Spaces:
Running
Running
File size: 55,193 Bytes
08c2fcd 4429488 207f9e0 a1797f0 08c2fcd a1797f0 08c2fcd 51940dd 08c2fcd 51940dd b149ffe 51940dd 477a402 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 4b79c41 ce3f373 f51f445 08c2fcd ce3f373 08c2fcd ce3f373 08c2fcd ce3f373 08c2fcd ce3f373 08c2fcd ce3f373 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 4b79c41 ce3f373 f51f445 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 4b79c41 08c2fcd ce3f373 4b79c41 ce3f373 f51f445 4b79c41 08c2fcd 4b79c41 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 51940dd 08c2fcd 4b79c41 ce3f373 f51f445 4b79c41 08c2fcd a640033 08c2fcd c6d454e 08c2fcd c6d454e 08c2fcd c6d454e 08c2fcd |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 |
"""
Product Comparison Multi-Agent System
This module implements a multi-agent system for product comparison based on images.
The system uses various specialized agents to process images, extract features,
compare products, and provide recommendations.
Main components:
- Coordinator: Orchestrates the multi-agent workflow
- ImageProcessingAgent: Handles image recognition and object detection
- FeatureExtractionAgent: Extracts product specifications and features
- ComparisonAgent: Compares products based on extracted features
- RecommendationAgent: Provides purchase recommendations
Each agent utilizes vision models and LLMs to perform its specialized tasks.
"""
import os
import uuid
import json
import time
from typing import Dict, List, Any, Optional, Tuple
import base64
from io import BytesIO
from datetime import datetime
from threading import Thread
import torch
from PIL import Image
import numpy as np
# Import LangChain components for agent implementation
try:
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser, JsonOutputParser
from langchain.agents import create_openai_functions_agent
from langchain.agents import AgentExecutor
from langchain.memory import ConversationBufferMemory
from langchain.tools.render import format_tool_to_openai_function
from langchain_openai import ChatOpenAI
from langchain_experimental.tools.python.tool import PythonAstREPLTool
LANGCHAIN_AVAILABLE = True
except ImportError:
print("Warning: LangChain components not available. Product comparison will work with limited functionality.")
# Set all LangChain components to None
ChatPromptTemplate = None
StrOutputParser = None
JsonOutputParser = None
create_openai_functions_agent = None
AgentExecutor = None
ConversationBufferMemory = None
format_tool_to_openai_function = None
ChatOpenAI = None
PythonAstREPLTool = None
LANGCHAIN_AVAILABLE = False
# Import vision models if available
try:
from ultralytics import YOLO
except ImportError:
YOLO = None
try:
from transformers import CLIPProcessor, CLIPModel
from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers import ViTImageProcessor, ViTForImageClassification
except ImportError:
CLIPProcessor = None
CLIPModel = None
AutoModelForCausalLM = None
AutoTokenizer = None
ViTImageProcessor = None
ViTForImageClassification = None
# Session storage for SSE communication
class SessionManager:
"""Manages active product comparison sessions and their event streams"""
def __init__(self):
self.active_sessions = {}
self.results = {}
def create_session(self, session_id=None):
"""Create a new session with unique ID"""
if session_id is None:
session_id = str(uuid.uuid4())
self.active_sessions[session_id] = {
"created_at": datetime.now(),
"messages": [],
"status": "created"
}
return session_id
def add_message(self, session_id, message, agent_type="system"):
"""Add a message to the session event stream"""
if session_id not in self.active_sessions:
return False
timestamp = datetime.now().strftime("%H:%M:%S")
self.active_sessions[session_id]["messages"].append({
"message": message,
"agent": agent_type,
"timestamp": timestamp
})
return True
def get_messages(self, session_id):
"""Get all messages for a session"""
if session_id not in self.active_sessions:
return []
return self.active_sessions[session_id]["messages"]
def set_final_result(self, session_id, result):
"""Store the final analysis result for a session"""
self.results[session_id] = result
self.active_sessions[session_id]["status"] = "completed"
def get_final_result(self, session_id):
"""Get the final result for a session"""
return self.results.get(session_id)
def set_status(self, session_id, status):
"""Update session status"""
if session_id in self.active_sessions:
self.active_sessions[session_id]["status"] = status
def get_status(self, session_id):
"""Get session status"""
if session_id in self.active_sessions:
return self.active_sessions[session_id]["status"]
return None
def clean_old_sessions(self, max_age_hours=24):
"""Clean up old sessions"""
now = datetime.now()
sessions_to_remove = []
for session_id, session_data in self.active_sessions.items():
age = now - session_data["created_at"]
if age.total_seconds() > max_age_hours * 3600:
sessions_to_remove.append(session_id)
for session_id in sessions_to_remove:
del self.active_sessions[session_id]
if session_id in self.results:
del self.results[session_id]
# Initialize the session manager
session_manager = SessionManager()
# Base Agent Class
class BaseAgent:
"""Base class for all specialized agents"""
def __init__(self, name, llm=None):
self.name = name
# Use LangChain ChatOpenAI as the default LLM if none is provided
if llm is None:
try:
if LANGCHAIN_AVAILABLE and ChatOpenAI is not None:
# Initialize ChatOpenAI with environment variable for API key
api_key = os.environ.get('OPENAI_API_KEY')
if api_key:
self.llm = ChatOpenAI(
model="gpt-5",
temperature=0.7,
api_key=api_key
)
print(f"Initialized {name} with ChatOpenAI (gpt-4)")
else:
print(f"Warning: OPENAI_API_KEY not found. {name} will use fallback mode.")
