File size: 29,575 Bytes
cd451ea 471986c cd451ea 599eb7e 471986c 807709a 66dc719 cd451ea 9bb36bd 599eb7e 471986c 807709a 9bb36bd cd451ea 471986c 807709a 471986c cd451ea 471986c 599eb7e 471986c 807709a 471986c 807709a 471986c 807709a 599eb7e 471986c 807709a 599eb7e 471986c 807709a 471986c 599eb7e 471986c 599eb7e 471986c 599eb7e 471986c 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 471986c cd451ea 471986c b1f5e34 471986c b1f5e34 471986c b1f5e34 471986c 807709a 471986c 599eb7e 471986c 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 9bb36bd 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e 9bb36bd 599eb7e 807709a 9bb36bd 807709a 599eb7e 807709a 599eb7e 9bb36bd 599eb7e 746d43a 599eb7e 746d43a 599eb7e 746d43a 599eb7e 746d43a 599eb7e 9bb36bd 599eb7e 9bb36bd 746d43a 471986c 9bb36bd 7bea63a 9bb36bd 7bea63a 9bb36bd 7bea63a 9bb36bd 7bea63a 9bb36bd 471986c 599eb7e 471986c 9bb36bd 746d43a 471986c 599eb7e 7bea63a 471986c cd451ea 471986c cd451ea 746d43a d7c2876 599eb7e cd451ea f3d7989 cd451ea 599eb7e cd451ea 807709a 599eb7e 9bb36bd cd451ea 9bb36bd 599eb7e 807709a 9bb36bd cd451ea 599eb7e 7bea63a 599eb7e 7bea63a 599eb7e 807709a 599eb7e 807709a 599eb7e 807709a 599eb7e cd451ea 807709a 7bea63a 599eb7e cd451ea 599eb7e cd451ea 9bb36bd cd451ea 9bb36bd cd451ea 9bb36bd c51b070 792d89e 9bb36bd cd451ea 9bb36bd 180a869 cd451ea 9bb36bd b6b1f39 180a869 b6b1f39 cd451ea 180a869 cd451ea b6b1f39 cd451ea 180a869 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 |
import os
import pandas as pd
from typing import Tuple, List, Dict
import ipfshttpclient
from utils import INC_TOOLS
from typing import List
from utils import TMP_DIR, ROOT_DIR
from cloud_storage import (
initialize_client,
download_tools_historical_files,
FILES_IN_TWO_MONTHS,
FILES_IN_FOUR_MONTHS,
)
import math
import time
ACCURACY_FILENAME = "tools_accuracy.csv"
IPFS_SERVER = "/dns/registry.autonolas.tech/tcp/443/https"
GCP_IPFS_SERVER = "/dns/registry.gcp.autonolas.tech/tcp/443/https"
SAMPLING_POPULATION_SIZE = 500
RECENTS_SAMPLES_SIZE = 5000
NR_SUBSETS = 100
SAMPLES_THRESHOLD = 50
DEFAULT_ACCURACY = 51.0
LAST_MODEL_UPDATE = "2025-06-03"
CLAUDE_TOOLS = [
"claude-prediction-online",
"claude-prediction-offline",
"prediction-request-rag-claude",
"prediction-request-reasoning-claude",
"prediction-url-cot-claude",
]
def mean_and_std(numbers):
mean = sum(numbers) / len(numbers)
squared_diff_sum = sum((x - mean) ** 2 for x in numbers)
std_dev = math.sqrt(squared_diff_sum / len(numbers))
return mean, std_dev
def clean_tools_dataset(tools_df: pd.DataFrame) -> pd.DataFrame:
# Remove tool_name and TEMP_TOOL
tools_non_error = tools_df[
tools_df["tool"].isin(["tool_name", "TEMP_TOOL"]) == False
].copy()
# Remove errors
tools_non_error = tools_non_error[tools_non_error["error"] == 0]
tools_non_error.loc[:, "currentAnswer"] = tools_non_error["currentAnswer"].replace(
{"no": "No", "yes": "Yes"}
)
tools_non_error = tools_non_error[
tools_non_error["currentAnswer"].isin(["Yes", "No"])
]
tools_non_error = tools_non_error[tools_non_error["vote"].isin(["Yes", "No"])]
tools_non_error["win"] = (
tools_non_error["currentAnswer"] == tools_non_error["vote"]
).astype(int)
tools_non_error.columns = tools_non_error.columns.astype(str)
return tools_non_error
def classify_tools(
tools_df: pd.DataFrame, recent_samples_size: int
) -> Tuple[Dict, Dict, Dict]:
"""
Separates tools into different groups based on two use-cases:
1. Tools with enough samples to reach sample_size
2. Tools with not enough samples to reach sample_size
Args:
tools_df: DataFrame containing the tools data
sample_size: Target number of samples for each tool
Returns:
