Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -9,8 +9,8 @@ import faiss
|
|
9 |
import numpy as np
|
10 |
from twilio.rest import Client
|
11 |
from groq import Groq
|
12 |
-
import re
|
13 |
-
|
14 |
# --- Page Configuration ---
|
15 |
st.set_page_config(page_title="RAG Customer Support Chatbot", layout="wide")
|
16 |
|
@@ -42,142 +42,142 @@ APP_TWILIO_BOT_WHATSAPP_IDENTITY_SECRET = st.secrets.get("TWILIO_BOT_WHATSAPP_ID
|
|
42 |
|
43 |
# --- RAG Processing Utilities ---
|
44 |
def load_json_data(file_path):
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
|
60 |
def load_pdf_data(file_path):
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
|
77 |
def chunk_text(text_pages, chunk_size=1000, chunk_overlap=200):
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
|
94 |
@st.cache_resource(show_spinner="Initializing embedding model...")
|
95 |
def initialize_embedding_model(model_name=DEFAULT_EMBEDDING_MODEL_NAME):
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
|
104 |
@st.cache_resource(show_spinner="Building FAISS index for PDF documents...")
|
105 |
def create_faiss_index(_text_chunks, _embedding_model):
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
|
129 |
def search_faiss_index(index, query_text, embedding_model, indexed_chunks, k=3):
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
|
148 |
def get_order_details(order_id, customer_orders_data):
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
|
157 |
def get_product_info(query, products_data):
|
158 |
-
|
159 |
-
|
160 |
-
|
161 |
-
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
|
172 |
# --- LLM Operations ---
|
173 |
@st.cache_data(show_spinner="Generating response with LLaMA3...")
|
174 |
def generate_response_groq(_groq_client, query, context, model="llama3-8b-8192"):
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
Use the following context to answer the user's question.
|
182 |
If the context doesn't contain the answer, state that you don't have enough information.
|
183 |
Do not make up information. Be concise and polite.
|
@@ -189,107 +189,107 @@ User Question: {query}
|
|
189 |
|
190 |
Assistant Answer:
|
191 |
"""
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
|
206 |
def initialize_groq_client(api_key_val):
|
207 |
-
|
208 |
-
|
209 |
-
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
|
218 |
# --- Twilio Operations ---
|
219 |
def initialize_twilio_client(acc_sid, auth_tkn): # Changed parameter names
|
220 |
-
|
221 |
-
|
222 |
-
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
230 |
|
231 |
def get_new_whatsapp_messages(twilio_client, conversation_service_sid_val, bot_start_time_utc, # Renamed
|
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 |
def send_whatsapp_message(twilio_client, conversation_service_sid_val, conversation_sid, message_body, bot_identity_val): # Renamed
|
272 |
-
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
|
277 |
-
|
278 |
-
|
279 |
-
|
280 |
-
|
281 |
-
|
282 |
-
|
283 |
-
|
284 |
-
|
285 |
-
|
286 |
-
|
287 |
-
|
288 |
-
|
289 |
-
|
290 |
-
|
291 |
-
|
292 |
-
|
293 |
|
294 |
# --- Main Application Logic & UI ---
|
295 |
st.title("🤖 RAG-Based Customer Support Chatbot")
|
@@ -300,47 +300,47 @@ st.sidebar.title("⚙️ Configurations")
|
|
300 |
|
301 |
# Use APP_ prefixed variables for values from secrets, then allow manual input if not found
|
302 |
if APP_TWILIO_ACCOUNT_SID:
|
303 |
-
|
304 |
-
|
305 |
else:
|
306 |
-
|
307 |
-
|
308 |
|
309 |
if APP_TWILIO_AUTH_TOKEN:
|
310 |
-
|
311 |
-
|
312 |
else:
|
313 |
-
|
314 |
-
|
315 |
|
316 |
if APP_GROQ_API_KEY:
|
317 |
-
|
318 |
-
|
319 |
else:
|
320 |
-
|
321 |
-
|
322 |
|
323 |
# For other configurations that can be overridden if secrets not found or for user preference
|
324 |
twilio_conversation_service_sid_to_use = st.sidebar.text_input(
|
325 |
-
|
326 |
-
|
327 |
-
|
328 |
-
|
329 |
)
|
330 |
twilio_bot_whatsapp_identity_to_use = st.sidebar.text_input(
|
331 |
-
|
332 |
-
|
333 |
-
|
334 |
)
|
335 |
embedding_model_name_to_use = st.sidebar.text_input( # Renamed
|
336 |
-
|
337 |
-
|
338 |
)
|
339 |
polling_interval_to_use = st.sidebar.number_input( # Renamed
|
340 |
-
|
341 |
-
|
342 |
-
|
343 |
-
|
344 |
)
|
345 |
|
346 |
# --- Initialize Session State ---
|
@@ -354,197 +354,197 @@ if "manual_chat_history" not in st.session_state: st.session_state.manual_chat_h
|
|
354 |
|
355 |
# --- Helper: Simple Intent Classifier ---
|
356 |
def simple_intent_classifier(query):
|
357 |
-
|
358 |
-
|
359 |
-
|
360 |
-
|
361 |
-
|
362 |
-
|
363 |
-
|
364 |
-
|
365 |
-
|
366 |
-
|
367 |
-
|
368 |
-
|
369 |
|
370 |
# --- Main Application Controls ---
|
371 |
col1, col2, col3, col4 = st.columns(4)
|
372 |
with col1:
|
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 |
with col2:
|
405 |
-
|
406 |
-
|
407 |
-
|
408 |
-
|
409 |
-
|
410 |
-
|
411 |
-
|
412 |
-
|
413 |
-
|
414 |
-
|
415 |
-
|
416 |
-
|
417 |
-
|
418 |
-
|
419 |
with col3:
|
420 |
-
|
421 |
-
|
422 |
-
|
423 |
-
|
424 |
-
|
425 |
-
|
426 |
-
|
427 |
-
|
428 |
-
|
429 |
-
|
430 |
-
|
431 |
-
|
432 |
-
|
433 |
-
|
434 |
with col4:
|
435 |
-
|
436 |
-
|
437 |
-
|
438 |
-
|
439 |
st.divider()
|
440 |
|
441 |
# --- Manual Query Interface ---
|
442 |
if st.session_state.get("app_started") and st.session_state.get("rag_pipeline_ready"):
|
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 |
# --- Twilio Bot Polling Logic ---
|
525 |
if st.session_state.get("bot_started") and st.session_state.get("rag_pipeline_ready"):
|
526 |
-
|
527 |
-
|
528 |
-
|
529 |
-
|
530 |
-
|
531 |
-
|
532 |
-
|
533 |
-
|
534 |
-
|
535 |
-
|
536 |
-
|
537 |
-
|
538 |
-
|
539 |
-
|
540 |
-
|
541 |
-
|
542 |
-
|
543 |
-
|
544 |
-
|
545 |
-
|
546 |
-
|
547 |
-
|
548 |
order_id_to_check_whatsapp = None
|
549 |
if potential_oid_whatsapp:
|
550 |
order_id_to_check_whatsapp = potential_oid_whatsapp
|
@@ -555,29 +555,29 @@ if st.session_state.get("bot_started") and st.session_state.get("rag_pipeline_re
|
|
555 |
order_id_to_check_whatsapp = possible_match_whatsapp
|
556 |
|
557 |
if order_id_to_check_whatsapp:
|
558 |
-
|
559 |
else:
|
560 |
context_whatsapp = "Please provide a valid Order ID."
|
561 |
-
|
562 |
-
|
563 |
-
|
564 |
-
|
565 |
-
|
566 |
-
|
567 |
-
|
568 |
-
|
569 |
-
|
570 |
-
|
571 |
-
|
572 |
-
|
573 |
-
|
574 |
-
|
575 |
-
|
576 |
|
577 |
# --- Footer & Status ---
|
578 |
st.sidebar.markdown("---")
|
579 |
st.sidebar.info("Ensure all keys and SIDs are correctly configured. Primary API keys (Twilio SID/Token, GROQ Key) are loaded from secrets if available.")
|
580 |
if st.session_state.get("app_started"):
|
581 |
-
|
582 |
else:
|
583 |
-
|
|
|
9 |
import numpy as np
|
10 |
from twilio.rest import Client
|
11 |
from groq import Groq
|
12 |
+
import re
|
13 |
+
|
14 |
# --- Page Configuration ---
|
15 |
st.set_page_config(page_title="RAG Customer Support Chatbot", layout="wide")
|
16 |
|
|
|
42 |
|
43 |
# --- RAG Processing Utilities ---
|
44 |
def load_json_data(file_path):
|
45 |
+
"""Loads data from a JSON file."""