self.llm = None
else:
print(f"Warning: LangChain not available. {name} will use fallback mode.")
self.llm = None
except Exception as e:
print(f"Error initializing ChatOpenAI for {name}: {e}")
self.llm = None
else:
self.llm = llm
def log(self, session_id, message):
"""Log a message to the session"""
return session_manager.add_message(session_id, message, agent_type=self.name)
async def process(self, session_id, data):
"""Process data with this agent - to be implemented by subclasses"""
raise NotImplementedError("Subclasses must implement this method")
class ImageProcessingAgent(BaseAgent):
"""Agent responsible for image processing and object recognition
This agent uses computer vision models to detect product type, brand, model,
and other visual characteristics from product images.
"""
def __init__(self, name="ImageProcessingAgent"):
super().__init__(name)
# Initialize vision models
self.models = self._load_vision_models()
def _load_vision_models(self):
"""Load vision models for product recognition"""
models = {}
# Try to load YOLO for object detection
try:
if YOLO is not None:
models["yolo"] = YOLO("yolov8n.pt")
except Exception as e:
print(f"Error loading YOLO: {e}")
# Try to load ViT for image classification
try:
if ViTImageProcessor is not None and ViTForImageClassification is not None:
models["vit_processor"] = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224")
models["vit_model"] = ViTForImageClassification.from_pretrained("google/vit-base-patch16-224")
except Exception as e:
print(f"Error loading ViT: {e}")
# Try to load CLIP for visual embedding
try:
if CLIPProcessor is not None and CLIPModel is not None:
models["clip_processor"] = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
models["clip_model"] = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
except Exception as e:
print(f"Error loading CLIP: {e}")
return models
def _process_with_yolo(self, image):
"""Process image with YOLO for object detection"""
if "yolo" not in self.models:
return {"error": "YOLO model not available"}
# Convert image to numpy array if it's a PIL image
if isinstance(image, Image.Image):
image_np = np.array(image)
else:
image_np = image
# Run inference
results = self.models["yolo"](image_np)
# Extract detection results
detections = []
for result in results:
boxes = result.boxes
for box in boxes:
class_id = int(box.cls[0].item())
class_name = result.names[class_id]
confidence = round(box.conf[0].item(), 2)
bbox = box.xyxy[0].tolist()
bbox = [round(x) for x in bbox]
detections.append({
"class": class_name,
"confidence": confidence,
"bbox": bbox
})
return {"detections": detections}
def _process_with_vit(self, image):
"""Process image with ViT for classification"""
if "vit_model" not in self.models or "vit_processor" not in self.models:
return {"error": "ViT model not available"}
# Prepare image for the model
inputs = self.models["vit_processor"](images=image, return_tensors="pt")
# Run inference
with torch.no_grad():
outputs = self.models["vit_model"](**inputs)
logits = outputs.logits
# Get top 5 predictions
probs = torch.nn.functional.softmax(logits, dim=-1)[0]
top5_prob, top5_indices = torch.topk(probs, 5)
results = []
for i, (prob, idx) in enumerate(zip(top5_prob, top5_indices)):
class_name = self.models["vit_model"].config.id2label[idx.item()]
results.append({
"rank": i+1,
"class": class_name,
"probability": round(prob.item(), 3)
})
return {"classifications": results}
def _extract_product_info_from_vision(self, image, results):
"""Extract product information using LLM and vision results"""
if self.llm is None:
# Provide a basic extraction based on detection results only
if "detections" in results and results["detections"]:
detections = results["detections"]
# Get the most confident detection
top_detection = max(detections, key=lambda x: x["confidence"])
return {
"product_type": top_detection["class"],
"confidence": top_detection["confidence"]
}
return {"product_type": "unknown"}
# If we have an LLM, we can do more sophisticated extraction
prompt = f"""Analyze this product image detection results and extract detailed product information.
Detection results: {json.dumps(results)}
Extract the following information in JSON format:
- product_type: The category of the product (car, smartphone, laptop, etc.)
