Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,6 +1,7 @@
|
|
| 1 |
# ================================================================
|
| 2 |
-
# Wine Quality
|
| 3 |
-
#
|
|
|
|
| 4 |
# ================================================================
|
| 5 |
|
| 6 |
import os
|
|
@@ -21,8 +22,6 @@ from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
|
| 21 |
APP_TITLE = "π· Wine Quality Research Assistant"
|
| 22 |
|
| 23 |
EMBEDDING_MODEL_ID = "sentence-transformers/all-MiniLM-L6-v2"
|
| 24 |
-
|
| 25 |
-
# Lightweight model for Hugging Face CPU Space
|
| 26 |
LLM_MODEL_ID = "google/flan-t5-small"
|
| 27 |
|
| 28 |
TOP_K = 3
|
|
@@ -33,42 +32,30 @@ DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
|
|
| 33 |
|
| 34 |
|
| 35 |
print("=" * 70)
|
| 36 |
-
print("
|
| 37 |
print("=" * 70)
|
| 38 |
|
| 39 |
print(f"Device: {DEVICE}")
|
| 40 |
-
print(f"Gradio
|
| 41 |
|
| 42 |
|
| 43 |
# ================================================================
|
| 44 |
-
# 2. FIND PDF
|
| 45 |
# ================================================================
|
| 46 |
|
| 47 |
def find_pdf():
|
| 48 |
-
"""
|
| 49 |
-
Automatically find the first PDF file in the
|
| 50 |
-
Hugging Face Space repository.
|
| 51 |
-
"""
|
| 52 |
-
|
| 53 |
-
files = os.listdir(".")
|
| 54 |
|
| 55 |
pdf_files = [
|
| 56 |
file
|
| 57 |
-
for file in
|
| 58 |
if file.lower().endswith(".pdf")
|
| 59 |
]
|
| 60 |
|
| 61 |
if not pdf_files:
|
| 62 |
|
| 63 |
-
print("\nFiles currently available in the Space:")
|
| 64 |
-
|
| 65 |
-
for file in files:
|
| 66 |
-
print(" -", file)
|
| 67 |
-
|
| 68 |
raise FileNotFoundError(
|
| 69 |
-
"
|
| 70 |
-
"
|
| 71 |
-
"folder of the Hugging Face Space."
|
| 72 |
)
|
| 73 |
|
| 74 |
print("\nPDF files detected:")
|
|
@@ -81,25 +68,25 @@ def find_pdf():
|
|
| 81 |
|
| 82 |
PDF_FILE = find_pdf()
|
| 83 |
|
| 84 |
-
print(f"\
|
| 85 |
|
| 86 |
|
| 87 |
# ================================================================
|
| 88 |
-
# 3.
|
| 89 |
# ================================================================
|
| 90 |
|
| 91 |
def clean_text(text):
|
| 92 |
-
"""
|
| 93 |
-
Clean text extracted from the PDF.
|
| 94 |
-
"""
|
| 95 |
|
| 96 |
if not text:
|
| 97 |
return ""
|
| 98 |
|
| 99 |
text = text.replace("\x00", " ")
|
| 100 |
|
| 101 |
-
|
| 102 |
-
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
return text.strip()
|
| 105 |
|
|
@@ -109,23 +96,21 @@ def clean_text(text):
|
|
| 109 |
# ================================================================
|
| 110 |
|
| 111 |
def read_pdf(pdf_path):
|
| 112 |
-
"""
|
| 113 |
-
Extract text from each PDF page while preserving page numbers.
|
| 114 |
-
"""
|
| 115 |
-
|
| 116 |
-
print("\nπ Reading PDF...")
|
| 117 |
|
| 118 |
reader = PdfReader(pdf_path)
|
| 119 |
|
| 120 |
pages = []
|
| 121 |
|
| 122 |
-
for page_number, page in enumerate(
|
|
|
|
|
|
|
|
|
|
| 123 |
|
| 124 |
try:
|
| 125 |
|
| 126 |
-
text =
|
| 127 |
-
|
| 128 |
-
|
| 129 |
|
| 130 |
if text:
|
| 131 |
|
|
@@ -139,7 +124,7 @@ def read_pdf(pdf_path):
|
|
| 139 |
except Exception as error:
|
| 140 |
|
| 141 |
print(
|
| 142 |
-
f"
|
| 143 |
f"{page_number}: {error}"
|
| 144 |
)
|
| 145 |
|
|
@@ -152,31 +137,27 @@ pages = read_pdf(PDF_FILE)
|
|
| 152 |
if not pages:
|
| 153 |
|
| 154 |
raise ValueError(
|
| 155 |
-
"
|
| 156 |
"could be extracted."
|
| 157 |
)
|
| 158 |
|
| 159 |
|
| 160 |
print(
|
| 161 |
-
f"
|
| 162 |
)
|
| 163 |
|
| 164 |
|
| 165 |
# ================================================================
|
| 166 |
-
# 5. CREATE
|
| 167 |
# ================================================================
|
| 168 |
|
| 169 |
def create_chunks(pages):
|
| 170 |
-
"""
|
| 171 |
-
Break extracted PDF text into overlapping chunks.
|
| 172 |
-
"""
|
| 173 |
|
| 174 |
chunks = []
|
| 175 |
|
| 176 |
for page_data in pages:
|
| 177 |
|
| 178 |
text = page_data["text"]
|
| 179 |
-
|
| 180 |
page_number = page_data["page"]
|
| 181 |
|
| 182 |
start = 0
|
|
@@ -185,7 +166,9 @@ def create_chunks(pages):
|
|
| 185 |
|
| 186 |
end = start + CHUNK_SIZE
|
| 187 |
|
| 188 |
-
chunk_text = text[
|
|
|
|
|
|
|
| 189 |
|
| 190 |
if len(chunk_text) > 50:
|
| 191 |
|
|
@@ -196,23 +179,19 @@ def create_chunks(pages):
|
|
| 196 |
}
|
| 197 |
)
|
| 198 |
|
| 199 |
-
start +=
|
|
|
|
|
|
|
|
|
|
| 200 |
|
| 201 |
return chunks
|
| 202 |
|
| 203 |
|
| 204 |
chunks = create_chunks(pages)
|
| 205 |
|
| 206 |
-
|
| 207 |
-
if not chunks:
|
| 208 |
-
|
| 209 |
-
raise ValueError(
|
| 210 |
-
"No searchable chunks could be created from the PDF."
