Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -9,7 +9,6 @@ from typing import List, Tuple, Dict, Optional
|
|
| 9 |
|
| 10 |
import gradio as gr
|
| 11 |
|
| 12 |
-
# Optional readers
|
| 13 |
try:
|
| 14 |
from pypdf import PdfReader
|
| 15 |
except Exception:
|
|
@@ -21,9 +20,6 @@ except Exception:
|
|
| 21 |
docx = None
|
| 22 |
|
| 23 |
|
| 24 |
-
# =========================================================
|
| 25 |
-
# CONFIG
|
| 26 |
-
# =========================================================
|
| 27 |
APP_TITLE = "BrainChat"
|
| 28 |
APP_SUBTITLE = "Neurology & neuroanatomy tutor"
|
| 29 |
NOT_FOUND_TEXT = "Not found in the course material."
|
|
@@ -48,9 +44,6 @@ STOPWORDS = {
|
|
| 48 |
}
|
| 49 |
|
| 50 |
|
| 51 |
-
# =========================================================
|
| 52 |
-
# HELPERS
|
| 53 |
-
# =========================================================
|
| 54 |
def normalize_spaces(text: str) -> str:
|
| 55 |
return re.sub(r"\s+", " ", text).strip()
|
| 56 |
|
|
@@ -64,8 +57,7 @@ def split_sentences(text: str) -> List[str]:
|
|
| 64 |
text = normalize_spaces(text)
|
| 65 |
if not text:
|
| 66 |
return []
|
| 67 |
-
|
| 68 |
-
return [p.strip() for p in parts if p.strip()]
|
| 69 |
|
| 70 |
|
| 71 |
def chunk_text(text: str, chunk_size: int = 900, overlap: int = 150) -> List[str]:
|
|
@@ -170,10 +162,7 @@ def file_to_data_uri(path: Optional[Path]) -> Optional[str]:
|
|
| 170 |
data = path.read_bytes()
|
| 171 |
mime, _ = mimetypes.guess_type(str(path))
|
| 172 |
if mime is None:
|
| 173 |
-
if path.suffix.lower() == ".svg"
|
| 174 |
-
mime = "image/svg+xml"
|
| 175 |
-
else:
|
| 176 |
-
mime = "image/png"
|
| 177 |
encoded = base64.b64encode(data).decode("utf-8")
|
| 178 |
return f"data:{mime};base64,{encoded}"
|
| 179 |
except Exception:
|
|
@@ -197,9 +186,6 @@ LOGO_FILE = find_asset([
|
|
| 197 |
LOGO_URI = file_to_data_uri(LOGO_FILE)
|
| 198 |
|
| 199 |
|
| 200 |
-
# =========================================================
|
| 201 |
-
# KNOWLEDGE BASE
|
| 202 |
-
# =========================================================
|
| 203 |
class LocalKnowledgeBase:
|
| 204 |
def __init__(self) -> None:
|
| 205 |
self.docs: List[Dict] = []
|
|
@@ -218,13 +204,12 @@ class LocalKnowledgeBase:
|
|
| 218 |
self.docs.append({"doc_id": doc_id, "source": source_name, "text": text})
|
| 219 |
|
| 220 |
for idx, chunk in enumerate(chunk_text(text)):
|
| 221 |
-
tokens = set(tokenize(chunk))
|
| 222 |
self.chunks.append({
|
| 223 |
"doc_id": doc_id,
|
| 224 |
"source": source_name,
|
| 225 |
"chunk_id": idx,
|
| 226 |
"text": chunk,
|
| 227 |
-
"tokens":
|
| 228 |
})
|
| 229 |
|
| 230 |
def load_from_directories(self) -> None:
|
|
@@ -251,7 +236,7 @@ class LocalKnowledgeBase:
|
|
| 251 |
if text.strip():
|
| 252 |
self.add_document(path.name, text)
|
| 253 |
|
| 254 |
-
def search(self, query: str, top_k: int =
|
| 255 |
q_tokens = set(tokenize(query))
|
| 256 |
if not q_tokens:
|
| 257 |
return []
|
|
@@ -261,7 +246,6 @@ class LocalKnowledgeBase:
|
|
| 261 |
overlap = len(q_tokens.intersection(item["tokens"]))
|
| 262 |
if overlap == 0:
|
| 263 |
continue
|
| 264 |
-
|
| 265 |
score = overlap / max(1, len(q_tokens))
|
| 266 |
scored.append((score, item))
|
| 267 |
|
|
@@ -289,10 +273,7 @@ KB = LocalKnowledgeBase()
|
|
| 289 |
KB.load_from_directories()
|
| 290 |
|
| 291 |
|
| 292 |
-
|
| 293 |
-
# CONTENT BUILDERS
|
| 294 |
-