- brand: The most likely brand of the product
- model: Any model information that can be determined
- color: The main color of the product
- key_features: List of notable visual features
Return only valid JSON format."""
try:
# Use LangChain's invoke method
response = self.llm.invoke(prompt)
# Extract content from LangChain response
if hasattr(response, 'content'):
response_text = response.content
else:
response_text = str(response)
# Try to parse as JSON
try:
extracted = json.loads(response_text)
return extracted
except json.JSONDecodeError:
# If LLM output is not valid JSON, extract key information using simple parsing
lines = response_text.split('\n')
extracted = {}
for line in lines:
if ':' in line:
key, value = line.split(':', 1)
key = key.strip().lower().replace(' ', '_').strip('"').strip("'")
# Clean up value more thoroughly
value = value.strip().rstrip(',').rstrip(';')
# Remove all types of quotes completely for cleaner values
value = value.strip('"').strip("'")
# Handle empty quoted strings
if value in ['""', "''", '""""', "''''", '']:
value = "Unknown"
# Clean up array-like strings - convert to proper arrays
if value.startswith('[') and value.endswith(']'):
try:
# Try to parse as actual array
import ast
parsed_array = ast.literal_eval(value)
if isinstance(parsed_array, list):
value = parsed_array
except:
# If parsing fails, clean up the string
value = value.strip('[]').split(',') if value != '[]' else []
elif value.startswith('[') and not value.endswith(']'):
value = []
if key and value not in ['null', 'None', None]:
extracted[key] = value
return extracted
except Exception as e:
print(f"Error extracting product info: {e}")
return {"product_type": "unknown", "error": str(e)}
async def process(self, session_id, data):
"""Process product images to identify products and extract features"""
self.log(session_id, "Starting image analysis to identify products...")
results = {}
product_info = {}
# Process each image if available
images = data.get("images", [])
for i, img in enumerate(images):
if img is None:
continue
image_key = f"image{i+1}"
self.log(session_id, f"Processing {image_key}...")
# Run object detection
yolo_results = self._process_with_yolo(img)
# Run classification
vit_results = self._process_with_vit(img)
# Combine results
vision_results = {
**yolo_results,
**vit_results
}
# Extract product information from vision results
info = self._extract_product_info_from_vision(img, vision_results)
self.log(session_id, f"Identified product in {image_key}: {info.get('product_type', 'unknown')}")
if "brand" in info:
self.log(session_id, f"Detected brand: {info['brand']}")
results[image_key] = {
"vision_results": vision_results,
"product_info": info
}
product_info[image_key] = info
self.log(session_id, "Image processing completed")
return {
"vision_results": results,
"product_info": product_info
}
class FeatureExtractionAgent(BaseAgent):
"""Agent responsible for extracting detailed product features and specifications
This agent analyzes image processing results and uses LLMs to extract detailed
product specifications, features, and technical details.
"""
def __init__(self, name="FeatureExtractionAgent"):
super().__init__(name)
def _extract_specifications(self, product_info):
"""Extract detailed specifications from product information"""
if self.llm is None:
# If no LLM is available, return basic specs from product info
return product_info
# Prepare prompt for specification extraction
product_type = product_info.get("product_type", "unknown")
prompt = f"""Based on this product information, generate a detailed list of specifications.
Product information: {json.dumps(product_info)}
For a {product_type}, extract or infer the following specifications.
IMPORTANT: Return ONLY valid JSON format. Do not include any explanatory text before or after the JSON.
Required JSON structure for a {product_type}:
"""
# Add product-specific specification types based on product type
if "car" in product_type.lower() or "vehicle" in product_type.lower():
prompt += """{
"make": "The manufacturer of the car",
"model": "The model name",
"year": "Estimated year of manufacture",
"body_type": "The body style (sedan, SUV, etc.)",
"fuel_type": "The fuel type if identifiable",
"engine": "Engine specifications if identifiable",
"color": "Exterior color",
"features": ["List", "of", "visible", "features"]
}
Example: {"make": "Toyota", "model": "Camry", "year": "2020", "body_type": "Sedan", "fuel_type": "Gasoline", "engine": "2.5L", "color": "Blue", "features": ["LED headlights", "Alloy wheels"]}"""
elif "phone" in product_type.lower() or "smartphone" in product_type.lower():
prompt += """{
"brand": "The manufacturer",
"model": "The phone model",
"screen_size": "Estimated screen size",
"camera": "Visible camera specifications",
"color": "Device color",
"generation": "Device generation if identifiable",
"features": ["List", "of", "visible", "features"]
}
Example: {"brand": "Apple", "model": "iPhone 14", "screen_size": "6.1 inches", "camera": "Dual camera", "color": "Blue", "generation": "14th", "features": ["Face ID", "Wireless charging"]}"""
elif "laptop" in product_type.lower() or "computer" in product_type.lower():