|
| 211 |
-
)
|
| 212 |
-
|
| 213 |
-
|
| 214 |
print(
|
| 215 |
-
f"
|
|
|
|
| 216 |
)
|
| 217 |
|
| 218 |
|
|
@@ -220,7 +199,7 @@ print(
|
|
| 220 |
# 6. LOAD EMBEDDING MODEL
|
| 221 |
# ================================================================
|
| 222 |
|
| 223 |
-
print("
|
| 224 |
|
| 225 |
|
| 226 |
embedding_model = SentenceTransformer(
|
|
@@ -229,16 +208,6 @@ embedding_model = SentenceTransformer(
|
|
| 229 |
)
|
| 230 |
|
| 231 |
|
| 232 |
-
print("β
Embedding model loaded.")
|
| 233 |
-
|
| 234 |
-
|
| 235 |
-
# ================================================================
|
| 236 |
-
# 7. CREATE DOCUMENT EMBEDDINGS
|
| 237 |
-
# ================================================================
|
| 238 |
-
|
| 239 |
-
print("\nπ§ Creating document embeddings...")
|
| 240 |
-
|
| 241 |
-
|
| 242 |
document_texts = [
|
| 243 |
chunk["text"]
|
| 244 |
for chunk in chunks
|
|
@@ -253,18 +222,15 @@ document_embeddings = embedding_model.encode(
|
|
| 253 |
)
|
| 254 |
|
| 255 |
|
| 256 |
-
print(
|
| 257 |
-
f"β
Embeddings created for "
|
| 258 |
-
f"{len(document_embeddings)} chunks."
|
| 259 |
-
)
|
| 260 |
|
| 261 |
|
| 262 |
# ================================================================
|
| 263 |
-
#
|
| 264 |
# ================================================================
|
| 265 |
|
| 266 |
print(
|
| 267 |
-
f"
|
| 268 |
f"{LLM_MODEL_ID}"
|
| 269 |
)
|
| 270 |
|
|
@@ -280,21 +246,17 @@ llm_model = AutoModelForSeq2SeqLM.from_pretrained(
|
|
| 280 |
|
| 281 |
|
| 282 |
llm_model.to(DEVICE)
|
| 283 |
-
|
| 284 |
llm_model.eval()
|
| 285 |
|
| 286 |
|
| 287 |
-
print("
|
| 288 |
|
| 289 |
|
| 290 |
# ================================================================
|
| 291 |
-
#
|
| 292 |
# ================================================================
|
| 293 |
|
| 294 |
-
def retrieve_context(question
|
| 295 |
-
"""
|
| 296 |
-
Retrieve the most relevant chunks from the PDF.
|
| 297 |
-
"""
|
| 298 |
|
| 299 |
question_embedding = embedding_model.encode(
|
| 300 |
[question],
|
|
@@ -311,7 +273,7 @@ def retrieve_context(question, top_k=TOP_K):
|
|
| 311 |
|
| 312 |
top_indices = np.argsort(
|
| 313 |
similarity_scores
|
| 314 |
-
)[-
|
| 315 |
|
| 316 |
|
| 317 |
results = []
|
|
@@ -321,11 +283,18 @@ def retrieve_context(question, top_k=TOP_K):
|
|
| 321 |
|
| 322 |
results.append(
|
| 323 |
{
|
| 324 |
-
"page":
|
| 325 |
-
|
| 326 |
-
|
| 327 |
-
|
| 328 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 329 |
}
|
| 330 |
)
|
| 331 |
|
|
@@ -334,59 +303,45 @@ def retrieve_context(question, top_k=TOP_K):
|
|
| 334 |
|
| 335 |
|
| 336 |
# ================================================================
|
| 337 |
-
#
|
| 338 |
# ================================================================
|
| 339 |
|
| 340 |
-
def build_context(
|
| 341 |
-
"""
|
| 342 |
-
Combine retrieved PDF chunks into a single context.
|
| 343 |
-
"""
|
| 344 |
-
|
| 345 |
-
context_parts = []
|
| 346 |
|
|
|
|
| 347 |
|
| 348 |
-
for item in
|
| 349 |
|
| 350 |
-
|
| 351 |
f"""
|
| 352 |
-
|
| 353 |
|
| 354 |
{item['text']}
|
| 355 |
"""
|
| 356 |
)
|
| 357 |
|
| 358 |
-
|
| 359 |
-
return "\n\n".join(context_parts)
|
| 360 |
|
| 361 |
|
| 362 |
# ================================================================
|
| 363 |
-
#
|
| 364 |
# ================================================================
|
| 365 |
|
| 366 |
def format_history(history):
|
| 367 |
-
"""
|
| 368 |
-
Convert recent chat messages into plain text for the LLM.