# =========================================================
|
| 295 |
-
def collect_context(hits: List[Dict], max_chars: int = 2000) -> Tuple[str, List[str]]:
|
| 296 |
if not hits:
|
| 297 |
return "", []
|
| 298 |
|
|
@@ -326,13 +307,7 @@ def simple_definition_style(text: str, short: bool = False) -> str:
|
|
| 326 |
sentences = split_sentences(text)
|
| 327 |
if not sentences:
|
| 328 |
return NOT_FOUND_TEXT
|
| 329 |
-
|
| 330 |
-
if short:
|
| 331 |
-
chosen = sentences[:3]
|
| 332 |
-
return "\n\n".join(chosen)
|
| 333 |
-
|
| 334 |
-
chosen = sentences[:6]
|
| 335 |
-
return "\n\n".join(chosen)
|
| 336 |
|
| 337 |
|
| 338 |
def detailed_teaching_style(query: str, text: str) -> str:
|
|
@@ -340,11 +315,11 @@ def detailed_teaching_style(query: str, text: str) -> str:
|
|
| 340 |
if not sentences:
|
| 341 |
return NOT_FOUND_TEXT
|
| 342 |
|
|
|
|
|
|
|
| 343 |
first = sentences[:2]
|
| 344 |
rest = sentences[2:6]
|
| 345 |
|
| 346 |
-
out = [f"**Topic:** {query.strip()}"]
|
| 347 |
-
|
| 348 |
if first:
|
| 349 |
out.append("\n**Simple explanation:**")
|
| 350 |
out.append(" ".join(first))
|
|
@@ -355,7 +330,7 @@ def detailed_teaching_style(query: str, text: str) -> str:
|
|
| 355 |
out.append(f"- {s}")
|
| 356 |
|
| 357 |
out.append("\n**Why this matters:**")
|
| 358 |
-
out.append("This concept is important because it helps students connect
|
| 359 |
|
| 360 |
return "\n".join(out)
|
| 361 |
|
|
@@ -379,7 +354,7 @@ def build_flashcards(query: str, text: str, count: int = 5) -> str:
|
|
| 379 |
|
| 380 |
out = ["**Flash Cards**"]
|
| 381 |
for i, term in enumerate(terms, 1):
|
| 382 |
-
meaning = sentences[min(i - 1, len(sentences) - 1)] if sentences else f"{term} is an important concept
|
| 383 |
out.append(f"\n**Card {i}**")
|
| 384 |
out.append(f"- **Term:** {term}")
|
| 385 |
out.append(f"- **Meaning:** {meaning}")
|
|
@@ -394,19 +369,17 @@ def build_case_study(query: str, text: str) -> str:
|
|
| 394 |
|
| 395 |
return (
|
| 396 |
"**Case Study**\n\n"
|
| 397 |
-
f"A student is
|
| 398 |
-
f"While studying, the student learns that: {base}\n\n"
|
| 399 |
"**Scenario:**\n"
|
| 400 |
-
"A patient
|
| 401 |
"The student must explain what structure or concept is involved, what it normally does, "
|
| 402 |
"and what may happen when it is affected.\n\n"
|
| 403 |
"**Questions to think about:**\n"
|
| 404 |
"1. What is the main structure or idea here?\n"
|
| 405 |
"2. What is its normal function?\n"
|
| 406 |
-
"3. What symptoms may appear if it is damaged
|
| 407 |
"4. Why is this topic important in neuroanatomy or neurology?\n\n"
|
| 408 |
-
"**
|
| 409 |
-
"This case study is for understanding and classroom learning, not for real medical diagnosis."
|
| 410 |
)
|
| 411 |
|
| 412 |
|
|
@@ -429,7 +402,7 @@ def build_quiz(query: str, text: str, n_questions: int = 5) -> str:
|
|
| 429 |
out = [f"**Quiz: {query.strip()}**"]
|
| 430 |
for i, w in enumerate(unique_words[:n_questions], 1):
|
| 431 |
out.append(f"{i}. What is **{w}**?")
|
| 432 |
-
out.append(f"{i}
|
| 433 |
return "\n".join(out)
|
| 434 |
|
| 435 |
|
|
@@ -448,26 +421,19 @@ def build_revision_questions(query: str, text: str, n_questions: int = 5) -> str
|
|
| 448 |
out.append(f"{i}. Give one example related to {query.strip()}.")
|
| 449 |
else:
|
| 450 |
if sentences:
|
| 451 |
-
out.append(f
|
| 452 |
else:
|
| 453 |
out.append(f"{i}. Write short notes on {query.strip()}.")