prompt += """{
"brand": "The manufacturer",
"model": "The computer model",
"screen_size": "Estimated screen size",
"form_factor": "Laptop, desktop, all-in-one, etc.",
"color": "Device color",
"features": ["List", "of", "visible", "features"]
}
Example: {"brand": "Dell", "model": "XPS 13", "screen_size": "13.3 inches", "form_factor": "Laptop", "color": "Silver", "features": ["Touchscreen", "Backlit keyboard"]}"""
else:
# Generic product specifications
prompt += """{
"brand": "The manufacturer if identifiable",
"model": "The product model if identifiable",
"color": "Product color",
"dimensions": "Estimated dimensions",
"features": ["List", "of", "visible", "features"],
"materials": "Visible materials used"
}
Example: {"brand": "Unknown", "model": "Unknown", "color": "Black", "dimensions": "Medium", "features": ["Modern design"], "materials": "Plastic"}"""
try:
# Use LangChain's invoke method
response = self.llm.invoke(prompt)
# Extract content from LangChain response
if hasattr(response, 'content'):
response_text = response.content
else:
response_text = str(response)
# Try to parse as JSON
try:
specs = json.loads(response_text)
return specs
except json.JSONDecodeError:
# If LLM output is not valid JSON, extract key information using simple parsing
lines = response_text.split('\n')
specs = {}
for line in lines:
if ':' in line:
key, value = line.split(':', 1)
key = key.strip().lower().replace(' ', '_').strip('"').strip("'")
# Clean up value more thoroughly
value = value.strip().rstrip(',').rstrip(';')
# Remove all types of quotes completely for cleaner values
value = value.strip('"').strip("'")
# Handle empty quoted strings
if value in ['""', "''", '""""', "''''", '']:
value = "Unknown"
# Clean up array-like strings - convert to proper arrays
if value.startswith('[') and value.endswith(']'):
try:
# Try to parse as actual array
import ast
parsed_array = ast.literal_eval(value)
if isinstance(parsed_array, list):
value = parsed_array
except:
# If parsing fails, clean up the string
value = value.strip('[]').split(',') if value != '[]' else []
elif value.startswith('[') and not value.endswith(']'):
value = []
if key and value not in ['null', 'None', None]:
specs[key] = value
return specs
except Exception as e:
print(f"Error extracting specifications: {e}")
return {"error": str(e)}
def _get_feature_highlights(self, specs):
"""Extract key feature highlights from specifications"""
if not specs or not isinstance(specs, dict):
return []
highlights = []
# Extract key features based on product type
product_type = specs.get("product_type", "").lower()
if "car" in product_type or "vehicle" in product_type:
# Highlight car features
if "make" in specs and "model" in specs:
highlights.append(f"{specs['make']} {specs['model']}")
if "year" in specs:
highlights.append(f"{specs['year']} model")
if "engine" in specs:
highlights.append(f"Engine: {specs['engine']}")
if "body_type" in specs:
highlights.append(f"{specs['body_type']} body style")
elif "phone" in product_type or "smartphone" in product_type:
# Highlight phone features
if "brand" in specs and "model" in specs:
highlights.append(f"{specs['brand']} {specs['model']}")
if "screen_size" in specs:
highlights.append(f"{specs['screen_size']} display")
if "camera" in specs:
highlights.append(f"Camera: {specs['camera']}")
elif "laptop" in product_type or "computer" in product_type:
# Highlight laptop features
if "brand" in specs and "model" in specs:
highlights.append(f"{specs['brand']} {specs['model']}")
if "screen_size" in specs:
highlights.append(f"{specs['screen_size']} display")
# Generic highlights for any product
if "features" in specs:
if isinstance(specs["features"], list):
highlights.extend(specs["features"][:3]) # Top 3 features
elif isinstance(specs["features"], str):
features = specs["features"].split(",")
highlights.extend([f.strip() for f in features[:3]]) # Top 3 features
# Add color as a feature if available
if "color" in specs:
highlights.append(f"{specs['color']} color")
return highlights
async def process(self, session_id, data):
"""Process product information to extract detailed specifications"""
self.log(session_id, "Extracting detailed product specifications...")
results = {}
product_info = data.get("product_info", {})
if not product_info:
self.log(session_id, "No product information available for specification extraction")
return {"specifications": {}}
# Process each product
for key, info in product_info.items():
self.log(session_id, f"Extracting specifications for {key}...")
# Extract detailed specifications
specs = self._extract_specifications(info)
# Get feature highlights
highlights = self._get_feature_highlights(specs)
# Log results
if highlights:
self.log(session_id, f"Key features for {key}: {', '.join(highlights[:3])}")
results[key] = {
"specifications": specs,
"highlights": highlights
}
self.log(session_id, "Feature extraction completed")
return {"specifications": results}
class ComparisonAgent(BaseAgent):
"""Agent responsible for comparing products based on their specifications
This agent analyzes the specifications of multiple products and identifies
the key differences, advantages, and disadvantages between them.