|
| 369 |
-
"""
|
| 370 |
|
| 371 |
if not history:
|
|
|
|
| 372 |
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
history_text = []
|
| 377 |
-
|
| 378 |
-
|
| 379 |
-
# Keep only recent messages to prevent prompt growth
|
| 380 |
-
recent_messages = history[-4:]
|
| 381 |
-
|
| 382 |
|
| 383 |
-
|
|
|
|
| 384 |
|
| 385 |
if isinstance(item, dict):
|
| 386 |
|
| 387 |
role = item.get(
|
| 388 |
"role",
|
| 389 |
-
"
|
| 390 |
)
|
| 391 |
|
| 392 |
content = item.get(
|
|
@@ -394,80 +349,62 @@ def format_history(history):
|
|
| 394 |
""
|
| 395 |
)
|
| 396 |
|
| 397 |
-
|
| 398 |
f"{role}: {content}"
|
| 399 |
)
|
| 400 |
|
| 401 |
-
|
| 402 |
-
elif isinstance(item, (list, tuple)) and len(item) == 2:
|
| 403 |
-
|
| 404 |
-
user_message = item[0]
|
| 405 |
-
assistant_message = item[1]
|
| 406 |
-
|
| 407 |
-
history_text.append(
|
| 408 |
-
f"user: {user_message}"
|
| 409 |
-
)
|
| 410 |
-
|
| 411 |
-
history_text.append(
|
| 412 |
-
f"assistant: {assistant_message}"
|
| 413 |
-
)
|
| 414 |
-
|
| 415 |
-
|
| 416 |
-
return "\n".join(history_text)
|
| 417 |
|
| 418 |
|
| 419 |
# ================================================================
|
| 420 |
-
#
|
| 421 |
# ================================================================
|
| 422 |
|
| 423 |
def generate_answer(
|
| 424 |
question,
|
| 425 |
context,
|
| 426 |
-
|
| 427 |
):
|
| 428 |
|
| 429 |
prompt = f"""
|
| 430 |
-
You are
|
| 431 |
|
| 432 |
-
|
| 433 |
-
Physicochemical Properties."
|
| 434 |
|
| 435 |
-
|
| 436 |
-
|
| 437 |
|
| 438 |
-
|
|
|
|
| 439 |
|
| 440 |
-
|
| 441 |
-
2. Do not invent facts.
|
| 442 |
-
3. Do not invent research findings.
|
| 443 |
-
4. Clearly distinguish planned methodology from completed work.
|
| 444 |
-
5. Do not claim that a model is the best-performing model unless
|
| 445 |
-
the supplied document explicitly provides that result.
|
| 446 |
-
6. Explain technical concepts in simple language when appropriate.
|
| 447 |
-
7. Keep answers clear and reasonably concise.
|
| 448 |
-
8. Use bullet points when they improve readability.
|
| 449 |
-
9. If the requested information is not available in the document,
|
| 450 |
-
say:
|
| 451 |
|
| 452 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 453 |
|
| 454 |
-
|
|
|
|
| 455 |
|
| 456 |
DOCUMENT CONTEXT:
|
| 457 |
|
| 458 |
{context}
|
| 459 |
|
| 460 |
-
|
| 461 |
RECENT CONVERSATION:
|
| 462 |
|
| 463 |
-
{
|
| 464 |
-
|
| 465 |
|
| 466 |
-
|
| 467 |
|
| 468 |
{question}
|
| 469 |
|
| 470 |
-
|
| 471 |
ANSWER:
|
| 472 |
"""
|
| 473 |
|
|
@@ -482,25 +419,25 @@ ANSWER:
|
|
| 482 |
|
| 483 |
inputs = {
|
| 484 |
key: value.to(DEVICE)
|
| 485 |
-
for key, value
|
|
|
|
| 486 |
}
|
| 487 |
|
| 488 |
|
| 489 |
with torch.no_grad():
|
| 490 |
|
| 491 |
-
|
| 492 |
**inputs,
|
| 493 |
max_new_tokens=220,
|
| 494 |
-
do_sample=False,
|
| 495 |
num_beams=2,
|
|
|
|
| 496 |
repetition_penalty=1.15,
|
| 497 |
-
no_repeat_ngram_size=3
|
| 498 |
-
early_stopping=True
|
| 499 |
)
|
| 500 |
|
| 501 |
|
| 502 |
answer = tokenizer.decode(
|
| 503 |
-
|
| 504 |
skip_special_tokens=True
|
| 505 |
)
|
| 506 |
|
|
@@ -509,113 +446,88 @@ ANSWER:
|
|
| 509 |
|
| 510 |
|
| 511 |
# ================================================================
|
| 512 |
-
#
|
| 513 |
# ================================================================
|
| 514 |
|
| 515 |
def chat(message, history):
|
| 516 |
-
"""
|
| 517 |
-
Main function called by Gradio ChatInterface.
|
| 518 |
-
"""
|
| 519 |
|
| 520 |
-
if not message
|
| 521 |
|
| 522 |
return (
|
| 523 |
-
"Please enter a question
|
| 524 |
-
"the
|
| 525 |
)
|
| 526 |
|
| 527 |
|
| 528 |
-
print("\n" + "=" *
|
| 529 |
-
|
| 530 |
print(f"QUESTION: {message}")
|
| 531 |
|
| 532 |
|
| 533 |
-
#
|
| 534 |
-
|
| 535 |
-
# ------------------------------------------------------------
|
| 536 |
-
|
| 537 |
-
retrieved_chunks = retrieve_context(
|
| 538 |
message
|
| 539 |
)
|
| 540 |
|
| 541 |
|
| 542 |
-
print("
|
| 543 |
|
| 544 |
-
|
| 545 |
-
for item in retrieved_chunks:
|
| 546 |
|
| 547 |
print(
|
| 548 |
f"Page {item['page']} | "
|
| 549 |
-
f"
|
| 550 |
)
|
| 551 |
|
| 552 |
|
| 553 |
-
# ------------------------------------------------------------
|
| 554 |
-
# Build context
|
| 555 |
-
# ------------------------------------------------------------
|
| 556 |
-
|
| 557 |
context = build_context(
|
| 558 |
-
|
| 559 |
)
|
| 560 |
|
| 561 |
|
| 562 |
-
|
| 563 |
history
|
| 564 |
)
|
| 565 |
|
| 566 |
|
| 567 |
-
# ------------------------------------------------------------
|
| 568 |
-
# Generate response
|
| 569 |
-
# ------------------------------------------------------------
|
| 570 |
-
|
| 571 |
answer = generate_answer(
|
| 572 |
message,
|
| 573 |
context,
|
| 574 |
-
|
| 575 |
)
|
| 576 |
|
| 577 |
|
| 578 |
-
#
|
| 579 |
-
# Get unique page numbers
|
| 580 |
-
# ------------------------------------------------------------
|
| 581 |
-
|
| 582 |
source_pages = sorted(
|
| 583 |
set(
|
| 584 |
item["page"]
|
| 585 |
-