|
| 454 |
|
| 455 |
return "\n".join(out)
|
| 456 |
|
| 457 |
|
| 458 |
-
def make_learning_output(
|
| 459 |
-
query: str,
|
| 460 |
-
learning_mode: str,
|
| 461 |
-
tutor_mode: str,
|
| 462 |
-
quiz_questions: str,
|
| 463 |
-
hits: List[Dict]
|
| 464 |
-
) -> Tuple[str, List[str]]:
|
| 465 |
if not hits:
|
| 466 |
return NOT_FOUND_TEXT, []
|
| 467 |
|
| 468 |
context, sources = collect_context(hits, max_chars=2400)
|
| 469 |
mode = (learning_mode or "Detailed Explanation").strip().lower()
|
| 470 |
-
|
| 471 |
qn = 5 if str(quiz_questions).lower() == "auto" else int(quiz_questions)
|
| 472 |
|
| 473 |
if mode == "normal answer":
|
|
@@ -504,97 +470,86 @@ def format_answer(answer_text: str, sources: List[str], show_sources: bool) -> s
|
|
| 504 |
return answer_text
|
| 505 |
|
| 506 |
|
| 507 |
-
def get_answer_and_sources(
|
| 508 |
-
message: str,
|
| 509 |
-
learning_mode: str,
|
| 510 |
-
tutor_mode: str,
|
| 511 |
-
answer_language: str,
|
| 512 |
-
quiz_questions: str
|
| 513 |
-
) -> Tuple[str, List[str]]:
|
| 514 |
msg = (message or "").strip()
|
| 515 |
if not msg:
|
| 516 |
return "Please type a question.", []
|
| 517 |
|
| 518 |
lower_msg = msg.lower().strip()
|
| 519 |
-
|
| 520 |
if lower_msg in {"hi", "hello", "hey"}:
|
| 521 |
return (
|
| 522 |
"Hello. Ask me anything from your neurology or neuroanatomy material. "
|
| 523 |
-
"You can
|
| 524 |
[]
|
| 525 |
)
|
| 526 |
|
| 527 |
hits = KB.search(msg, top_k=6)
|
| 528 |
-
return make_learning_output(msg, learning_mode,
|
| 529 |
|
| 530 |
|
| 531 |
-
# =========================================================
|
| 532 |
-
# UI
|
| 533 |
-
# =========================================================
|
| 534 |
CUSTOM_CSS = """
|
| 535 |
:root{
|
| 536 |
-
--
|
| 537 |
-
--
|
| 538 |
-
--
|
| 539 |
-
--
|
| 540 |
-
--
|
| 541 |
-
--
|
| 542 |
-
--
|
| 543 |
-
--
|
| 544 |
-
--
|
| 545 |
-
--
|
| 546 |
-
--
|
|
|
|
| 547 |
}
|
| 548 |
|
| 549 |
html, body, .gradio-container {
|
| 550 |
-
background: linear-gradient(180deg, #
|
| 551 |
-
color: var(--text) !important;
|
| 552 |
font-family: "Segoe UI", Arial, sans-serif !important;
|
| 553 |
}
|
| 554 |
|
| 555 |
#main_shell {
|
| 556 |
-
max-width:
|
| 557 |
margin: 18px auto;
|
| 558 |
-
padding: 0
|
| 559 |
}
|
| 560 |
|
| 561 |
#topbar {
|
| 562 |
-
background: linear-gradient(90deg, #
|
| 563 |
-
border-radius:
|
| 564 |
-
padding:
|
| 565 |
-
box-shadow: 0
|
| 566 |
-
|
| 567 |
-
margin-bottom: 18px;
|
| 568 |
}
|
| 569 |
|
| 570 |
#brand_row {
|
| 571 |
display: flex;
|
| 572 |
align-items: center;
|
| 573 |
-
gap:
|
| 574 |
}
|
| 575 |
|
| 576 |
#brand_logo {
|
| 577 |
-
width:
|
| 578 |
-
height:
|
| 579 |
-
border-radius:
|
| 580 |
object-fit: contain;
|
| 581 |
-
background:
|
| 582 |
-
padding:
|
| 583 |
-
box-shadow: 0
|
| 584 |
}
|
| 585 |
|
| 586 |
#brand_fallback {
|
| 587 |
-
width:
|
| 588 |
-
height:
|
| 589 |
-
border-radius:
|
| 590 |
display: flex;
|
| 591 |
align-items: center;
|
| 592 |
justify-content: center;
|
| 593 |
-
background:
|
| 594 |
-
color: #
|
| 595 |
-
font-size:
|
| 596 |
font-weight: 800;
|
| 597 |
-
box-shadow: 0 6px 18px rgba(0,0,0,0.12);
|
| 598 |
}
|
| 599 |
|
| 600 |
#brand_title {
|
|
@@ -607,41 +562,36 @@ html, body, .gradio-container {
|
|
| 607 |
|
| 608 |
#brand_subtitle {
|
| 609 |
font-size: 15px;
|
| 610 |
-
color: #
|
| 611 |
font-weight: 600;
|
| 612 |
margin-top: 3px;
|
| 613 |
}
|
| 614 |
|
| 615 |
#settings_card, #chat_card {
|
| 616 |
-
background: rgba(
|
| 617 |
-
border:
|
