autogpt / benchmark /reports /send_to_googledrive.py
kakumusic's picture
Upload folder using huggingface_hub
b225a21 verified
raw
history blame
6.47 kB
import base64
import json
import os
import re
from datetime import datetime, timedelta
import gspread
import pandas as pd
from dotenv import load_dotenv
from oauth2client.service_account import ServiceAccountCredentials
# Load environment variables from .env file
load_dotenv()
# Get the base64 string from the environment variable
base64_creds = os.getenv("GDRIVE_BASE64")
if base64_creds is None:
raise ValueError("The GDRIVE_BASE64 environment variable is not set")
# Decode the base64 string into bytes
creds_bytes = base64.b64decode(base64_creds)
# Convert the bytes into a string
creds_string = creds_bytes.decode("utf-8")
# Parse the string into a JSON object
creds_info = json.loads(creds_string)
# Define the base directory containing JSON files
base_dir = "reports"
# Get the current working directory
current_dir = os.getcwd()
# Check if the current directory ends with 'reports'
if current_dir.endswith("reports"):
base_dir = "/"
else:
base_dir = "reports"
# Create a list to store each row of data
rows = []
def process_test(
test_name: str, test_info: dict, agent_name: str, common_data: dict
) -> None:
"""Recursive function to process test data."""
parts = test_name.split("_", 1) # Split by underscore only once
test_suite = parts[0] if len(parts) > 1 else None
# transform array into string with | as separator
separator = "|"
categories = separator.join(
test_info.get("category", []),
)
row = {
"Agent": agent_name,
"Command": common_data.get("command", ""),
"Completion Time": common_data.get("completion_time", ""),
"Benchmark Start Time": common_data.get("benchmark_start_time", ""),
"Total Run Time": common_data.get("metrics", {}).get("run_time", ""),
"Highest Difficulty": common_data.get("metrics", {}).get(
"highest_difficulty", ""
),
"Workspace": common_data.get("config", {}).get("workspace", ""),
"Test Name": test_name,
"Data Path": test_info.get("data_path", ""),
"Is Regression": test_info.get("is_regression", ""),
"Difficulty": test_info.get("metrics", {}).get("difficulty", ""),
"Success": test_info.get("metrics", {}).get("success", ""),
"Success %": test_info.get("metrics", {}).get("success_%", ""),
"Non mock success %": test_info.get("metrics", {}).get(
"non_mock_success_%", ""
),
"Run Time": test_info.get("metrics", {}).get("run_time", ""),
"Benchmark Git Commit Sha": common_data.get("benchmark_git_commit_sha", None),
"Agent Git Commit Sha": common_data.get("agent_git_commit_sha", None),
"Cost": test_info.get("metrics", {}).get("cost", ""),
"Attempted": test_info.get("metrics", {}).get("attempted", ""),
"Test Suite": test_suite,
"Category": categories,
"Task": test_info.get("task", ""),
"Answer": test_info.get("answer", ""),
"Description": test_info.get("description", ""),
"Fail Reason": test_info.get("metrics", {}).get("fail_reason", ""),
"Reached Cutoff": test_info.get("reached_cutoff", ""),
}
rows.append(row)
# Check for nested tests and process them if present
nested_tests = test_info.get("tests")
if nested_tests:
for nested_test_name, nested_test_info in nested_tests.items():
process_test(nested_test_name, nested_test_info, agent_name, common_data)
# Usage:
# Loop over each directory in the base directory
for agent_dir in os.listdir(base_dir):
agent_dir_path = os.path.join(base_dir, agent_dir)
# Ensure the agent_dir_path is a directory
if os.path.isdir(agent_dir_path):
# Loop over each sub-directory in the agent directory (e.g., "folder49_07-28-03-53")
for report_folder in os.listdir(agent_dir_path):
report_folder_path = os.path.join(agent_dir_path, report_folder)
# Ensure the report_folder_path is a directory
if os.path.isdir(report_folder_path):
# Check for a file named "report.json" in the sub-directory
report_path = os.path.join(report_folder_path, "report.json")
if os.path.exists(report_path):
# Load the JSON data from the file
with open(report_path, "r") as f:
data = json.load(f)
benchmark_start_time = data.get("benchmark_start_time", "")
# Check if benchmark_start_time complies with the required format
pattern = re.compile(r"\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\+00:00")
if not pattern.fullmatch(benchmark_start_time):
continue # Skip processing this report if the date is not in the correct format
# Parse the benchmark_start_time to a datetime object
benchmark_datetime = datetime.strptime(
benchmark_start_time, "%Y-%m-%dT%H:%M:%S+00:00"
)
# Check if benchmark_start_time is older than 3 days
current_datetime = datetime.utcnow()
if current_datetime - benchmark_datetime > timedelta(days=3):
continue # Skip processing this report if it's more than 3 days old
# Loop through each test
for test_name, test_info in data["tests"].items():
process_test(test_name, test_info, agent_dir, data)
# Convert the list of rows into a DataFrame
df = pd.DataFrame(rows)
# Define the scope
scope = [
"https://spreadsheets.google.com/feeds",
"https://www.googleapis.com/auth/drive",
]
# Add your service account credentials
creds = ServiceAccountCredentials.from_json_keyfile_dict(creds_info, scope)
# Authorize the clientsheet
client = gspread.authorize(creds)
# Get the instance of the Spreadsheet
branch_name = os.getenv("GITHUB_REF_NAME")
sheet = client.open(f"benchmark-{branch_name}")
# Get the first sheet of the Spreadsheet
sheet_instance = sheet.get_worksheet(0)
# Convert dataframe to list of lists for uploading to Google Sheets
values = df.values.tolist()
# Prepend the header to the values list
values.insert(0, df.columns.tolist())
# Clear the existing values in the worksheet
sheet_instance.clear()
# Update the worksheet with the new values
sheet_instance.append_rows(values)