for item in
|
| 586 |
)
|
| 587 |
)
|
| 588 |
|
| 589 |
|
| 590 |
-
|
| 591 |
str(page)
|
| 592 |
for page in source_pages
|
| 593 |
)
|
| 594 |
|
| 595 |
|
| 596 |
-
# ------------------------------------------------------------
|
| 597 |
-
# Return answer + source pages
|
| 598 |
-
# ------------------------------------------------------------
|
| 599 |
-
|
| 600 |
return f"""
|
| 601 |
{answer}
|
| 602 |
|
| 603 |
---
|
| 604 |
|
| 605 |
-
π **
|
| 606 |
-
|
| 607 |
-
π *Answer generated from the uploaded QM640 Capstone Synopsis.*
|
| 608 |
"""
|
| 609 |
|
| 610 |
|
| 611 |
# ================================================================
|
| 612 |
-
#
|
| 613 |
# ================================================================
|
| 614 |
|
| 615 |
css = """
|
| 616 |
|
| 617 |
.gradio-container {
|
| 618 |
-
max-width:
|
| 619 |
margin: auto !important;
|
| 620 |
}
|
| 621 |
|
|
@@ -623,7 +535,7 @@ css = """
|
|
| 623 |
|
| 624 |
|
| 625 |
# ================================================================
|
| 626 |
-
#
|
| 627 |
# ================================================================
|
| 628 |
|
| 629 |
with gr.Blocks(
|
|
@@ -632,108 +544,52 @@ with gr.Blocks(
|
|
| 632 |
|
| 633 |
|
| 634 |
# ------------------------------------------------------------
|
| 635 |
-
#
|
| 636 |
# ------------------------------------------------------------
|
| 637 |
|
| 638 |
gr.Markdown(
|
| 639 |
"""
|
| 640 |
# π· Wine Quality Research Assistant
|
| 641 |
|
| 642 |
-
###
|
| 643 |
-
|
| 644 |
-
Welcome to the **QM640 Data Analytics Capstone Research Assistant**.
|
| 645 |
-
|
| 646 |
-
This chatbot searches the uploaded capstone synopsis and retrieves
|
| 647 |
-
the most relevant sections before generating an answer.
|
| 648 |
-
|
| 649 |
-
You can ask questions about:
|
| 650 |
|
| 651 |
-
|
|
|
|
| 652 |
|
| 653 |
-
|
| 654 |
-
- Purpose of the Study
|
| 655 |
-
- Research Objectives
|
| 656 |
-
- Research Questions
|
| 657 |
-
- Research Hypotheses
|
| 658 |
-
|
| 659 |
-
### π· Dataset
|
| 660 |
-
|
| 661 |
-
- Wine Quality Dataset
|
| 662 |
-
- Physicochemical Variables
|
| 663 |
-
- Target Variable
|
| 664 |
-
- Sample Size
|
| 665 |
-
- Dataset Source
|
| 666 |
-
|
| 667 |
-
### π€ Machine Learning
|
| 668 |
-
|
| 669 |
-
- Linear Regression
|
| 670 |
-
- Decision Tree Regression
|
| 671 |
-
- Random Forest Regression
|
| 672 |
-
- Gradient Boosting Regression
|
| 673 |
-
- XGBoost Regression
|
| 674 |
-
|
| 675 |
-
### π Model Evaluation
|
| 676 |
-
|
| 677 |
-
- MAE
|
| 678 |
-
- RMSE
|
| 679 |
-
- RΒ²
|
| 680 |
-
- 80/20 Train-Test Split
|
| 681 |
-
- 5-Fold Cross Validation
|
| 682 |
-
- Hyperparameter Optimisation
|
| 683 |
-
|
| 684 |
-
### π Explainable AI
|
| 685 |
-
|
| 686 |
-
- Feature Importance
|
| 687 |
-
- SHAP
|
| 688 |
-
- LIME
|
| 689 |
-
|
| 690 |
-
### πΌ Business
|
| 691 |
-
|
| 692 |
-
- Business Impact
|
| 693 |
-
- Quality Assessment
|
| 694 |
-
- Production Optimisation
|
| 695 |
-
- Study Limitations
|
| 696 |
-
|
| 697 |
-
---
|
| 698 |
"""
|
| 699 |
)
|
| 700 |
|
| 701 |
|
| 702 |
# ------------------------------------------------------------
|
| 703 |
-
#
|
| 704 |
# ------------------------------------------------------------
|
| 705 |
|
| 706 |
gr.Markdown(
|
| 707 |
f"""
|
| 708 |
-
|
| 709 |
-
|
| 710 |
-
|
| 711 |
-
|
| 712 |
-
π **Pages loaded:** {len(pages)}
|
| 713 |
-
|
| 714 |
-
π§© **Searchable chunks:** {len(chunks)}
|
| 715 |
-
|
| 716 |
-
π **Embedding model:** `{EMBEDDING_MODEL_ID}`
|
| 717 |
-
|
| 718 |
-
π€ **Language model:** `{LLM_MODEL_ID}`
|
| 719 |
-
|
| 720 |
-
π» **Device:** `{DEVICE}`
|
| 721 |
"""
|
| 722 |
)
|
| 723 |
|
| 724 |
|
| 725 |
# ------------------------------------------------------------
|
| 726 |
-
#
|
| 727 |
# ------------------------------------------------------------
|
| 728 |
|
| 729 |
chatbot_component = gr.Chatbot(
|
| 730 |
-
height=
|
| 731 |
-
show_label=False
|
|
|
|
|
|
|
|
|
|
|
|
|
| 732 |
)
|
| 733 |
|
| 734 |
|
| 735 |
# ------------------------------------------------------------
|
| 736 |
-
#
|
| 737 |
# ------------------------------------------------------------
|
| 738 |
|
| 739 |
chat_interface = gr.ChatInterface(
|
|
@@ -742,41 +598,31 @@ You can ask questions about:
|
|
| 742 |
|
| 743 |
chatbot=chatbot_component,
|
| 744 |
|
| 745 |
-
|
| 746 |
-
|
| 747 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 748 |
|
| 749 |
-
|
| 750 |
|
| 751 |
-
|
| 752 |
|
| 753 |
-
|
| 754 |
|
| 755 |
-
"What
|
| 756 |
|
| 757 |
-
"What is
|
| 758 |
|
| 759 |
-
"
|
| 760 |
|
| 761 |
"Which machine learning models will be compared?",
|
| 762 |
|
| 763 |
-
"How will
|
| 764 |
-
|
| 765 |
-
"Explain 5-fold cross-validation in simple terms.",
|
| 766 |
-
|
| 767 |
-
"What is MAE?",
|
| 768 |
-
|
| 769 |
-
"What is RMSE?",
|
| 770 |
-
|
| 771 |
-
"What does RΒ² mean?",
|
| 772 |
-
|
| 773 |
-
"How will SHAP be used?",
|
| 774 |
-
|
| 775 |
-
"What is LIME?",
|
| 776 |
|
| 777 |
-
"What is the business value of this research?"