"""
def __init__(self, name="ComparisonAgent"):
super().__init__(name)
def _compare_specifications(self, specs1, specs2):
"""Compare two sets of product specifications"""
if not specs1 or not specs2:
return {"error": "Insufficient specification data for comparison"}
# If we have an LLM, use it for more sophisticated comparison
if self.llm is not None:
prompt = f"""Compare these two products based on their specifications.
Product 1: {json.dumps(specs1)}
Product 2: {json.dumps(specs2)}
Provide a detailed comparison in JSON format with the following structure:
{{"differences": [...], "product1_advantages": [...], "product2_advantages": [...], "summary": "..."}}
- differences: List key differences between the products
- product1_advantages: List advantages of Product 1 over Product 2
- product2_advantages: List advantages of Product 2 over Product 1
- summary: A concise summary of the comparison
"""
try:
# Use LangChain's invoke method
response = self.llm.invoke(prompt)
# Extract content from LangChain response
if hasattr(response, 'content'):
response_text = response.content
else:
response_text = str(response)
# Try to parse as JSON
try:
comparison = json.loads(response_text)
return comparison
except json.JSONDecodeError:
# If LLM output is not valid JSON, extract key sections using simple parsing
lines = response_text.split('\n')
comparison = {}
for line in lines:
if ':' in line:
key, value = line.split(':', 1)
key = key.strip().lower().replace(' ', '_').strip('"').strip("'")
# Clean up value more thoroughly
value = value.strip().rstrip(',').rstrip(';')
# Remove all types of quotes completely for cleaner values
value = value.strip('"').strip("'")
# Handle empty quoted strings
if value in ['""', "''", '""""', "''''", '']:
value = "Unknown"
# Clean up array-like strings - convert to proper arrays
if value.startswith('[') and value.endswith(']'):
try:
# Try to parse as actual array
import ast
parsed_array = ast.literal_eval(value)
if isinstance(parsed_array, list):
value = parsed_array
except:
# If parsing fails, clean up the string
value = value.strip('[]').split(',') if value != '[]' else []
elif value.startswith('[') and not value.endswith(']'):
value = []
if key and value not in ['null', 'None', None]:
comparison[key] = value
return comparison
except Exception as e:
print(f"Error in LLM comparison: {e}")
# Fall back to simple comparison
pass
# Simple comparison logic as fallback
differences = []
product1_advantages = []
product2_advantages = []
# Identify common keys to compare
all_keys = set(list(specs1.keys()) + list(specs2.keys()))
# Exclude utility keys like 'error'
exclude_keys = {'error', 'product_type', 'confidence'}
compare_keys = all_keys - exclude_keys
for key in compare_keys:
val1 = specs1.get(key)
val2 = specs2.get(key)
if val1 is None and val2 is not None:
differences.append(f"Product 2 has {key}: {val2}, but Product 1 doesn't")
product2_advantages.append(f"Has {key}: {val2}")
elif val1 is not None and val2 is None:
differences.append(f"Product 1 has {key}: {val1}, but Product 2 doesn't")
product1_advantages.append(f"Has {key}: {val1}")
elif val1 != val2:
differences.append(f"Different {key}: Product 1 has {val1}, Product 2 has {val2}")
# Try to determine advantages based on common metrics
if key in ['price', 'weight']:
# Lower is generally better
try:
num1 = float(str(val1).split()[0])
num2 = float(str(val2).split()[0])
if num1 < num2:
product1_advantages.append(f"Lower {key}: {val1}")
else:
product2_advantages.append(f"Lower {key}: {val2}")
except (ValueError, IndexError):
pass
elif key in ['screen_size', 'storage', 'memory', 'ram', 'battery', 'capacity']:
# Higher is generally better
try:
num1 = float(str(val1).split()[0])
num2 = float(str(val2).split()[0])
if num1 > num2:
product1_advantages.append(f"Higher {key}: {val1}")
else:
product2_advantages.append(f"Higher {key}: {val2}")
except (ValueError, IndexError):
pass
# Create a simple summary
product1_type = specs1.get('product_type', 'Product 1')
product2_type = specs2.get('product_type', 'Product 2')
summary = f"Comparison between {product1_type} and {product2_type} reveals {len(differences)} key differences."
if len(product1_advantages) > len(product2_advantages):
summary += f" {product1_type} appears to have more advantages."
elif len(product2_advantages) > len(product1_advantages):
summary += f" {product2_type} appears to have more advantages."
else:
summary += " Both products have similar number of advantages."
return {
"differences": differences,
"product1_advantages": product1_advantages,
"product2_advantages": product2_advantages,
"summary": summary
}
async def process(self, session_id, data):
"""Compare products based on their specifications"""
self.log(session_id, "Starting product comparison analysis...")
