Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,6 +1,7 @@
|
|
| 1 |
import os
|
| 2 |
import glob
|
| 3 |
import yaml
|
|
|
|
| 4 |
from typing import List, Tuple
|
| 5 |
|
| 6 |
import faiss
|
|
@@ -37,7 +38,7 @@ def get_default_config():
|
|
| 37 |
"index_directory": "./index",
|
| 38 |
},
|
| 39 |
"models": {
|
| 40 |
-
"embedding": "all-MiniLM-L6-v2",
|
| 41 |
"qa": "deepset/roberta-base-squad2",
|
| 42 |
},
|
| 43 |
"chunking": {
|
|
@@ -237,6 +238,8 @@ class RAGIndex:
|
|
| 237 |
print("⚠️ No documents found in knowledge base")
|
| 238 |
print(f" Please add .txt, .md, .pdf, or .docx files to: {KB_DIR}")
|
| 239 |
self.index = None
|
|
|
|
|
|
|
| 240 |
return
|
| 241 |
|
| 242 |
all_chunks: List[str] = []
|
|
@@ -251,6 +254,8 @@ class RAGIndex:
|
|
| 251 |
if not all_chunks:
|
| 252 |
print("⚠️ No valid chunks created from documents")
|
| 253 |
self.index = None
|
|
|
|
|
|
|
| 254 |
return
|
| 255 |
|
| 256 |
print(f"Created {len(all_chunks)} chunks from {len(docs)} documents")
|
|
@@ -299,7 +304,10 @@ class RAGIndex:
|
|
| 299 |
try:
|
| 300 |
q_emb = self.embedder.encode([query], convert_to_numpy=True)
|
| 301 |
faiss.normalize_L2(q_emb)
|
| 302 |
-
|
|
|
|
|
|
|
|
|
|
| 303 |
|
| 304 |
results: List[Tuple[str, str, float]] = []
|
| 305 |
for score, idx in zip(scores[0], idxs[0]):
|
|
@@ -325,7 +333,7 @@ class RAGIndex:
|
|
| 325 |
if not question or not question.strip():
|
| 326 |
return "Please ask a question."
|
| 327 |
|
| 328 |
-
if self.index is None:
|
| 329 |
return (
|
| 330 |
f"📚 Knowledge base is empty.\n\n"
|
| 331 |
f"Please add documents to: `{KB_DIR}`\n"
|
|
@@ -390,24 +398,77 @@ print("=" * 50)
|
|
| 390 |
|
| 391 |
|
| 392 |
# -----------------------------
|
| 393 |
-
# GRADIO
|
| 394 |
# -----------------------------
|
| 395 |
|
| 396 |
def rag_respond(message, history):
|
| 397 |
-
"""Handle chat messages"""
|
| 398 |
if not message or not str(message).strip():
|
| 399 |
-
return "
|
| 400 |
-
|
| 401 |
-
return rag_index.answer(str(message))
|
| 402 |
|
| 403 |
-
|
|
|
|
|
|
|
|
|
|
| 404 |
|
| 405 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 406 |
description = WELCOME_MSG
|
| 407 |
-
if not rag_index.initialized or rag_index.index is None:
|
| 408 |
description += (
|
| 409 |
-
f"\n\n⚠️ **Note:** Knowledge base is empty
|
| 410 |
-
f"
|
| 411 |
)
|
| 412 |
|
| 413 |
examples = [
|
|
@@ -415,27 +476,85 @@ examples = [
|
|
| 415 |
for qa in CONFIG.get("quick_actions", [])
|
| 416 |
if qa.get("query")
|
| 417 |
]
|
| 418 |
-
if not examples and rag_index.initialized and rag_index.index is not None:
|
| 419 |
examples = [
|
| 420 |
"What is this document about?",
|
| 421 |
"Can you summarize the main points?",
|
| 422 |
"What are the key findings?",
|
| 423 |
]
|
| 424 |
|
| 425 |
-
chat = gr.ChatInterface(
|
| 426 |
-
fn=rag_respond,
|
| 427 |
-
title=CONFIG["client"]["name"],
|
| 428 |
-
description=description,
|
| 429 |
-
examples=examples if examples else None,
|
| 430 |
-
)
|
| 431 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 432 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 433 |
|
| 434 |
|
| 435 |
if __name__ == "__main__":
|
| 436 |
port = int(os.environ.get("PORT", 7860))
|
| 437 |
-
|
| 438 |
server_name="0.0.0.0",
|
| 439 |
server_port=port,
|
| 440 |
share=False,
|
| 441 |
-
)
|
|
|
|
| 1 |
import os
|
| 2 |
import glob
|
| 3 |
import yaml
|
| 4 |
+
import shutil
|
| 5 |
from typing import List, Tuple
|
| 6 |
|
| 7 |
import faiss
|
|
|
|
| 38 |
"index_directory": "./index",
|
| 39 |
},
|
| 40 |
"models": {
|
| 41 |
+
"embedding": "sentence-transformers/all-MiniLM-L6-v2",