|
| 778 |
-
|
| 779 |
-
"What are the limitations of the study?"
|
| 780 |
|
| 781 |
]
|
| 782 |
|
|
@@ -784,35 +630,15 @@ You can ask questions about:
|
|
| 784 |
|
| 785 |
|
| 786 |
# ------------------------------------------------------------
|
| 787 |
-
#
|
| 788 |
# ------------------------------------------------------------
|
| 789 |
|
| 790 |
gr.Markdown(
|
| 791 |
"""
|
| 792 |
---
|
| 793 |
|
| 794 |
-
|
| 795 |
-
|
| 796 |
-
**Your Question**
|
| 797 |
-
|
| 798 |
-
β
|
| 799 |
-
|
| 800 |
-
**Semantic Search**
|
| 801 |
-
|
| 802 |
-
β
|
| 803 |
-
|
| 804 |
-
**Relevant Sections Retrieved from the PDF**
|
| 805 |
-
|
| 806 |
-
β
|
| 807 |
-
|
| 808 |
-
**Relevant Content Sent to FLAN-T5**
|
| 809 |
-
|
| 810 |
-
β
|
| 811 |
-
|
| 812 |
-
**AI Generated Answer**
|
| 813 |
-
|
| 814 |
-
This is a lightweight **Retrieval-Augmented Generation (RAG)**
|
| 815 |
-
implementation.
|
| 816 |
|
| 817 |
*QM640 Data Analytics Capstone*
|
| 818 |
"""
|
|
@@ -820,13 +646,11 @@ implementation.
|
|
| 820 |
|
| 821 |
|
| 822 |
# ================================================================
|
| 823 |
-
#
|
| 824 |
# ================================================================
|
| 825 |
|
| 826 |
print("\n" + "=" * 70)
|
| 827 |
-
|
| 828 |
-
print("β
Wine Quality Research Assistant is ready.")
|
| 829 |
-
|
| 830 |
print("=" * 70)
|
| 831 |
|
| 832 |
|
|
|
|
| 1 |
# ================================================================
|
| 2 |
+
# Wine Quality Research Assistant
|
| 3 |
+
# Simple RAG Chatbot for Hugging Face Spaces
|
| 4 |
+
# Gradio 6.x Compatible
|
| 5 |
# ================================================================
|
| 6 |
|
| 7 |
import os
|
|
|
|
| 22 |
APP_TITLE = "π· Wine Quality Research Assistant"
|
| 23 |
|
| 24 |
EMBEDDING_MODEL_ID = "sentence-transformers/all-MiniLM-L6-v2"
|
|
|
|
|
|
|
| 25 |
LLM_MODEL_ID = "google/flan-t5-small"
|
| 26 |
|
| 27 |
TOP_K = 3
|
|
|
|
| 32 |
|
| 33 |
|
| 34 |
print("=" * 70)
|
| 35 |
+
print("Starting Wine Quality Research Assistant")
|
| 36 |
print("=" * 70)
|
| 37 |
|
| 38 |
print(f"Device: {DEVICE}")
|
| 39 |
+
print(f"Gradio: {gr.__version__}")
|
| 40 |
|
| 41 |
|
| 42 |
# ================================================================
|
| 43 |
+
# 2. FIND PDF
|
| 44 |
# ================================================================
|
| 45 |
|
| 46 |
def find_pdf():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
pdf_files = [
|
| 49 |
file
|
| 50 |
+
for file in os.listdir(".")
|
| 51 |
if file.lower().endswith(".pdf")
|
| 52 |
]
|
| 53 |
|
| 54 |
if not pdf_files:
|
| 55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
raise FileNotFoundError(
|
| 57 |
+
"No PDF found. Upload the capstone PDF "
|
| 58 |
+
"to the root folder of the Space."