Tuple of two dictionaries containing tools that are from group 1 and group 2
"""
# Get count of samples per tool
tool_counts = tools_df["tool"].value_counts()
# Separate tools based on sample size
valid_tools = {
tool: count
for tool, count in tool_counts.items()
if count >= recent_samples_size
}
more_samples_tools = {
tool: count
for tool, count in tool_counts.items()
if count < recent_samples_size
}
return valid_tools, more_samples_tools
def downsample_tool_multiple(
tool_samples: pd.DataFrame, sample_size: int, n_subsets: int, replace: bool = False
) -> List[pd.DataFrame]:
"""
Creates multiple downsampled subsets from the tool samples
Args:
tool_samples: DataFrame containing samples for a specific tool
sample_size: Number of samples to select for each subset
n_subsets: Number of subsets to create
Returns:
List of DataFrames containing the downsampled subsets
"""
total_samples = len(tool_samples)
if total_samples < RECENTS_SAMPLES_SIZE:
raise ValueError(
f"Not enough samples available: {total_samples} < {RECENTS_SAMPLES_SIZE}"
)
print(f"total samples available: {total_samples}")
downsampled_sets = []
available_samples = tool_samples.copy()
available_samples.set_index("request_id", inplace=True, drop=False)
for i in range(n_subsets):
# If we don't have enough samples left, reset the available samples
if len(available_samples) < sample_size:
available_samples = tool_samples.copy()
# Sample without replacement from available samples
current_subset = available_samples.sample(n=sample_size, replace=replace)
downsampled_sets.append(current_subset)
# Remove used samples from available pool
available_samples = available_samples.drop(current_subset.index)
return downsampled_sets
def get_recent_samples(
tool_samples: pd.DataFrame, recent_samples_size: int
) -> pd.DataFrame:
"""
Gets the most recent samples from the tool samples
Args:
tool_samples: DataFrame containing samples for a specific tool
sample_size: Number of samples to select
Returns:
DataFrame containing the most recent samples
"""
if len(tool_samples) < recent_samples_size:
raise ValueError(
f"Not enough samples available: {len(tool_samples)} < {recent_samples_size}"
)
return tool_samples.nlargest(recent_samples_size, "request_time")
def upsample_tool_multiple(
tool_samples: pd.DataFrame, sample_size: int, n_subsets: int
) -> List[pd.DataFrame]:
"""
Creates multiple upsampled subsets from the tool samples
Args:
tool_samples: DataFrame containing samples for a specific tool
sample_size: Target number of samples for each subset
n_subsets: Number of subsets to create
Returns:
List of DataFrames containing the upsampled subsets
"""
upsampled_sets = []
for _ in range(n_subsets):
# Calculate how many additional samples we need
n_samples = len(tool_samples)
n_additional = sample_size - n_samples
# Get additional samples
additional_samples = tool_samples.sample(n=n_additional, replace=True)
# Combine original and additional samples
upsampled_set = pd.concat([tool_samples, additional_samples])
upsampled_sets.append(upsampled_set)
return upsampled_sets
def compute_subset_accuracy(tools_subset: pd.DataFrame) -> float:
"""
Computes the accuracy of a subset of tools.
Args:
tools_subset: DataFrame containing a subset of tools
Returns:
Accuracy of the subset
"""
wins = tools_subset["win"].value_counts().reindex([0, 1], fill_value=0)
# Ensure both 0 and 1 columns exist
if 0 not in wins.index:
wins[0] = 0
if 1 not in wins.index:
wins[1] = 0
return (wins[1] / (wins[0] + wins[1])) * 100
def compute_tool_estimated_accuracy(subset_list: List[pd.DataFrame]) -> Tuple:
"""
Computes the accuracy for the tool averaging the accuracy achieved within the list of subsets.