|
| 42 |
"qa": "deepset/roberta-base-squad2",
|
| 43 |
},
|
| 44 |
"chunking": {
|
|
|
|
| 238 |
print("⚠️ No documents found in knowledge base")
|
| 239 |
print(f" Please add .txt, .md, .pdf, or .docx files to: {KB_DIR}")
|
| 240 |
self.index = None
|
| 241 |
+
self.chunks = []
|
| 242 |
+
self.chunk_sources = []
|
| 243 |
return
|
| 244 |
|
| 245 |
all_chunks: List[str] = []
|
|
|
|
| 254 |
if not all_chunks:
|
| 255 |
print("⚠️ No valid chunks created from documents")
|
| 256 |
self.index = None
|
| 257 |
+
self.chunks = []
|
| 258 |
+
self.chunk_sources = []
|
| 259 |
return
|
| 260 |
|
| 261 |
print(f"Created {len(all_chunks)} chunks from {len(docs)} documents")
|
|
|
|
| 304 |
try:
|
| 305 |
q_emb = self.embedder.encode([query], convert_to_numpy=True)
|
| 306 |
faiss.normalize_L2(q_emb)
|
| 307 |
+
k = min(top_k, len(self.chunks)) if self.chunks else 0
|
| 308 |
+
if k == 0:
|
| 309 |
+
return []
|
| 310 |
+
scores, idxs = self.index.search(q_emb, k)
|
| 311 |
|
| 312 |
results: List[Tuple[str, str, float]] = []
|
| 313 |
for score, idx in zip(scores[0], idxs[0]):
|
|
|
|
| 333 |
if not question or not question.strip():
|
| 334 |
return "Please ask a question."
|
| 335 |
|
| 336 |
+
if self.index is None or not self.chunks:
|
| 337 |
return (
|
| 338 |
f"📚 Knowledge base is empty.\n\n"
|
| 339 |
f"Please add documents to: `{KB_DIR}`\n"
|
|
|
|
| 398 |
|
| 399 |
|
| 400 |
# -----------------------------
|
| 401 |
+
# GRADIO APP (BLOCKS)
|
| 402 |
# -----------------------------
|
| 403 |
|
| 404 |
def rag_respond(message, history):
|
| 405 |
+
"""Handle chat messages for chatbot UI"""
|
| 406 |
if not message or not str(message).strip():
|
| 407 |
+
return "", history
|
|
|
|
|
|
|
| 408 |
|
| 409 |
+
user_msg = str(message)
|
| 410 |
+
bot_reply = rag_index.answer(user_msg)
|
| 411 |
+
history = history + [[user_msg, bot_reply]]
|
| 412 |
+
return "", history
|
| 413 |
|
| 414 |
+
|
| 415 |
+
def upload_to_kb(files):
|
| 416 |
+
"""Save uploaded files into the KB directory"""
|
| 417 |
+
if not files:
|
| 418 |
+
return "No files uploaded."
|
| 419 |
+
|
| 420 |
+
if not isinstance(files, list):
|
| 421 |
+
files = [files]
|
| 422 |
+
|
| 423 |
+
os.makedirs(KB_DIR, exist_ok=True)
|
| 424 |
+
saved_files = []
|
| 425 |
+
|
| 426 |
+
for f in files:
|
| 427 |
+
# Gradio File object or temp file path
|
| 428 |
+
# In older Gradio, `f.name` is the temp file path
|
| 429 |
+
src_path = getattr(f, "name", None) or str(f)
|
| 430 |
+
if not os.path.exists(src_path):
|
| 431 |
+
continue
|
| 432 |
+
|
| 433 |
+
filename = os.path.basename(src_path)
|
| 434 |
+
dest_path = os.path.join(KB_DIR, filename)
|
| 435 |
+
|
| 436 |
+
try:
|
| 437 |
+
shutil.copy(src_path, dest_path)
|
| 438 |
+
saved_files.append(filename)
|
| 439 |
+
except Exception as e:
|
| 440 |
+
print(f"Error saving file {filename}: {e}")
|
| 441 |
+
|
| 442 |
+
if not saved_files:
|
| 443 |
+
return "No files could be saved. Check logs."
|
| 444 |
+
|
| 445 |
+
return (
|
| 446 |
+
f"✅ Saved {len(saved_files)} file(s) to knowledge base:\n- "
|
| 447 |
+
+ "\n- ".join(saved_files)
|
| 448 |
+
+ "\n\nClick **Rebuild index** to include them in search."
|
| 449 |
+
)
|
| 450 |
+
|
| 451 |
+
|
| 452 |
+
def rebuild_index():
|
| 453 |
+
"""Trigger index rebuild from UI"""
|
| 454 |
+
rag_index._build_or_load_index()
|
| 455 |
+
if rag_index.index is None or not rag_index.chunks:
|
| 456 |
+
return (
|
| 457 |
+
"⚠️ Index rebuild finished, but no documents or chunks were found.\n"
|
| 458 |
+
f"Add files to `{KB_DIR}` and try again."