| 618 |
-
border-radius: 24px !important;
|
| 619 |
-
box-shadow: 0 10px 24px rgba(120, 100, 180, 0.08);
|
| 620 |
-
}
|
| 621 |
-
|
| 622 |
-
#chatbot {
|
| 623 |
-
background: linear-gradient(180deg, #fffefb 0%, #fff7ff 100%) !important;
|
| 624 |
border-radius: 20px !important;
|
|
|
|
| 625 |
}
|
| 626 |
|
| 627 |
-
#chatbot
|
| 628 |
-
background:
|
| 629 |
-
|
| 630 |
-
|
| 631 |
-
|
| 632 |
-
|
| 633 |
-
|
| 634 |
-
|
| 635 |
-
|
| 636 |
-
|
| 637 |
border-radius: 18px !important;
|
| 638 |
}
|
| 639 |
|
| 640 |
textarea, input, .wrap textarea {
|
| 641 |
border-radius: 18px !important;
|
| 642 |
-
border:
|
| 643 |
-
background:
|
| 644 |
-
color: var(--text) !important;
|
| 645 |
}
|
| 646 |
|
| 647 |
button {
|
|
@@ -652,37 +602,27 @@ button {
|
|
| 652 |
}
|
| 653 |
|
| 654 |
#send_btn {
|
| 655 |
-
background: linear-gradient(90deg, #
|
| 656 |
color: white !important;
|
| 657 |
}
|
| 658 |
|
| 659 |
#clear_btn {
|
| 660 |
-
background:
|
| 661 |
-
color: #
|
| 662 |
-
|
| 663 |
-
|
| 664 |
-
#upload_btn {
|
| 665 |
-
background: linear-gradient(90deg, #67e8f9, #6ee7b7) !important;
|
| 666 |
-
color: #17324d !important;
|
| 667 |
}
|
| 668 |
|
| 669 |
#reload_btn {
|
| 670 |
-
background:
|
| 671 |
-
color: #
|
| 672 |
-
border:
|
| 673 |
}
|
| 674 |
|
| 675 |
.small_hint {
|
| 676 |
-
color: var(--muted);
|
| 677 |
font-size: 13px;
|
| 678 |
}
|
| 679 |
|
| 680 |
-
.gr-file, .gr-file * {
|
| 681 |
-
background: #fffaf0 !important;
|
| 682 |
-
color: #2f2a4d !important;
|
| 683 |
-
border-color: #f3df8c !important;
|
| 684 |
-
}
|
| 685 |
-
|
| 686 |
footer {
|
| 687 |
display: none !important;
|
| 688 |
}
|
|
@@ -708,19 +648,20 @@ def build_header_html() -> str:
|
|
| 708 |
"""
|
| 709 |
|
| 710 |
|
| 711 |
-
def respond(message, history, learning_mode,
|
| 712 |
history = history or []
|
| 713 |
|
| 714 |
answer_text, sources = get_answer_and_sources(
|
| 715 |
message=message,
|
| 716 |
learning_mode=learning_mode,
|
| 717 |
-
tutor_mode=tutor_mode,
|
| 718 |
-
answer_language=answer_language,
|
| 719 |
quiz_questions=quiz_questions
|
| 720 |
)
|
| 721 |
|
| 722 |
final_text = format_answer(answer_text, sources, show_sources)
|
| 723 |
-
|
|
|
|
|
|
|
|
|
|
| 724 |
return history, ""
|
| 725 |
|
| 726 |
|
|
@@ -735,27 +676,7 @@ def reload_materials():
|
|
| 735 |
return f"Materials reloaded. Loaded {len(KB.docs)} document(s)."
|
| 736 |
|
| 737 |
|
| 738 |
-
|
| 739 |
-
if not files:
|
| 740 |
-
return "No file uploaded."
|
| 741 |
-
|
| 742 |
-
added = 0
|
| 743 |
-
for f in files:
|
| 744 |
-
try:
|
| 745 |
-
path = Path(f.name)
|
| 746 |
-
text = extract_text_from_file(path)
|
| 747 |
-
if text.strip():
|
| 748 |
-
KB.add_document(path.name, text)
|
| 749 |
-
added += 1
|
| 750 |
-
except Exception:
|
| 751 |
-
continue
|
| 752 |
-
|
| 753 |
-
if added == 0:
|
| 754 |
-
return "No readable text found in uploaded files."
|
| 755 |
-
return f"{added} file(s) added successfully."
|
| 756 |
-
|
| 757 |
-
|
| 758 |
-
with gr.Blocks(css=CUSTOM_CSS, title=APP_TITLE) as demo:
|
| 759 |
with gr.Column(elem_id="main_shell"):
|
| 760 |
gr.HTML(build_header_html())
|
| 761 |
|
|
@@ -774,18 +695,6 @@ with gr.Blocks(css=CUSTOM_CSS, title=APP_TITLE) as demo:
|
|
| 774 |
label="Learning Mode"
|
| 775 |
)
|
| 776 |
|
| 777 |
-
tutor_mode = gr.Dropdown(
|
| 778 |
-
["Brief", "Detailed"],
|
| 779 |
-
value="Detailed",
|
| 780 |
-
label="Tutor Mode"
|
| 781 |
-
)
|
| 782 |
-
|
| 783 |
-
answer_language = gr.Dropdown(
|
| 784 |
-
["Auto", "English", "Spanish"],