Args:
subset_list: List of DataFrames containing subsets of the same tool
Returns:
float with the average accuracy of the tool across all subsets
"""
tool_accuracies = []
for subset in subset_list:
# only one tool per subset, all
accuracy = compute_subset_accuracy(subset)
tool_accuracies.append(accuracy)
# print(f"tool accuracies: {tool_accuracies}")
accuracy_mean, accuracy_std = mean_and_std(tool_accuracies)
return accuracy_mean, accuracy_std
def update_global_accuracy(
valid_tools: Dict,
tools_df: pd.DataFrame,
global_accuracies: Dict,
sample_size: int,
nr_subsets: int = NR_SUBSETS,
one_tool: str = None,
) -> Dict:
"""
Updates the global accuracies for tools that have enough samples.
Args:
valid_tools: Dictionary of tools with enough samples
tools_df: DataFrame containing the tools data
global_accuracies: Dictionary to store global accuracies
sample_size: Target number of samples for each subset
nr_subsets: Number of subsets to create
one_tool: Optional specific tool to compute accuracy for
Returns:
Updated global accuracies dictionary
"""
if one_tool:
print(f"Computing accuracy for tool: {one_tool}")
tool_samples = tools_df[tools_df["tool"] == one_tool].copy()
# recent_samples = get_recent_samples(tool_samples, RECENTS_SAMPLES_SIZE)
downsampled_sets = downsample_tool_multiple(
tool_samples, sample_size, nr_subsets
)
tool_mean_accuracy, tool_std = compute_tool_estimated_accuracy(downsampled_sets)
global_accuracies[one_tool] = {
"mean": float(tool_mean_accuracy),
"std": tool_std,
}
return global_accuracies
for tool in valid_tools.keys():
print(f"Processing tool: {tool}")
tool_samples = tools_df[tools_df["tool"] == tool]
# recent_samples = get_recent_samples(tool_samples, RECENTS_SAMPLES_SIZE)
downsampled_sets = downsample_tool_multiple(
tool_samples, sample_size, nr_subsets
)
tool_mean_accuracy, tool_std = compute_tool_estimated_accuracy(downsampled_sets)
global_accuracies[tool] = {"mean": float(tool_mean_accuracy), "std": tool_std}
return global_accuracies
def check_upgrade_dates(
tools_df,
tools_list,
new_tools,
claude_upgrade_date="30-07-2025",
gpt_upgrade_date="03-06-2025",
) -> None:
# Convert upgrade dates to datetime
claude_upgrade_date = pd.to_datetime(claude_upgrade_date, format="%d-%m-%Y").date()
gpt_upgrade_date = pd.to_datetime(gpt_upgrade_date, format="%d-%m-%Y").date()
for tool in tools_list.keys():
print(f"checking tool {tool}")
# take the RECENT_SAMPLES from tools_df
tool_data = tools_df[tools_df["tool"] == tool]
# sort tool_data by request date in ascending order
tool_data = tool_data.sort_values(by="request_date", ascending=True)
if len(tool_data) < RECENTS_SAMPLES_SIZE:
new_tools.append(tool)
continue
recent_samples = get_recent_samples(
tool_data, recent_samples_size=RECENTS_SAMPLES_SIZE
)
recent_samples = recent_samples.sort_values(by="request_date", ascending=True)
print(recent_samples.head())
oldest_sample_date = recent_samples.iloc[0].request_date
if isinstance(oldest_sample_date, str):
oldest_sample_date = pd.to_datetime(oldest_sample_date).date()
print(f"tool {tool}: oldest sample date {oldest_sample_date}")
if tool in CLAUDE_TOOLS:
# if oldest_sample_date is before claude_upgrade_date then remove the tool
# from valid_tools and add it to the list of other_tools
if oldest_sample_date < claude_upgrade_date:
print(f"the oldest sample found is older than {claude_upgrade_date}")
new_tools.append(tool)
elif oldest_sample_date < gpt_upgrade_date:
print(f"the oldest sample found is older than {gpt_upgrade_date}")
new_tools.append(tool)
return
def check_upgraded_tools(
tools_df,
valid_tools,
other_tools,
):
"""
Function to update the input lists and remove from valid tools any tools whose oldest date is before the upgrade dates
"""
new_tools = []
# Check and remove tools from valid_tools
check_upgrade_dates(tools_df, valid_tools, new_tools)
for tool in new_tools:
if tool in valid_tools.keys():
print(f"removing tool {tool} from valid tools")
del valid_tools[tool]
# Check and remove tools from other_tools
check_upgrade_dates(tools_df, other_tools, new_tools)
for tool in new_tools:
if tool in other_tools.keys():
print(f"removing tool {tool} from other tools")
del other_tools[tool]
return new_tools
def add_historical_data(
tools_historical_file: str,
tools_df: pd.DataFrame,
more_sample_tools: Dict,
recent_samples_size: int,
valid_tools: Dict,
completed_tools: List[str],
) -> pd.DataFrame:
"""
It searches into the historical cloud files to get more samples for the tools in more_sample_tools.