|
|
|
|
| 59 |
)
|
| 60 |
|
| 61 |
print("\nPDF files detected:")
|
|
|
|
| 68 |
|
| 69 |
PDF_FILE = find_pdf()
|
| 70 |
|
| 71 |
+
print(f"\nUsing PDF: {PDF_FILE}")
|
| 72 |
|
| 73 |
|
| 74 |
# ================================================================
|
| 75 |
+
# 3. CLEAN PDF TEXT
|
| 76 |
# ================================================================
|
| 77 |
|
| 78 |
def clean_text(text):
|
|
|
|
|
|
|
|
|
|
| 79 |
|
| 80 |
if not text:
|
| 81 |
return ""
|
| 82 |
|
| 83 |
text = text.replace("\x00", " ")
|
| 84 |
|
| 85 |
+
text = re.sub(
|
| 86 |
+
r"\s+",
|
| 87 |
+
" ",
|
| 88 |
+
text
|
| 89 |
+
)
|
| 90 |
|
| 91 |
return text.strip()
|
| 92 |
|
|
|
|
| 96 |
# ================================================================
|
| 97 |
|
| 98 |
def read_pdf(pdf_path):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 99 |
|
| 100 |
reader = PdfReader(pdf_path)
|
| 101 |
|
| 102 |
pages = []
|
| 103 |
|
| 104 |
+
for page_number, page in enumerate(
|
| 105 |
+
reader.pages,
|
| 106 |
+
start=1
|
| 107 |
+
):
|
| 108 |
|
| 109 |
try:
|
| 110 |
|
| 111 |
+
text = clean_text(
|
| 112 |
+
page.extract_text()
|
| 113 |
+
)
|
| 114 |
|
| 115 |
if text:
|
| 116 |
|
|
|
|
| 124 |
except Exception as error:
|
| 125 |
|
| 126 |
print(
|
| 127 |
+
f"Could not read page "
|
| 128 |
f"{page_number}: {error}"
|
| 129 |
)
|
| 130 |
|
|
|
|
| 137 |
if not pages:
|
| 138 |
|
| 139 |
raise ValueError(
|
| 140 |
+
"PDF found, but no readable text "
|
| 141 |
"could be extracted."
|
| 142 |
)
|
| 143 |
|
| 144 |
|
| 145 |
print(
|
| 146 |
+
f"Loaded {len(pages)} PDF pages."
|
| 147 |
)
|
| 148 |
|
| 149 |
|
| 150 |
# ================================================================
|
| 151 |
+
# 5. CREATE CHUNKS
|
| 152 |
# ================================================================
|
| 153 |
|
| 154 |
def create_chunks(pages):
|
|
|
|
|
|
|
|
|
|
| 155 |
|
| 156 |
chunks = []
|
| 157 |
|
| 158 |
for page_data in pages:
|
| 159 |
|
| 160 |
text = page_data["text"]
|
|
|
|
| 161 |
page_number = page_data["page"]
|
| 162 |
|
| 163 |
start = 0
|
|
|
|
| 166 |
|
| 167 |
end = start + CHUNK_SIZE
|
| 168 |
|
| 169 |
+
chunk_text = text[
|
| 170 |
+
start:end
|
| 171 |
+
].strip()
|
| 172 |
|
| 173 |
if len(chunk_text) > 50:
|
| 174 |
|
|
|
|
| 179 |
}
|
| 180 |
)
|
| 181 |
|
| 182 |
+
start += (
|
| 183 |
+
CHUNK_SIZE -
|
| 184 |
+
CHUNK_OVERLAP
|
| 185 |
+
)
|
| 186 |
|
| 187 |
return chunks
|
| 188 |
|
| 189 |
|
| 190 |
chunks = create_chunks(pages)
|
| 191 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 192 |
print(
|
| 193 |
+
f"Created {len(chunks)} "
|
| 194 |
+
f"knowledge chunks."
|
| 195 |
)
|
| 196 |
|
| 197 |
|
|
|
|
| 199 |
# 6. LOAD EMBEDDING MODEL
|
| 200 |
# ================================================================
|
| 201 |
|
| 202 |
+
print("Loading embedding model...")
|
| 203 |
|
| 204 |
|
| 205 |
embedding_model = SentenceTransformer(
|
|
|
|
| 208 |
)
|
| 209 |
|
| 210 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 211 |
document_texts = [
|
| 212 |
chunk["text"]
|
| 213 |
for chunk in chunks
|
|
|
|
| 222 |
)
|
| 223 |
|
| 224 |
|
| 225 |
+
print("Embedding model ready.")
|
|
|
|
|
|
|
|
|
|
| 226 |
|
| 227 |
|
| 228 |
# ================================================================
|
| 229 |
+
# 7. LOAD LANGUAGE MODEL
|
| 230 |
# ================================================================
|
| 231 |
|
| 232 |
print(
|
| 233 |
+
f"Loading language model: "
|
| 234 |
f"{LLM_MODEL_ID}"
|
| 235 |
)
|
| 236 |
|
|
|
|
| 246 |
|
| 247 |
|
| 248 |
llm_model.to(DEVICE)
|
|
|
|
| 249 |
llm_model.eval()
|
| 250 |
|
| 251 |
|
| 252 |
+
print("Language model ready.")