|
| 785 |
-
value="Auto",
|
| 786 |
-
label="Answer Language"
|
| 787 |
-
)
|
| 788 |
-
|
| 789 |
quiz_questions = gr.Dropdown(
|
| 790 |
["Auto", "5", "10"],
|
| 791 |
value="Auto",
|
|
@@ -804,26 +713,18 @@ with gr.Blocks(css=CUSTOM_CSS, title=APP_TITLE) as demo:
|
|
| 804 |
|
| 805 |
with gr.Column(elem_id="chat_card"):
|
| 806 |
chatbot = gr.Chatbot(
|
| 807 |
-
height=
|
| 808 |
elem_id="chatbot",
|
| 809 |
-
show_label=False
|
|
|
|
| 810 |
)
|
| 811 |
|
| 812 |
-
|
| 813 |
-
|
| 814 |
-
|
| 815 |
-
|
| 816 |
-
label="Add material files (optional)",
|
| 817 |
-
scale=2
|
| 818 |
-
)
|
| 819 |
-
msg = gr.Textbox(
|
| 820 |
-
placeholder="Ask a question about neurology or neuroanatomy...",
|
| 821 |
-
show_label=False,
|
| 822 |
-
scale=6
|
| 823 |
-
)
|
| 824 |
|
| 825 |
with gr.Row():
|
| 826 |
-
upload_btn = gr.Button("Upload Files", elem_id="upload_btn")
|
| 827 |
reload_btn = gr.Button("Reload Materials", elem_id="reload_btn")
|
| 828 |
clear_btn = gr.Button("Clear Chat", elem_id="clear_btn")
|
| 829 |
send_btn = gr.Button("Send", elem_id="send_btn")
|
|
@@ -832,13 +733,13 @@ with gr.Blocks(css=CUSTOM_CSS, title=APP_TITLE) as demo:
|
|
| 832 |
|
| 833 |
send_btn.click(
|
| 834 |
respond,
|
| 835 |
-
inputs=[msg, chatbot, learning_mode,
|
| 836 |
outputs=[chatbot, msg]
|
| 837 |
)
|
| 838 |
|
| 839 |
msg.submit(
|
| 840 |
respond,
|
| 841 |
-
inputs=[msg, chatbot, learning_mode,
|
| 842 |
outputs=[chatbot, msg]
|
| 843 |
)
|
| 844 |
|
|
@@ -852,11 +753,5 @@ with gr.Blocks(css=CUSTOM_CSS, title=APP_TITLE) as demo:
|
|
| 852 |
outputs=[status_box]
|
| 853 |
)
|
| 854 |
|
| 855 |
-
upload_btn.click(
|
| 856 |
-
upload_files,
|
| 857 |
-
inputs=[file_input],
|
| 858 |
-
outputs=[status_box]
|
| 859 |
-
)
|
| 860 |
-
|
| 861 |
if __name__ == "__main__":
|
| 862 |
-
demo.launch()
|
|
|
|
| 9 |
|
| 10 |
import gradio as gr
|
| 11 |
|
|
|
|
| 12 |
try:
|
| 13 |
from pypdf import PdfReader
|
| 14 |
except Exception:
|
|
|
|
| 20 |
docx = None
|
| 21 |
|
| 22 |
|
|
|
|
|
|
|
|
|
|
| 23 |
APP_TITLE = "BrainChat"
|
| 24 |
APP_SUBTITLE = "Neurology & neuroanatomy tutor"
|
| 25 |
NOT_FOUND_TEXT = "Not found in the course material."
|
|
|
|
| 44 |
}
|
| 45 |
|
| 46 |
|
|
|
|
|
|
|
|
|
|
| 47 |
def normalize_spaces(text: str) -> str:
|
| 48 |
return re.sub(r"\s+", " ", text).strip()
|
| 49 |
|
|
|
|
| 57 |
text = normalize_spaces(text)
|
| 58 |
if not text:
|
| 59 |
return []
|
| 60 |
+
return [p.strip() for p in re.split(r"(?<=[.!?])\s+", text) if p.strip()]
|
|
|
|
| 61 |
|
| 62 |
|
| 63 |
def chunk_text(text: str, chunk_size: int = 900, overlap: int = 150) -> List[str]:
|
|
|
|
| 162 |
data = path.read_bytes()
|
| 163 |
mime, _ = mimetypes.guess_type(str(path))
|
| 164 |
if mime is None:
|
| 165 |
+
mime = "image/svg+xml" if path.suffix.lower() == ".svg" else "image/png"
|
|
|
|
|
|
|
|
|
|
| 166 |
encoded = base64.b64encode(data).decode("utf-8")
|
| 167 |
return f"data:{mime};base64,{encoded}"
|
| 168 |
except Exception:
|
|
|
|
| 186 |
LOGO_URI = file_to_data_uri(LOGO_FILE)
|
| 187 |
|
| 188 |
|
|
|
|
|
|
|
|
|
|
| 189 |
class LocalKnowledgeBase:
|
| 190 |
def __init__(self) -> None:
|
| 191 |
self.docs: List[Dict] = []
|
|
|
|
| 204 |
self.docs.append({"doc_id": doc_id, "source": source_name, "text": text})
|
| 205 |
|
| 206 |
for idx, chunk in enumerate(chunk_text(text)):
|
|
|
|
| 207 |
self.chunks.append({
|
| 208 |
"doc_id": doc_id,
|
| 209 |
"source": source_name,
|
| 210 |
"chunk_id": idx,
|
| 211 |
"text": chunk,
|
| 212 |
+