"""
if not tools_historical_file:
raise ValueError(
"No historical tools file found, skipping adding historical data."
)
# get the historical tools data
tool_names = list(more_sample_tools.keys())
print(f"Downloaded historical file in tmp folder: {tools_historical_file}")
# Load the historical tools data
historical_tools_df = pd.read_parquet(TMP_DIR / tools_historical_file)
# check if the historical tools data has samples from the tools that need more samples
historical_tools_df = historical_tools_df[
historical_tools_df["tool"].isin(tool_names)
]
# check the volume of samples for the tools in the historical data
historical_tool_counts = historical_tools_df["tool"].value_counts()
for tool, count in historical_tool_counts.items():
historical_tool_data = historical_tools_df[historical_tools_df["tool"] == tool]
current_count = more_sample_tools.get(tool, 0)
# compute the minimal amount of needed samples
needed_samples = recent_samples_size - current_count
if count < needed_samples:
# update the count in more_sample_tools but do not remove the tool
print(
f"Tool {tool} has not enough samples in historical data: {count} < {needed_samples}"
)
more_sample_tools[tool] = current_count + count
print(f"updated count for tool {tool}: {more_sample_tools[tool]}")
# Combine the current tools with the historical data for that tool
tools_df = pd.concat([tools_df, historical_tool_data], ignore_index=True)
continue
print(f"Tool {tool} has enough samples now:. Adding {needed_samples} samples")
# take only the latest needed_samples in historical_tool_data
recent_samples = get_recent_samples(historical_tool_data, needed_samples)
# Combine the current tools with the historical ones
tools_df = pd.concat([tools_df, recent_samples], ignore_index=True)
tools_df = tools_df.sort_values(by="request_date", ascending=True)
valid_tools[tool] = count + needed_samples
completed_tools.append(tool)
# Remove the tool from more_sample_tools
if tool in more_sample_tools:
print(f"Deleting tool {tool} from more_sample_tools")
del more_sample_tools[tool]
return tools_df
def check_historical_samples(
global_accuracies, more_sample_tools, valid_tools, sample_size, n_subsets
):
client = initialize_client()
# first attempt: historical file download
tool_names = list(more_sample_tools.keys())
print(f"Tools with not enough samples: {tool_names}")
completed_tools = []
if len(tool_names) > 0:
print("First attempt to complete the population size")
# first attempt: historical file from 4 months ago
tools_historical_file = download_tools_historical_files(
client, skip_files_count=FILES_IN_TWO_MONTHS
)
tools_df = add_historical_data(
tools_historical_file,
tools_df,
more_sample_tools,
RECENTS_SAMPLES_SIZE,
valid_tools,
completed_tools,
)
print(more_sample_tools)
# second attempt: historical file from 6 months ago
if len(more_sample_tools) > 0:
print("Second attempt to complete the population size")
# second historical file download
tools_historical_file2 = download_tools_historical_files(
client,
skip_files_count=FILES_IN_FOUR_MONTHS,
)
tools_df = add_historical_data(
tools_historical_file2,
tools_df,
more_sample_tools,
RECENTS_SAMPLES_SIZE,
valid_tools,
completed_tools,
)
print(more_sample_tools)
if len(completed_tools) > 0:
for tool in completed_tools:
update_global_accuracy(
valid_tools,
tools_df,
global_accuracies,
sample_size,
n_subsets,
one_tool=tool,
)
return
def compute_global_accuracy_same_population(
tools_df: pd.DataFrame,
recent_samples_size: int = RECENTS_SAMPLES_SIZE,
sample_size: int = SAMPLING_POPULATION_SIZE,
n_subsets: int = NR_SUBSETS,
) -> Tuple[Dict, Dict]:
"""
For the tools in the dataset, it creates different subsets of the same size (using downsampling or upsampling) and
computes the accuracy for each subset. Finally it averages the accuracies across all subsets.