|
| 253 |
|
| 254 |
|
| 255 |
# ================================================================
|
| 256 |
+
# 8. RETRIEVE RELEVANT PDF CONTENT
|
| 257 |
# ================================================================
|
| 258 |
|
| 259 |
+
def retrieve_context(question):
|
|
|
|
|
|
|
|
|
|
| 260 |
|
| 261 |
question_embedding = embedding_model.encode(
|
| 262 |
[question],
|
|
|
|
| 273 |
|
| 274 |
top_indices = np.argsort(
|
| 275 |
similarity_scores
|
| 276 |
+
)[-TOP_K:][::-1]
|
| 277 |
|
| 278 |
|
| 279 |
results = []
|
|
|
|
| 283 |
|
| 284 |
results.append(
|
| 285 |
{
|
| 286 |
+
"page":
|
| 287 |
+
chunks[index]["page"],
|
| 288 |
+
|
| 289 |
+
"text":
|
| 290 |
+
chunks[index]["text"],
|
| 291 |
+
|
| 292 |
+
"score":
|
| 293 |
+
float(
|
| 294 |
+
similarity_scores[
|
| 295 |
+
index
|
| 296 |
+
]
|
| 297 |
+
)
|
| 298 |
}
|
| 299 |
)
|
| 300 |
|
|
|
|
| 303 |
|
| 304 |
|
| 305 |
# ================================================================
|
| 306 |
+
# 9. BUILD CONTEXT
|
| 307 |
# ================================================================
|
| 308 |
|
| 309 |
+
def build_context(results):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 310 |
|
| 311 |
+
sections = []
|
| 312 |
|
| 313 |
+
for item in results:
|
| 314 |
|
| 315 |
+
sections.append(
|
| 316 |
f"""
|
| 317 |
+
PAGE {item['page']}:
|
| 318 |
|
| 319 |
{item['text']}
|
| 320 |
"""
|
| 321 |
)
|
| 322 |
|
| 323 |
+
return "\n".join(sections)
|
|
|
|
| 324 |
|
| 325 |
|
| 326 |
# ================================================================
|
| 327 |
+
# 10. FORMAT CHAT HISTORY
|
| 328 |
# ================================================================
|
| 329 |
|
| 330 |
def format_history(history):
|
|
|
|
|
|
|
|
|
|
| 331 |
|
| 332 |
if not history:
|
| 333 |
+
return ""
|
| 334 |
|
| 335 |
+
conversation = []
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 336 |
|
| 337 |
+
# Only retain recent conversation
|
| 338 |
+
for item in history[-4:]:
|
| 339 |
|
| 340 |
if isinstance(item, dict):
|
| 341 |
|
| 342 |
role = item.get(
|
| 343 |
"role",
|
| 344 |
+
""
|
| 345 |
)
|
| 346 |
|
| 347 |
content = item.get(
|
|
|
|
| 349 |
""
|
| 350 |
)
|
| 351 |
|
| 352 |
+
conversation.append(
|
| 353 |
f"{role}: {content}"
|
| 354 |
)
|
| 355 |
|
| 356 |
+
return "\n".join(conversation)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 357 |
|
| 358 |
|
| 359 |
# ================================================================
|
| 360 |
+
# 11. GENERATE ANSWER
|
| 361 |
# ================================================================
|
| 362 |
|
| 363 |
def generate_answer(
|
| 364 |
question,
|
| 365 |
context,
|
| 366 |
+
history
|
| 367 |
):
|
| 368 |
|
| 369 |
prompt = f"""
|
| 370 |
+
You are a helpful academic research assistant.
|
| 371 |
|
| 372 |
+
You answer questions about the research project:
|
|
|
|
| 373 |
|
| 374 |
+
"Machine Learning-Based Prediction of Wine Quality
|
| 375 |
+
Using Physicochemical Properties."
|
| 376 |
|
| 377 |
+
Answer the user's question using ONLY the information
|
| 378 |
+
provided in the DOCUMENT CONTEXT.
|
| 379 |
|
| 380 |
+
RULES:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 381 |
|
| 382 |
+
1. Answer the question directly.
|
| 383 |
+
2. Use clear and simple language.
|
| 384 |
+
3. Be academically accurate.
|
| 385 |
+
4. Do not invent information.
|
| 386 |
+
5. Do not invent research results.
|
| 387 |
+
6. Clearly distinguish proposed methodology from
|
| 388 |
+
completed findings.
|
| 389 |
+
7. Use bullet points when helpful.
|
| 390 |
+
8. Keep the answer focused on the user's question.
|
| 391 |
+
9. If the document does not contain the answer, say:
|
| 392 |
|
| 393 |
+
"This information is not specified in the current
|
| 394 |
+
capstone synopsis."
|
| 395 |
|
| 396 |
DOCUMENT CONTEXT:
|
| 397 |
|
| 398 |
{context}
|
| 399 |
|
|
|
|
| 400 |
RECENT CONVERSATION:
|
| 401 |
|
| 402 |
+
{history}
|
|
|
|
| 403 |
|
| 404 |
+
QUESTION:
|
| 405 |
|
| 406 |
{question}
|
| 407 |
|
|
|
|
| 408 |
ANSWER:
|
| 409 |
"""
|
| 410 |
|
|
|
|
| 419 |
|
| 420 |
inputs = {
|
| 421 |
key: value.to(DEVICE)
|
| 422 |
+
for key, value
|
| 423 |
+
in inputs.items()
|
| 424 |
}
|
| 425 |
|
| 426 |
|
| 427 |
with torch.no_grad():
|
| 428 |
|
| 429 |
+
output = llm_model.generate(
|
| 430 |
**inputs,
|
| 431 |
max_new_tokens=220,
|
|
|
|
| 432 |
num_beams=2,
|
| 433 |
+
do_sample=False,
|
| 434 |
repetition_penalty=1.15,
|
| 435 |
+
no_repeat_ngram_size=3
|
|
|
|
| 436 |
)
|
| 437 |
|
| 438 |
|
| 439 |
answer = tokenizer.decode(
|
| 440 |
+
output[0],
|
| 441 |
skip_special_tokens=True
|
| 442 |
)
|
| 443 |
|
|
|
|
| 446 |
|
| 447 |
|
| 448 |
# ================================================================
|
| 449 |
+
# 12. CHAT FUNCTION
|
| 450 |
# ================================================================
|
| 451 |
|
| 452 |
def chat(message, history):
|
|
|
|
|
|
|
|
|
|
| 453 |
|
| 454 |
+
if not message.strip():
|
| 455 |
|
| 456 |
return (
|
| 457 |
+
"Please enter a question "
|
| 458 |
+
"about the document."