specifications = data.get("specifications", {})
if len(specifications) < 2:
self.log(session_id, "Not enough products to compare")
return {"comparison": {"error": "Need at least two products to compare"}}
# Get the product keys (image1, image2, etc.)
product_keys = list(specifications.keys())
if len(product_keys) > 2:
self.log(session_id, f"Found {len(product_keys)} products, comparing the first two only")
product_keys = product_keys[:2]
# Get the specifications for each product
product1_specs = specifications.get(product_keys[0], {}).get("specifications", {})
product2_specs = specifications.get(product_keys[1], {}).get("specifications", {})
# Perform comparison
comparison = self._compare_specifications(product1_specs, product2_specs)
# Log comparison results
if "differences" in comparison:
num_diff = len(comparison["differences"])
self.log(session_id, f"Found {num_diff} key differences between the products")
# Log a few example differences
if num_diff > 0:
for i, diff in enumerate(comparison["differences"][:3]):
self.log(session_id, f"Difference {i+1}: {diff}")
# Log summary if available
if "summary" in comparison:
self.log(session_id, f"Comparison summary: {comparison['summary']}")
self.log(session_id, "Comparison analysis completed")
return {
"comparison": comparison,
"product_keys": product_keys
}
class RecommendationAgent(BaseAgent):
"""Agent responsible for providing purchase recommendations
This agent analyzes product comparisons and provides personalized recommendations
based on the user's needs and preferences.
"""
def __init__(self, name="RecommendationAgent"):
super().__init__(name)
def _generate_recommendation(self, comparison, product_keys, specifications):
"""Generate a purchase recommendation based on comparison results"""
if not comparison or "error" in comparison:
return {
"recommendation": "Insufficient data to make a recommendation",
"reason": "Could not compare the products",
"confidence": 0.0
}
# Get product information
product1_key = product_keys[0]
product2_key = product_keys[1]
product1_specs = specifications.get(product1_key, {}).get("specifications", {})
product2_specs = specifications.get(product2_key, {}).get("specifications", {})
product1_type = product1_specs.get("product_type", product1_key)
product2_type = product2_specs.get("product_type", product2_key)
# If we have an LLM, use it for more sophisticated recommendation
if self.llm is not None:
prompt = f"""Based on this product comparison, provide a purchase recommendation.
Product 1: {json.dumps(product1_specs)}
Product 2: {json.dumps(product2_specs)}
Comparison: {json.dumps(comparison)}
Provide a recommendation in JSON format with the following structure:
{{"recommended_product": "1 or 2", "recommendation": "...", "reason": "...", "confidence": 0.0-1.0, "use_cases": [...]}}
- recommended_product: Either "1" or "2" indicating which product is recommended
- recommendation: A concise recommendation statement
- reason: The main reason for the recommendation
- confidence: A confidence score between 0 and 1
- use_cases: List of ideal use cases for the recommended product
"""
try:
# Use LangChain's invoke method
response = self.llm.invoke(prompt)
# Extract content from LangChain response
if hasattr(response, 'content'):
response_text = response.content
else:
response_text = str(response)
# Try to parse as JSON
try:
recommendation = json.loads(response_text)
return recommendation
except json.JSONDecodeError:
# If LLM output is not valid JSON, extract key information using simple parsing
lines = response_text.split('\n')
recommendation = {}
for line in lines:
if ':' in line:
key, value = line.split(':', 1)
key = key.strip().lower().replace(' ', '_').strip('"').strip("'")
# Clean up value more thoroughly
value = value.strip().rstrip(',').rstrip(';')
# Remove all types of quotes completely for cleaner values
value = value.strip('"').strip("'")
# Handle empty quoted strings
if value in ['""', "''", '""""', "''''", '']:
value = "Unknown"
# Clean up array-like strings - convert to proper arrays
if value.startswith('[') and value.endswith(']'):
try:
# Try to parse as actual array
import ast
parsed_array = ast.literal_eval(value)
if isinstance(parsed_array, list):
value = parsed_array
except:
# If parsing fails, clean up the string
value = value.strip('[]').split(',') if value != '[]' else []
elif value.startswith('[') and not value.endswith(']'):
value = []
if key and value not in ['null', 'None', None]:
recommendation[key] = value
return recommendation
except Exception as e:
print(f"Error in LLM recommendation: {e}")
# Fall back to simple recommendation
pass
# Simple recommendation logic as fallback
product1_advantages = comparison.get("product1_advantages", [])
product2_advantages = comparison.get("product2_advantages", [])
# Count advantages
p1_count = len(product1_advantages)
p2_count = len(product2_advantages)
# Simple confidence calculation
total_advantages = p1_count + p2_count
if total_advantages == 0:
confidence = 0.5 # Can't determine
else:
# Maximum confidence is 0.95, minimum is 0.55
max_confidence = 0.95
min_confidence = 0.55
base_confidence = max(p1_count, p2_count) / total_advantages if total_advantages > 0 else 0.5
confidence = min_confidence + base_confidence * (max_confidence - min_confidence)
# Make recommendation
if p1_count > p2_count:
recommended = "1"
recommendation = f"{product1_type} is recommended over {product2_type}"
reason = f"It has {p1_count} advantages compared to {p2_count} for the alternative"
elif p2_count > p1_count:
recommended = "2"
recommendation = f"{product2_type} is recommended over {product1_type}"
reason = f"It has {p2_count} advantages compared to {p1_count} for the alternative"
else:
# Equal advantages, slight preference for first product
recommended = "1"
recommendation = f"Both {product1_type} and {product2_type} appear equally matched"
reason = "Consider your specific needs as both have similar advantages"
confidence = 0.5 # Equal confidence
# Determine use cases based on advantages
use_cases = []
if recommended == "1" and product1_advantages:
# Extract use cases from product 1's advantages
use_cases = [f"When {adv.lower()}" for adv in product1_advantages[:3]]
elif recommended == "2" and product2_advantages:
# Extract use cases from product 2's advantages
use_cases = [f"When {adv.lower()}" for adv in product2_advantages[:3]]
# Add generic use case if none found
if not use_cases:
use_cases = [f"General {product1_type if recommended == '1' else product2_type} usage"]
return {
"recommended_product": recommended,
"recommendation": recommendation,
"reason": reason,
"confidence": round(confidence, 2),
"use_cases": use_cases
}
async def process(self, session_id, data):
"""Generate purchase recommendations based on product comparison"""
self.log(session_id, "Generating product recommendations...")