Args:
tools_df: DataFrame containing the tools data
sample_size: Target number of samples per tool
n_subsets: Number of balanced datasets to create
Returns:
List of global accuracies for the tools
"""
valid_tools, more_sample_tools = classify_tools(tools_df, recent_samples_size)
# check tools that were upgraded recently
# otherwise they will be moved as new_tools
new_tools = check_upgraded_tools(tools_df, valid_tools, more_sample_tools)
print(f"new tools {new_tools} after checking the upgraded tools")
global_accuracies = {}
if len(valid_tools) > 0:
# Compute the accuracy for tools in valid_tools
update_global_accuracy(
valid_tools, tools_df, global_accuracies, sample_size, n_subsets
)
# Check historical files for tools that need more samples
if len(more_sample_tools) > 0:
new_tools.extend(more_sample_tools.keys())
# check_historical_samples(
# global_accuracies, more_sample_tools, valid_tools, sample_size, n_subsets
# )
# for tool in more_sample_tools.keys():
# # tool but not reaching yet the population size so treated as a new tool
# new_tools.append(tool)
return global_accuracies, new_tools
def compute_global_weekly_accuracy(clean_tools_df):
"""
Compute accuracy following version 5.0 of spec"""
# get the information in clean_tools_df from last two weeks only, timestamp column is request_time
three_weeks_ago = pd.Timestamp.now(tz="UTC") - pd.Timedelta(days=7)
recent_df = clean_tools_df[clean_tools_df["request_time"] >= three_weeks_ago]
daily_requests = recent_df.groupby(clean_tools_df["request_time"].dt.date)["request_id"].count()
print("\nDaily request counts:")
print(daily_requests)
# compute at the tool level (using "tool" column) the volume of requests per tool
tool_volumes = (
recent_df.groupby("tool")["request_id"].count().reset_index(name="volume")
)
min_volume = tool_volumes["volume"].min()
max_volume = tool_volumes["volume"].max()
# compute the average volume of tool requests in the last two weeks, excluding min and max
filtered_volumes = tool_volumes[
(tool_volumes["volume"] != min_volume) & (tool_volumes["volume"] != max_volume)
]
avg_volume = filtered_volumes["volume"].mean()
print("Tool volumes in last three weeks:")
print(tool_volumes)
print(
f"Average volume of tool requests in last three weeks excluding min and max: {avg_volume}"
)
sampling_size = int(avg_volume / 2)
print(f"Sampling size = {sampling_size}")
global_accuracies, new_tools = compute_global_accuracy_same_population(
tools_df=recent_df,
recent_samples_size=avg_volume,
sample_size=sampling_size,
n_subsets=50,
)
def save_weekly_top_3_metric(recent_df):
"""Save weekly top 3 tools percentage metric"""
tool_volumes = (
recent_df.groupby("tool")["request_id"].count().reset_index(name="volume")
)
global_accuracies, _ = compute_global_accuracy_same_population(
tools_df=recent_df,
recent_samples_size=len(recent_df) // len(recent_df['tool'].unique()) if len(recent_df) > 0 else 1000,
sample_size=sampling_size,
n_subsets=50
)
tools_data = []
for _, row in tool_volumes.iterrows():
tool = row['tool']
volume = row['volume']
accuracy = global_accuracies.get(tool, {}).get('mean', 0)
tools_data.append({
'tool': tool,
'volume': volume,
'accuracy': accuracy
})
tools_df = pd.DataFrame(tools_data)
# sort by accuracy descending and get top 3
top_3_tools = tools_df.nlargest(3, 'accuracy')
top_3_requests = top_3_tools['volume'].sum()
total_requests = tools_df['volume'].sum()
top_3_percentage = (top_3_requests / total_requests) * 100 if total_requests > 0 else 0
# Create weekly report
weekly_report = {
'timestamp': pd.Timestamp.now(tz="UTC").strftime("%Y-%m-%d %H:%M:%S"),