"tokens": set(tokenize(chunk))
|
| 213 |
})
|
| 214 |
|
| 215 |
def load_from_directories(self) -> None:
|
|
|
|
| 236 |
if text.strip():
|
| 237 |
self.add_document(path.name, text)
|
| 238 |
|
| 239 |
+
def search(self, query: str, top_k: int = 6) -> List[Dict]:
|
| 240 |
q_tokens = set(tokenize(query))
|
| 241 |
if not q_tokens:
|
| 242 |
return []
|
|
|
|
| 246 |
overlap = len(q_tokens.intersection(item["tokens"]))
|
| 247 |
if overlap == 0:
|
| 248 |
continue
|
|
|
|
| 249 |
score = overlap / max(1, len(q_tokens))
|
| 250 |
scored.append((score, item))
|
| 251 |
|
|
|
|
| 273 |
KB.load_from_directories()
|
| 274 |
|
| 275 |
|
| 276 |
+
def collect_context(hits: List[Dict], max_chars: int = 2400) -> Tuple[str, List[str]]:
|
|
|
|
|
|
|
|
|
|
| 277 |
if not hits:
|
| 278 |
return "", []
|
| 279 |
|
|
|
|
| 307 |
sentences = split_sentences(text)
|
| 308 |
if not sentences:
|
| 309 |
return NOT_FOUND_TEXT
|
| 310 |
+
return "\n\n".join(sentences[:3] if short else sentences[:6])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 311 |
|
| 312 |
|
| 313 |
def detailed_teaching_style(query: str, text: str) -> str:
|
|
|
|
| 315 |
if not sentences:
|
| 316 |
return NOT_FOUND_TEXT
|
| 317 |
|
| 318 |
+
out = [f"**Topic:** {query.strip()}"]
|
| 319 |
+
|
| 320 |
first = sentences[:2]
|
| 321 |
rest = sentences[2:6]
|
| 322 |
|
|
|
|
|
|
|
| 323 |
if first:
|
| 324 |
out.append("\n**Simple explanation:**")
|
| 325 |
out.append(" ".join(first))
|
|
|
|
| 330 |
out.append(f"- {s}")
|
| 331 |
|
| 332 |
out.append("\n**Why this matters:**")
|
| 333 |
+
out.append("This concept is important because it helps students connect structure, function, and clinical understanding.")
|
| 334 |
|
| 335 |
return "\n".join(out)
|
| 336 |
|
|
|
|
| 354 |
|
| 355 |
out = ["**Flash Cards**"]
|
| 356 |
for i, term in enumerate(terms, 1):
|
| 357 |
+
meaning = sentences[min(i - 1, len(sentences) - 1)] if sentences else f"{term} is an important concept."
|
| 358 |
out.append(f"\n**Card {i}**")
|
| 359 |
out.append(f"- **Term:** {term}")
|
| 360 |
out.append(f"- **Meaning:** {meaning}")
|
|
|
|
| 369 |
|
| 370 |
return (
|
| 371 |
"**Case Study**\n\n"
|
| 372 |
+
f"A student is studying **{query}**. From the material, the student learns that: {base}\n\n"
|
|
|
|
| 373 |
"**Scenario:**\n"
|
| 374 |
+
"A patient shows symptoms that may involve this part of the nervous system. "
|
| 375 |
"The student must explain what structure or concept is involved, what it normally does, "
|
| 376 |
"and what may happen when it is affected.\n\n"
|
| 377 |
"**Questions to think about:**\n"
|
| 378 |
"1. What is the main structure or idea here?\n"
|
| 379 |
"2. What is its normal function?\n"
|
| 380 |
+
"3. What symptoms may appear if it is damaged?\n"
|
| 381 |
"4. Why is this topic important in neuroanatomy or neurology?\n\n"
|
| 382 |
+
"**Note:** This is for classroom learning, not real diagnosis."
|
|
|
|
| 383 |
)
|
| 384 |
|
| 385 |
|
|
|
|
| 402 |
out = [f"**Quiz: {query.strip()}**"]
|
| 403 |
for i, w in enumerate(unique_words[:n_questions], 1):
|
| 404 |
out.append(f"{i}. What is **{w}**?")
|
| 405 |
+
out.append(f"{i}a. Why is **{w}** important?")
|
| 406 |
return "\n".join(out)
|
| 407 |
|
| 408 |
|
|
|
|
| 421 |
out.append(f"{i}. Give one example related to {query.strip()}.")
|
| 422 |
else:
|
| 423 |
if sentences:
|
| 424 |
+
out.append(f'{i}. Explain this statement: "{sentences[min(i-5, len(sentences)-1)]}"')
|
| 425 |
else:
|
| 426 |
out.append(f"{i}. Write short notes on {query.strip()}.")