# Get comparison data
comparison = data.get("comparison", {})
product_keys = data.get("product_keys", [])
specifications = data.get("specifications", {})
if not comparison or not product_keys or len(product_keys) < 2:
self.log(session_id, "Insufficient data for recommendation")
return {"recommendation": {"error": "Not enough data to generate recommendation"}}
# Generate recommendation
recommendation = self._generate_recommendation(comparison, product_keys, specifications)
# Log recommendation
self.log(session_id, f"Recommendation: {recommendation.get('recommendation', 'No recommendation')}")
self.log(session_id, f"Reason: {recommendation.get('reason', 'No reason provided')}")
# Log confidence
confidence = recommendation.get('confidence', 0)
try:
# Handle various confidence formats
if isinstance(confidence, str):
# Extract first valid number from string like '0.70.70.70...'
confidence_clean = confidence.split('.')[0] + '.' + confidence.split('.')[1] if '.' in confidence else confidence
confidence = float(confidence_clean)
confidence = float(confidence)
confidence_percent = f"{int(confidence * 100)}%"
except (ValueError, IndexError):
confidence_percent = "Unknown"
self.log(session_id, f"Confidence in recommendation: {confidence_percent}")
self.log(session_id, "Recommendation generation completed")
return {"recommendation": recommendation}
class ProductComparisonCoordinator:
"""Main coordinator for the product comparison multi-agent system
This class orchestrates the entire product comparison workflow by managing
all the specialized agents and their interactions.
"""
def __init__(self):
# Initialize all agents
self.image_processor = ImageProcessingAgent()
self.feature_extractor = FeatureExtractionAgent()
self.comparison_agent = ComparisonAgent()
self.recommendation_agent = RecommendationAgent()
async def process_images(self, session_id, images, session_metadata=None):
"""Process images through the entire multi-agent workflow
Args:
session_id: Unique session identifier
images: List of image data (PIL Images or numpy arrays)
session_metadata: Optional dictionary with additional session information
Returns:
Dictionary containing the final analysis results
"""
# Initialize session if it doesn't exist
if session_manager.get_status(session_id) is None:
session_manager.create_session(session_id)
# Set default metadata if not provided
if session_metadata is None:
session_metadata = {}
# Get analysis type from metadata
analysis_type = session_metadata.get('analysis_type', 'info')
session_manager.set_status(session_id, "processing")
session_manager.add_message(session_id, f"Starting product {analysis_type} analysis")
try:
# Step 1: Process images with Image Processing Agent
session_manager.add_message(session_id, "Step 1: Analyzing product images...")
image_results = await self.image_processor.process(session_id, {"images": images})
# Check if we have enough products to compare
product_info = image_results.get("product_info", {})
if len(product_info) < 1:
session_manager.add_message(session_id, "Error: No products detected in images")
session_manager.set_status(session_id, "error")
return {"error": "No products detected in images"}
# Step 2: Extract features with Feature Extraction Agent
session_manager.add_message(session_id, "Step 2: Extracting product specifications...")
try:
feature_results = await self.feature_extractor.process(session_id, image_results)
except Exception as e:
session_manager.add_message(session_id, f"Warning: Feature extraction failed: {str(e)}")
feature_results = {"specifications": {"error": f"Feature extraction failed: {str(e)}"}}
# Step 3: Compare products with Comparison Agent if we have multiple products
comparison_results = {}
if len(product_info) >= 2:
session_manager.add_message(session_id, "Step 3: Comparing products...")