'calculation_period': '3_weeks',
'top_3_percentage': round(top_3_percentage, 2),
'total_requests': int(total_requests),
'top_3_total_requests': int(top_3_requests),
'top_1_tool': top_3_tools.iloc[0]['tool'] if len(top_3_tools) > 0 else None,
'top_1_accuracy': round(top_3_tools.iloc[0]['accuracy'], 2) if len(top_3_tools) > 0 else None,
'top_1_volume': int(top_3_tools.iloc[0]['volume']) if len(top_3_tools) > 0 else None,
'top_2_tool': top_3_tools.iloc[1]['tool'] if len(top_3_tools) > 1 else None,
'top_2_accuracy': round(top_3_tools.iloc[1]['accuracy'], 2) if len(top_3_tools) > 1 else None,
'top_2_volume': int(top_3_tools.iloc[1]['volume']) if len(top_3_tools) > 1 else None,
'top_3_tool': top_3_tools.iloc[2]['tool'] if len(top_3_tools) > 2 else None,
'top_3_accuracy': round(top_3_tools.iloc[2]['accuracy'], 2) if len(top_3_tools) > 2 else None,
'top_3_volume': int(top_3_tools.iloc[2]['volume']) if len(top_3_tools) > 2 else None,
}
# Save to CSV
report_df = pd.DataFrame([weekly_report])
filename = ROOT_DIR / "weekly_top_3_tools_report.csv"
# Append to existing file or create new one
if os.path.exists(filename):
existing_df = pd.read_csv(filename)
combined_df = pd.concat([existing_df, report_df], ignore_index=True)
combined_df.to_csv(filename, index=False)
else:
report_df.to_csv(filename, index=False)
print(f"Weekly top 3 tools report saved: {top_3_percentage:.2f}% requests served by top 3 most accurate tools")
print(f"Report saved to: {filename}")
return weekly_report
save_weekly_top_3_metric(recent_df)
return global_accuracies, new_tools
def get_accuracy_info(clean_tools_df: pd.DataFrame) -> [pd.DataFrame, bool, List]:
"""
Extracts accuracy information from the tools DataFrame.
"""
# compute global accuracy information for the tools
# global_accuracies, new_tools = compute_global_accuracy_same_population(
# tools_df=clean_tools_df,
# )
global_accuracies, new_tools = compute_global_weekly_accuracy(
clean_tools_df=clean_tools_df
)
# transform the dictionary global_accuracies into a DataFrame
wins = pd.DataFrame(
[
{
"tool": tool,
"tool_accuracy": global_accuracies[tool]["mean"],
"std_accuracy": global_accuracies[tool]["std"],
"nr_responses": SAMPLING_POPULATION_SIZE,
}
for tool in global_accuracies.keys()
]
)
no_timeline_info = False
try:
timeline = clean_tools_df.groupby(["tool"])["request_time"].agg(["min", "max"])
print(timeline.head())
acc_info = wins.merge(timeline, how="left", on="tool")
except:
print("NO REQUEST TIME INFORMATION AVAILABLE")
no_timeline_info = True
acc_info = wins
return acc_info, no_timeline_info, new_tools
def update_tools_accuracy_same_model(
tools_acc: pd.DataFrame, tools_df: pd.DataFrame, inc_tools: List[str]
) -> pd.DataFrame:
"""To compute/update the latest accuracy information for the different mech tools"""
# computation of the accuracy information
tools_inc = tools_df[tools_df["tool"].isin(inc_tools)]
clean_tools_df = clean_tools_dataset(tools_inc)
acc_info, no_timeline_info, new_tools = get_accuracy_info(clean_tools_df)
if tools_acc is None:
print("Creating accuracy file for the first time")
return acc_info
# update the old information
print(f"accuracy information: {acc_info.head()}")
tools_to_update = list(acc_info["tool"].values)
print(f"Tools to update: {tools_to_update}")
existing_tools = list(tools_acc["tool"].values)
# dt.strftime("%Y-%m-%d %H:%M:%S")
acc_info["min"] = acc_info["min"].dt.strftime("%Y-%m-%d %H:%M:%S")
acc_info["max"] = acc_info["max"].dt.strftime("%Y-%m-%d %H:%M:%S")
all_accuracies = []
final_acc_df = pd.DataFrame(columns=tools_acc.columns)