|
| 427 |
|
| 428 |
return "\n".join(out)
|
| 429 |
|
| 430 |
|
| 431 |
+
def make_learning_output(query: str, learning_mode: str, quiz_questions: str, hits: List[Dict]) -> Tuple[str, List[str]]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 432 |
if not hits:
|
| 433 |
return NOT_FOUND_TEXT, []
|
| 434 |
|
| 435 |
context, sources = collect_context(hits, max_chars=2400)
|
| 436 |
mode = (learning_mode or "Detailed Explanation").strip().lower()
|
|
|
|
| 437 |
qn = 5 if str(quiz_questions).lower() == "auto" else int(quiz_questions)
|
| 438 |
|
| 439 |
if mode == "normal answer":
|
|
|
|
| 470 |
return answer_text
|
| 471 |
|
| 472 |
|
| 473 |
+
def get_answer_and_sources(message: str, learning_mode: str, quiz_questions: str) -> Tuple[str, List[str]]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 474 |
msg = (message or "").strip()
|
| 475 |
if not msg:
|
| 476 |
return "Please type a question.", []
|
| 477 |
|
| 478 |
lower_msg = msg.lower().strip()
|
|
|
|
| 479 |
if lower_msg in {"hi", "hello", "hey"}:
|
| 480 |
return (
|
| 481 |
"Hello. Ask me anything from your neurology or neuroanatomy material. "
|
| 482 |
+
"You can choose flash cards, case study, quiz, short explanation, or detailed explanation.",
|
| 483 |
[]
|
| 484 |
)
|
| 485 |
|
| 486 |
hits = KB.search(msg, top_k=6)
|
| 487 |
+
return make_learning_output(msg, learning_mode, quiz_questions, hits)
|
| 488 |
|
| 489 |
|
|
|
|
|
|
|
|
|
|
| 490 |
CUSTOM_CSS = """
|
| 491 |
:root{
|
| 492 |
+
--wa-bg: #efeae2;
|
| 493 |
+
--wa-panel: #f7f5f2;
|
| 494 |
+
--wa-header: #075e54;
|
| 495 |
+
--wa-header-2: #128c7e;
|
| 496 |
+
--wa-text: #1f2c34;
|
| 497 |
+
--wa-muted: #667781;
|
| 498 |
+
--wa-border: #d8e0dc;
|
| 499 |
+
--wa-user: #d9fdd3;
|
| 500 |
+
--wa-bot: #ffffff;
|
| 501 |
+
--wa-btn: #25d366;
|
| 502 |
+
--wa-btn-dark: #128c7e;
|
| 503 |
+
--wa-input: #ffffff;
|
| 504 |
}
|
| 505 |
|
| 506 |
html, body, .gradio-container {
|
| 507 |
+
background: linear-gradient(180deg, #e6ddd4 0%, #efeae2 100%) !important;
|
| 508 |
+
color: var(--wa-text) !important;
|
| 509 |
font-family: "Segoe UI", Arial, sans-serif !important;
|
| 510 |
}
|
| 511 |
|
| 512 |
#main_shell {
|
| 513 |
+
max-width: 1100px;
|
| 514 |
margin: 18px auto;
|
| 515 |
+
padding: 0 10px 18px 10px;
|
| 516 |
}
|
| 517 |
|
| 518 |
#topbar {
|
| 519 |
+
background: linear-gradient(90deg, #075e54 0%, #128c7e 100%);
|
| 520 |
+
border-radius: 22px;
|
| 521 |
+
padding: 16px 18px;
|
| 522 |
+
box-shadow: 0 10px 22px rgba(7, 94, 84, 0.20);
|
| 523 |
+
margin-bottom: 16px;
|
|
|
|
| 524 |
}
|
| 525 |
|
| 526 |
#brand_row {
|
| 527 |
display: flex;
|
| 528 |
align-items: center;
|
| 529 |
+
gap: 14px;
|
| 530 |
}
|
| 531 |
|
| 532 |
#brand_logo {
|
| 533 |
+
width: 68px;
|
| 534 |
+
height: 68px;
|
| 535 |
+
border-radius: 16px;
|
| 536 |
object-fit: contain;
|
| 537 |
+
background: white;
|
| 538 |
+
padding: 5px;
|
| 539 |
+
box-shadow: 0 4px 10px rgba(0,0,0,0.12);
|
| 540 |
}
|
| 541 |
|
| 542 |
#brand_fallback {
|
| 543 |
+
width: 68px;
|
| 544 |
+
height: 68px;
|
| 545 |
+
border-radius: 16px;
|
| 546 |
display: flex;
|
| 547 |
align-items: center;
|
| 548 |
justify-content: center;
|
| 549 |
+
background: white;
|
| 550 |
+
color: #075e54;
|
| 551 |
+
font-size: 26px;
|
| 552 |
font-weight: 800;
|
|
|
|
| 553 |
}
|
| 554 |
|
| 555 |
#brand_title {
|
|
|
|
| 562 |
|
| 563 |
#brand_subtitle {
|
| 564 |
font-size: 15px;
|
| 565 |
+
color: #dcf8c6;
|
| 566 |
font-weight: 600;
|
| 567 |
margin-top: 3px;
|
| 568 |
}
|
| 569 |
|
| 570 |
#settings_card, #chat_card {
|
| 571 |
+
background: rgba(247,245,242,0.96) !important;
|
| 572 |
+
border: 1px solid var(--wa-border) !important;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 573 |
border-radius: 20px !important;
|
| 574 |
+
box-shadow: 0 8px 20px rgba(80, 90, 90, 0.08);
|
| 575 |
}
|