|
| 459 |
+
)
|
| 460 |
+
return (
|
| 461 |
+
f"✅ Index rebuilt successfully.\n"
|
| 462 |
+
f"Chunks in index: {len(rag_index.chunks)}"
|
| 463 |
+
)
|
| 464 |
+
|
| 465 |
+
|
| 466 |
+
# Description + examples
|
| 467 |
description = WELCOME_MSG
|
| 468 |
+
if not rag_index.initialized or rag_index.index is None or not rag_index.chunks:
|
| 469 |
description += (
|
| 470 |
+
f"\n\n⚠️ **Note:** Knowledge base is currently empty or index is not built.\n"
|
| 471 |
+
f"Upload documents in the **Knowledge Base** tab and click **Rebuild index**."
|
| 472 |
)
|
| 473 |
|
| 474 |
examples = [
|
|
|
|
| 476 |
for qa in CONFIG.get("quick_actions", [])
|
| 477 |
if qa.get("query")
|
| 478 |
]
|
| 479 |
+
if not examples and rag_index.initialized and rag_index.index is not None and rag_index.chunks:
|
| 480 |
examples = [
|
| 481 |
"What is this document about?",
|
| 482 |
"Can you summarize the main points?",
|
| 483 |
"What are the key findings?",
|
| 484 |
]
|
| 485 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 486 |
|
| 487 |
+
with gr.Blocks(title=CONFIG["client"]["name"]) as demo:
|
| 488 |
+
gr.Markdown(f"# {CONFIG['client']['name']}")
|
| 489 |
+
gr.Markdown(description)
|
| 490 |
+
|
| 491 |
+
with gr.Tab("Chat"):
|
| 492 |
+
chatbot = gr.Chatbot(label="RAG Chat")
|
| 493 |
+
with gr.Row():
|
| 494 |
+
txt = gr.Textbox(
|
| 495 |
+
show_label=False,
|
| 496 |
+
placeholder="Ask a question about your documents...",
|
| 497 |
+
lines=2,
|
| 498 |
+
)
|
| 499 |
+
with gr.Row():
|
| 500 |
+
send_btn = gr.Button("Send")
|
| 501 |
+
clear_btn = gr.Button("Clear")
|
| 502 |
+
|
| 503 |
+
# Pre-fill example buttons if available
|
| 504 |
+
if examples:
|
| 505 |
+
gr.Markdown("### Example questions")
|
| 506 |
+
example_btns = []
|
| 507 |
+
with gr.Row():
|
| 508 |
+
for ex in examples:
|
| 509 |
+
example_btns.append(gr.Button(ex))
|
| 510 |
+
|
| 511 |
+
def use_example(example, history):
|
| 512 |
+
"""When clicking an example, send it as a message"""
|
| 513 |
+
bot_reply = rag_index.answer(example)
|
| 514 |
+
history = history + [[example, bot_reply]]
|
| 515 |
+
return history
|
| 516 |
+
|
| 517 |
+
for btn, ex in zip(example_btns, examples):
|
| 518 |
+
btn.click(
|
| 519 |
+
use_example,
|
| 520 |
+
inputs=[gr.State(ex), chatbot],
|
| 521 |
+
outputs=chatbot,
|
| 522 |
+
)
|
| 523 |
+
|
| 524 |
+
# Chat logic wiring
|
| 525 |
+
txt.submit(rag_respond, [txt, chatbot], [txt, chatbot])
|
| 526 |
+
send_btn.click(rag_respond, [txt, chatbot], [txt, chatbot])
|
| 527 |
+
clear_btn.click(lambda: ([], ""), None, [chatbot, txt])
|
| 528 |
+
|
| 529 |
+
with gr.Tab("Knowledge Base"):
|
| 530 |
+
gr.Markdown(
|
| 531 |
+
f"""
|
| 532 |
+
### Manage Knowledge Base
|
| 533 |
|
| 534 |
+
- Supported formats: `.txt`, `.md`, `.pdf`, `.docx`, `.doc`
|
| 535 |
+
- Files are stored in: `{KB_DIR}`
|
| 536 |
+
- After uploading, click **Rebuild index** so the assistant can use the new content.
|
| 537 |
+
"""
|
| 538 |
+
)
|
| 539 |
+
kb_upload = gr.File(
|
| 540 |
+
label="Upload documents",
|
| 541 |
+
file_count="multiple",
|
| 542 |
+
)
|
| 543 |
+
kb_status = gr.Textbox(
|
| 544 |
+
label="Status",
|
| 545 |
+
lines=6,
|
| 546 |
+
interactive=False,
|
| 547 |
+
)
|
| 548 |
+
rebuild_btn = gr.Button("Rebuild index")
|
| 549 |
+
|
| 550 |
+
kb_upload.change(upload_to_kb, inputs=kb_upload, outputs=kb_status)
|
| 551 |
+
rebuild_btn.click(rebuild_index, inputs=None, outputs=kb_status)
|
| 552 |
|
| 553 |
|
| 554 |
if __name__ == "__main__":
|
| 555 |
port = int(os.environ.get("PORT", 7860))
|
| 556 |
+
demo.launch(
|
| 557 |
server_name="0.0.0.0",
|
| 558 |
server_port=port,
|
| 559 |
share=False,
|
| 560 |
+
)
|