|
| 459 |
)
|
| 460 |
|
| 461 |
|
| 462 |
+
print("\n" + "=" * 60)
|
|
|
|
| 463 |
print(f"QUESTION: {message}")
|
| 464 |
|
| 465 |
|
| 466 |
+
# Retrieve relevant content
|
| 467 |
+
results = retrieve_context(
|
|
|
|
|
|
|
|
|
|
| 468 |
message
|
| 469 |
)
|
| 470 |
|
| 471 |
|
| 472 |
+
print("Retrieved:")
|
| 473 |
|
| 474 |
+
for item in results:
|
|
|
|
| 475 |
|
| 476 |
print(
|
| 477 |
f"Page {item['page']} | "
|
| 478 |
+
f"Score: {item['score']:.3f}"
|
| 479 |
)
|
| 480 |
|
| 481 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 482 |
context = build_context(
|
| 483 |
+
results
|
| 484 |
)
|
| 485 |
|
| 486 |
|
| 487 |
+
history_text = format_history(
|
| 488 |
history
|
| 489 |
)
|
| 490 |
|
| 491 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 492 |
answer = generate_answer(
|
| 493 |
message,
|
| 494 |
context,
|
| 495 |
+
history_text
|
| 496 |
)
|
| 497 |
|
| 498 |
|
| 499 |
+
# Unique page numbers
|
|
|
|
|
|
|
|
|
|
| 500 |
source_pages = sorted(
|
| 501 |
set(
|
| 502 |
item["page"]
|
| 503 |
+
for item in results
|
| 504 |
)
|
| 505 |
)
|
| 506 |
|
| 507 |
|
| 508 |
+
page_text = ", ".join(
|
| 509 |
str(page)
|
| 510 |
for page in source_pages
|
| 511 |
)
|
| 512 |
|
| 513 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 514 |
return f"""
|
| 515 |
{answer}
|
| 516 |
|
| 517 |
---
|
| 518 |
|
| 519 |
+
π **Source page(s): {page_text}**
|
|
|
|
|
|
|
| 520 |
"""
|
| 521 |
|
| 522 |
|
| 523 |
# ================================================================
|
| 524 |
+
# 13. CSS
|
| 525 |
# ================================================================
|
| 526 |
|
| 527 |
css = """
|
| 528 |
|
| 529 |
.gradio-container {
|
| 530 |
+
max-width: 1000px !important;
|
| 531 |
margin: auto !important;
|
| 532 |
}
|
| 533 |
|
|
|
|
| 535 |
|
| 536 |
|
| 537 |
# ================================================================
|
| 538 |
+
# 14. USER INTERFACE
|
| 539 |
# ================================================================
|
| 540 |
|
| 541 |
with gr.Blocks(
|
|
|
|
| 544 |
|
| 545 |
|
| 546 |
# ------------------------------------------------------------
|
| 547 |
+
# SIMPLE HEADER
|
| 548 |
# ------------------------------------------------------------
|
| 549 |
|
| 550 |
gr.Markdown(
|
| 551 |
"""
|
| 552 |
# π· Wine Quality Research Assistant
|
| 553 |
|
| 554 |
+
### Ask questions about the QM640 Wine Quality Capstone
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 555 |
|
| 556 |
+
This chatbot searches the uploaded capstone document and
|
| 557 |
+
answers your questions using the relevant information from it.
|
| 558 |
|
| 559 |
+
**Type your question below to begin.**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 560 |
"""
|
| 561 |
)
|
| 562 |
|
| 563 |
|
| 564 |
# ------------------------------------------------------------
|
| 565 |
+
# SMALL STATUS LINE
|
| 566 |
# ------------------------------------------------------------
|
| 567 |
|
| 568 |
gr.Markdown(
|
| 569 |
f"""
|
| 570 |
+
π **Document loaded:** `{PDF_FILE}`
|
| 571 |
+
π **Pages:** {len(pages)} |
|
| 572 |
+
π§© **Knowledge chunks:** {len(chunks)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 573 |
"""
|
| 574 |
)
|
| 575 |
|
| 576 |
|
| 577 |
# ------------------------------------------------------------
|
| 578 |
+
# CHATBOT
|
| 579 |
# ------------------------------------------------------------
|
| 580 |
|
| 581 |
chatbot_component = gr.Chatbot(
|
| 582 |
+
height=480,
|
| 583 |
+
show_label=False,
|
| 584 |
+
placeholder=(
|
| 585 |
+
"π· Ask me anything about "
|
| 586 |
+
"the Wine Quality Capstone."
|
| 587 |
+
)
|
| 588 |
)
|
| 589 |
|
| 590 |
|
| 591 |
# ------------------------------------------------------------
|
| 592 |
+
# CHAT INTERFACE
|
| 593 |
# ------------------------------------------------------------
|
| 594 |
|
| 595 |
chat_interface = gr.ChatInterface(
|
|
|
|
| 598 |
|
| 599 |
chatbot=chatbot_component,
|
| 600 |
|
| 601 |
+
textbox=gr.Textbox(
|
| 602 |
+
placeholder=(
|
| 603 |
+
"Ask a question about "
|
| 604 |
+
"the document..."
|
| 605 |
+
),
|
| 606 |
+
show_label=False
|
| 607 |
+
),
|
| 608 |
|
| 609 |
+
submit_btn="Ask",
|
| 610 |
|
| 611 |
+
stop_btn=False,
|
| 612 |
|
| 613 |
+
examples=[
|
| 614 |
|
| 615 |
+
"What is the main objective of the study?",
|
| 616 |
|
| 617 |
+
"What dataset is used in this project?",
|
| 618 |
|
| 619 |
+
"What are the research questions?",
|
| 620 |
|
| 621 |
"Which machine learning models will be compared?",
|
| 622 |
|
| 623 |
+
"How will model performance be evaluated?",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 624 |
|
| 625 |
+
"What is the business value of this research?"
|
|
|
|
|
|
|
| 626 |
|
| 627 |
]
|
| 628 |
|
|
|
|
| 630 |
|
| 631 |
|
| 632 |
# ------------------------------------------------------------
|
| 633 |
+
# FOOTER
|
| 634 |
# ------------------------------------------------------------
|
| 635 |
|
| 636 |
gr.Markdown(
|
| 637 |
"""
|
| 638 |
---
|
| 639 |
|
| 640 |
+
π **RAG-powered:** Your question β Relevant PDF sections β
|
| 641 |
+
AI-generated answer
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 642 |
|
| 643 |
*QM640 Data Analytics Capstone*
|
| 644 |
"""
|
|
|
|
| 646 |
|
| 647 |
|
| 648 |
# ================================================================
|
| 649 |
+
# 15. LAUNCH
|
| 650 |
# ================================================================
|
| 651 |
|
| 652 |
print("\n" + "=" * 70)
|
| 653 |
+
print("Wine Quality Research Assistant is ready.")
|
|
|
|
|
|
|
| 654 |
print("=" * 70)
|
| 655 |
|
| 656 |
|