# accuracy has been computed over the same population size
new_volume = SAMPLING_POPULATION_SIZE
for tool in tools_to_update:
if tool in new_tools or tool not in existing_tools:
new_tools.append(tool)
continue
new_accuracy = round(
acc_info[acc_info["tool"] == tool]["tool_accuracy"].values[0], 2
)
all_accuracies.append(new_accuracy)
if no_timeline_info:
new_min_timeline = None
new_max_timeline = None
else:
new_min_timeline = acc_info[acc_info["tool"] == tool]["min"].values[0]
new_max_timeline = acc_info[acc_info["tool"] == tool]["max"].values[0]
old_accuracy = tools_acc[tools_acc["tool"] == tool]["tool_accuracy"].values[0]
old_volume = tools_acc[tools_acc["tool"] == tool]["nr_responses"].values[0]
# update the accuracy information
print(f"Updating tool {tool} with new accuracy {new_accuracy}")
print(f"Old volume: {old_volume}, New volume: {new_volume}")
print(f"Old accuracy: {old_accuracy}, New accuracy: {new_accuracy}")
new_row = {
"tool": tool,
"tool_accuracy": new_accuracy,
"nr_responses": SAMPLING_POPULATION_SIZE,
"min": new_min_timeline,
"max": new_max_timeline,
}
final_acc_df = pd.concat(
[final_acc_df, pd.DataFrame([new_row])], ignore_index=True
)
# compute the average accuracy for the existing tools
if len(new_tools) > 0:
print(f"Tools with not enough samples: {new_tools}")
# compute the average accuracy for the new tools
avg_accuracy = (
round(sum(all_accuracies) / len(all_accuracies), 2)
if len(all_accuracies) > 0
else DEFAULT_ACCURACY
)
timeline = (
clean_tools_df.groupby(["tool"])["request_time"]
.agg(["min", "max"])
.reset_index()
)
print(f"Updating with average accuracy: {avg_accuracy}")
print(f"Timeline: {timeline}")
for tool in new_tools:
# timeline info?
try:
new_min_timeline = timeline[timeline["tool"] == tool]["min"].values[0]
new_max_timeline = timeline[timeline["tool"] == tool]["max"].values[0]
except:
new_min_timeline = None
new_max_timeline = None
# tool,tool_accuracy,total_requests,min,max
new_row = {
"tool": tool,
"tool_accuracy": avg_accuracy,
"nr_responses": SAMPLING_POPULATION_SIZE,
"min": new_min_timeline,
"max": new_max_timeline,
}
final_acc_df = pd.concat(
[final_acc_df, pd.DataFrame([new_row])], ignore_index=True
)
print(final_acc_df)
return final_acc_df
def compute_tools_accuracy():
"""To compute/update the latest accuracy information. Relevant dates
-- 3rd of June 2025 when the gpt 4.1 update happened
-- 30th of July 2025 when the Claude 4 update happened"""
print("Computing accuracy of tools")
print("Reading tools parquet file")
tools_df = pd.read_parquet(TMP_DIR / "tools.parquet")
# Check if the file exists
acc_data = None
if os.path.exists(ROOT_DIR / ACCURACY_FILENAME):
acc_data = pd.read_csv(ROOT_DIR / ACCURACY_FILENAME)
tools_df["request_time"] = pd.to_datetime(tools_df["request_time"])
tools_df["request_date"] = tools_df["request_time"].dt.date
tools_df["request_date"] = pd.to_datetime(tools_df["request_date"])
tools_df["request_date"] = tools_df["request_date"].dt.strftime("%Y-%m-%d")
new_acc_data = update_tools_accuracy_same_model(acc_data, tools_df, INC_TOOLS)
print("Saving into a csv file")
new_acc_data.to_csv(ROOT_DIR / ACCURACY_FILENAME, index=False)
# save the data into IPFS
# push_csv_file_to_ipfs()
def push_csv_file_to_ipfs(filename: str = ACCURACY_FILENAME) -> str:
"""Push the tools accuracy CSV file to IPFS."""
client = ipfshttpclient.connect(IPFS_SERVER)
result = client.add(ROOT_DIR / filename)
print(f"HASH of the tools accuracy file: {result['Hash']}")
return result["Hash"]
if __name__ == "__main__":
compute_tools_accuracy()
# push_csv_file_to_ipfs()
|