try:
comparison_results = await self.comparison_agent.process(
session_id,
{**feature_results, "specifications": feature_results.get("specifications", {})}
)
except Exception as e:
session_manager.add_message(session_id, f"Warning: Comparison failed: {str(e)}")
comparison_results = {"comparison": {"error": f"Comparison failed: {str(e)}"}}
# Step 4: Generate recommendation with Recommendation Agent
session_manager.add_message(session_id, "Step 4: Generating purchase recommendation...")
try:
recommendation_results = await self.recommendation_agent.process(
session_id,
{**comparison_results, "specifications": feature_results.get("specifications", {})}
)
except Exception as e:
session_manager.add_message(session_id, f"Warning: Recommendation failed: {str(e)}")
recommendation_results = {"recommendation": {"error": f"Recommendation failed: {str(e)}"}}
else:
session_manager.add_message(session_id, "Skipping comparison: Only one product detected")
comparison_results = {"comparison": {"error": "Need at least two products to compare"}}
recommendation_results = {"recommendation": {"error": "Need at least two products to compare"}}
# Tailor results based on analysis type
final_results = {
"status": "completed"
}
# Include results based on analysis type
if analysis_type == 'info':
final_results["productInfo"] = image_results.get("product_info", {})
final_results["specifications"] = feature_results.get("specifications", {})
elif analysis_type == 'compare':
final_results["comparison"] = comparison_results.get("comparison", {})
final_results["productInfo"] = image_results.get("product_info", {})
elif analysis_type == 'value':
# Value analysis combines specs and comparison data
final_results["valueAnalysis"] = {
"priceA": "$" + str(1000 + int(hash(str(session_id)) % 500)), # Mock price for demo
"valueScoreA": 7 + (int(hash(str(session_id)) % 3)), # Mock score between 7-9
"analysis": "Based on the specifications and market positioning, this product offers good value for money."
}
if len(product_info) >= 2:
final_results["valueAnalysis"]["priceB"] = "$" + str(1200 + int(hash(str(session_id + "B")) % 500))
final_results["valueAnalysis"]["valueScoreB"] = 6 + (int(hash(str(session_id + "B")) % 4))
elif analysis_type == 'recommend':
if recommendation_results.get("recommendation", {}).get("error"):
# Create a mock recommendation if only one product
if len(product_info) == 1:
product_name = next(iter(product_info.values())).get("name", "Product")
final_results["recommendation"] = {
"recommendedProduct": product_name,
"reason": "This is the only product analyzed and appears to meet standard quality benchmarks.",
"confidence": 0.85,
"alternatives": [
{"name": "Similar model with higher storage", "reason": "If you need more storage capacity"},
{"name": "Budget alternative", "reason": "If price is your primary concern"}
],
"buyingTips": [
"Wait for seasonal sales for the best price",
"Check for warranty terms before purchasing"
]
}
else:
final_results["recommendation"] = recommendation_results.get("recommendation", {})
else:
final_results["recommendation"] = recommendation_results.get("recommendation", {})
# Always include basic product info and vision results for context
final_results["vision_results"] = image_results.get("vision_results", {})
if "productInfo" not in final_results: # Don't duplicate if already added
final_results["productInfo"] = image_results.get("product_info", {})
# Set final results in session manager
session_manager.set_final_result(session_id, final_results)
session_manager.set_status(session_id, "completed")
session_manager.add_message(session_id, "Product comparison analysis completed successfully")
return final_results
except Exception as e:
error_msg = f"Error during product comparison: {str(e)}"
session_manager.add_message(session_id, error_msg)
session_manager.set_status(session_id, "error")
return {"error": error_msg}
def get_session_messages(self, session_id):
"""Get all messages for a session"""
return session_manager.get_messages(session_id)
def get_session_result(self, session_id):
"""Get the final result for a session"""
return session_manager.get_final_result(session_id)
def get_session_status(self, session_id):
"""Get the status of a session"""
return session_manager.get_status(session_id)
# Helper function to create a coordinator instance
def get_product_comparison_coordinator():
"""Get a singleton instance of the ProductComparisonCoordinator"""
if not hasattr(get_product_comparison_coordinator, "_instance"):
get_product_comparison_coordinator._instance = ProductComparisonCoordinator()
return get_product_comparison_coordinator._instance
# Helper function to convert base64 image data to PIL Image
def decode_base64_image(base64_data):
"""Convert base64 image data to PIL Image"""
try:
# Check if the base64 data includes a data URL prefix
if base64_data.startswith('data:image'):
# Extract the actual base64 data after the comma
base64_data = base64_data.split(',', 1)[1]
# Decode base64 data
image_data = base64.b64decode(base64_data)
# Convert to PIL Image
image = Image.open(BytesIO(image_data))
return image
except Exception as e:
print(f"Error decoding base64 image: {e}")
return None
|