| 576 |
|
| 577 |
+
#chatbot {
|
| 578 |
+
background:
|
| 579 |
+
linear-gradient(rgba(239,234,226,0.94), rgba(239,234,226,0.94)),
|
| 580 |
+
repeating-linear-gradient(
|
| 581 |
+
45deg,
|
| 582 |
+
rgba(255,255,255,0.08) 0px,
|
| 583 |
+
rgba(255,255,255,0.08) 12px,
|
| 584 |
+
rgba(0,0,0,0.02) 12px,
|
| 585 |
+
rgba(0,0,0,0.02) 24px
|
| 586 |
+
) !important;
|
| 587 |
border-radius: 18px !important;
|
| 588 |
}
|
| 589 |
|
| 590 |
textarea, input, .wrap textarea {
|
| 591 |
border-radius: 18px !important;
|
| 592 |
+
border: 1px solid #d5ddd9 !important;
|
| 593 |
+
background: var(--wa-input) !important;
|
| 594 |
+
color: var(--wa-text) !important;
|
| 595 |
}
|
| 596 |
|
| 597 |
button {
|
|
|
|
| 602 |
}
|
| 603 |
|
| 604 |
#send_btn {
|
| 605 |
+
background: linear-gradient(90deg, #25d366, #128c7e) !important;
|
| 606 |
color: white !important;
|
| 607 |
}
|
| 608 |
|
| 609 |
#clear_btn {
|
| 610 |
+
background: #ffffff !important;
|
| 611 |
+
color: #075e54 !important;
|
| 612 |
+
border: 1px solid #bfd6cf !important;
|
|
|
|
|
|
|
|
|
|
|
|
|
| 613 |
}
|
| 614 |
|
| 615 |
#reload_btn {
|
| 616 |
+
background: #e7f7ee !important;
|
| 617 |
+
color: #075e54 !important;
|
| 618 |
+
border: 1px solid #bfe6d0 !important;
|
| 619 |
}
|
| 620 |
|
| 621 |
.small_hint {
|
| 622 |
+
color: var(--wa-muted);
|
| 623 |
font-size: 13px;
|
| 624 |
}
|
| 625 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 626 |
footer {
|
| 627 |
display: none !important;
|
| 628 |
}
|
|
|
|
| 648 |
"""
|
| 649 |
|
| 650 |
|
| 651 |
+
def respond(message, history, learning_mode, quiz_questions, show_sources):
|
| 652 |
history = history or []
|
| 653 |
|
| 654 |
answer_text, sources = get_answer_and_sources(
|
| 655 |
message=message,
|
| 656 |
learning_mode=learning_mode,
|
|
|
|
|
|
|
| 657 |
quiz_questions=quiz_questions
|
| 658 |
)
|
| 659 |
|
| 660 |
final_text = format_answer(answer_text, sources, show_sources)
|
| 661 |
+
|
| 662 |
+
history.append({"role": "user", "content": message})
|
| 663 |
+
history.append({"role": "assistant", "content": final_text})
|
| 664 |
+
|
| 665 |
return history, ""
|
| 666 |
|
| 667 |
|
|
|
|
| 676 |
return f"Materials reloaded. Loaded {len(KB.docs)} document(s)."
|
| 677 |
|
| 678 |
|
| 679 |
+
with gr.Blocks(title=APP_TITLE) as demo:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 680 |
with gr.Column(elem_id="main_shell"):
|
| 681 |
gr.HTML(build_header_html())
|
| 682 |
|
|
|
|
| 695 |
label="Learning Mode"
|
| 696 |
)
|
| 697 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 698 |
quiz_questions = gr.Dropdown(
|
| 699 |
["Auto", "5", "10"],
|
| 700 |
value="Auto",
|
|
|
|
| 713 |
|
| 714 |
with gr.Column(elem_id="chat_card"):
|
| 715 |
chatbot = gr.Chatbot(
|
| 716 |
+
height=500,
|
| 717 |
elem_id="chatbot",
|
| 718 |
+
show_label=False,
|
| 719 |
+
type="messages"
|
| 720 |
)
|
| 721 |
|
| 722 |
+
msg = gr.Textbox(
|
| 723 |
+
placeholder="Type your question here...",
|
| 724 |
+
show_label=False
|
| 725 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 726 |
|
| 727 |
with gr.Row():
|
|
|
|
| 728 |
reload_btn = gr.Button("Reload Materials", elem_id="reload_btn")
|
| 729 |
clear_btn = gr.Button("Clear Chat", elem_id="clear_btn")
|
| 730 |
send_btn = gr.Button("Send", elem_id="send_btn")
|
|
|
|
| 733 |
|
| 734 |
send_btn.click(
|
| 735 |
respond,
|
| 736 |
+
inputs=[msg, chatbot, learning_mode, quiz_questions, show_sources],
|
| 737 |
outputs=[chatbot, msg]
|
| 738 |
)
|
| 739 |
|
| 740 |
msg.submit(
|
| 741 |
respond,
|
| 742 |
+
inputs=[msg, chatbot, learning_mode, quiz_questions, show_sources],
|
| 743 |
outputs=[chatbot, msg]
|
| 744 |
)
|
| 745 |
|
|
|
|
| 753 |
outputs=[status_box]
|
| 754 |
)
|
| 755 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 756 |
if __name__ == "__main__":
|
| 757 |
+
demo.launch(css=CUSTOM_CSS)
|