym-x2024@outlook.com>',subject='Reminder of Payment',body='$(cat /home/user/.payment-reminder-mail-body.html)'\"",
"shell": true
}
}
],
"trajectory": "trajectories/f5c13cdd-205c-4719-a562-348ae5cd1d91",
"related_apps": ["thunderbird", "os", "libreoffice_calc"],
"evaluator": {
"func": "check_accessibility_tree",
"result": {
"type": "accessibility_tree"
},
"expected": {
"type": "rule",
"rules": [
{"selectors": ["tool-bar[attr|id=MsgHeadersToolbar] label[attr|class=\"pill-label\"][name*=\"fox@someuniversity.edu\"]"]},
{"selectors": ["tool-bar[attr|id=MsgHeadersToolbar] label[attr|class=\"pill-label\"][name*=\"iron@someuniversity.edu\"]"]},
{"selectors": ["tool-bar[attr|id=MsgHeadersToolbar] label[attr|class=\"pill-label\"][name*=\"nancy@someuniversity.edu\"]"]},
{"selectors": ["tool-bar[attr|id=MsgHeadersToolbar] label[attr|class=\"pill-label\"][name*=\"stella@someuniversity.edu\"]"]}
]
}
}
}

31
evaluation_examples/examples/multi_apps/f8369178-fafe-40c2-adc4-b9b08a125456.json Normal file
View File

@@ -0,0 +1,31 @@
{
"id": "f8369178-fafe-40c2-adc4-b9b08a125456",
"snapshot": "chrome",
"instruction": "Help me to install Orchis theme from gnome-look.org and change to it for my GNOME desktop.",
"source": "https://itsfoss.com/install-switch-themes-gnome-shell",
"config": [],
"trajectory": "trajectories/f8369178-fafe-40c2-adc4-b9b08a125456",
"related_apps": ["chrome", "os"],
"evaluator": {
"postconfig": [
{
"type": "execute",
"parameters": {
"command": ["gsettings", "get", "org.gnome.desktop.interface", "gtk-theme"],
"stdout": "gsettings.out"
}
}
],
"func": "check_list",
"result": {
"type": "cache_file",
"path": "gsettings.out"
},
"expected": {
"type": "rule",
"rules": {
"expect": ["Orchis"]
}
}
}
}

12
evaluation_examples/examples/thunderbird/12086550-11c0-466b-b367-1d9e75b3910e.json
View File

@@ -48,11 +48,13 @@
},
"expected": {
"type": "rule",
"rules": {
"selectors": [
"application[name=Thunderbird] page-tab-list[attr|id=\"tabmail-tabs\"]>page-tab[name=\"About Profiles\"]"
]
}
"rules": [
{
"selectors": [
"application[name=Thunderbird] page-tab-list[attr|id=\"tabmail-tabs\"]>page-tab[name=\"About Profiles\"]"
]
}
]
},
"func": "check_accessibility_tree"
}

1
requirements.txt
View File

@@ -42,3 +42,4 @@ func-timeout
beautifulsoup4
dashscope
google-generativeai
PyYaml