|
46 |
+
try:
|
47 |
+
with open(file_path, 'r', encoding='utf-8') as f:
|
48 |
+
data = json.load(f)
|
49 |
+
return data
|
50 |
+
except FileNotFoundError:
|
51 |
+
st.error(f"Error: JSON file not found at {file_path}")
|
52 |
+
return None
|
53 |
+
except json.JSONDecodeError:
|
54 |
+
st.error(f"Error: Could not decode JSON from {file_path}")
|
55 |
+
return None
|
56 |
+
except Exception as e:
|
57 |
+
st.error(f"An unexpected error occurred while loading {file_path}: {e}")
|
58 |
+
return None
|
59 |
|
60 |
def load_pdf_data(file_path):
|
61 |
+
"""Extracts text from a PDF file, page by page."""
|
62 |
+
try:
|
63 |
+
with open(file_path, 'rb') as f:
|
64 |
+
reader = PyPDF2.PdfReader(f)
|
65 |
+
text_pages = []
|
66 |
+
for page_num in range(len(reader.pages)):
|
67 |
+
page = reader.pages[page_num]
|
68 |
+
text_pages.append(page.extract_text() or "")
|
69 |
+
return text_pages
|
70 |
+
except FileNotFoundError:
|
71 |
+
st.error(f"Error: PDF file not found at {file_path}")
|
72 |
+
return []
|
73 |
+
except Exception as e:
|
74 |
+
st.error(f"An error occurred while processing PDF {file_path}: {e}")
|
75 |
+
return []
|
76 |
|
77 |
def chunk_text(text_pages, chunk_size=1000, chunk_overlap=200):
|
78 |
+
"""Chunks text from PDF pages into smaller, overlapping pieces."""
|
79 |
+
full_text = "\n".join(text_pages)
|
80 |
+
if not full_text.strip():
|
81 |
+
return []
|
82 |
+
chunks = []
|
83 |
+
start = 0
|
84 |
+
while start < len(full_text):
|
85 |
+
end = start + chunk_size
|
86 |
+
chunks.append(full_text[start:end])
|
87 |
+
if end >= len(full_text):
|
88 |
+
break
|
89 |
+
start += (chunk_size - chunk_overlap)
|
90 |
+
if start >= len(full_text):
|
91 |
+
break
|
92 |
+
return [chunk for chunk in chunks if chunk.strip()]
|
93 |
|
94 |
@st.cache_resource(show_spinner="Initializing embedding model...")
|
95 |
def initialize_embedding_model(model_name=DEFAULT_EMBEDDING_MODEL_NAME):
|
96 |
+
"""Initializes and returns a SentenceTransformer model."""
|
97 |
+
try:
|
98 |
+
model = SentenceTransformer(model_name)
|
99 |
+
return model
|
100 |
+
except Exception as e:
|
101 |
+
st.error(f"Error initializing embedding model '{model_name}': {e}")
|
102 |
+
return None
|
103 |
|
104 |
@st.cache_resource(show_spinner="Building FAISS index for PDF documents...")
|
105 |
def create_faiss_index(_text_chunks, _embedding_model):
|
106 |
+
"""Creates a FAISS index from text chunks and an embedding model."""
|
107 |
+
if not _text_chunks or _embedding_model is None:
|
108 |
+
st.warning("Cannot create FAISS index: No text chunks or embedding model available.")
|
109 |
+
return None, []
|
110 |
+
try:
|
111 |
+
valid_chunks = [str(chunk) for chunk in _text_chunks if chunk and isinstance(chunk, str) and chunk.strip()]
|
112 |
+
if not valid_chunks:
|
113 |
+
st.warning("No valid text chunks to embed for FAISS index.")
|
114 |
+
return None, []
|
115 |
+
embeddings = _embedding_model.encode(valid_chunks, convert_to_tensor=False)
|
116 |
+
if embeddings.ndim == 1:
|
117 |
+
embeddings = embeddings.reshape(1, -1)
|
118 |
+
if embeddings.shape[0] == 0:
|
119 |
+
st.warning("No embeddings were generated for FAISS index.")
|
120 |
+
return None, []
|
121 |
+
dimension = embeddings.shape[1]
|
122 |
+
index = faiss.IndexFlatL2(dimension)
|
123 |
+
index.add(np.array(embeddings, dtype=np.float32))
|
124 |
+
return index, valid_chunks
|
125 |
+
except Exception as e:
|
126 |
+
st.error(f"Error creating FAISS index: {e}")
|
127 |
+
return None, []
|
128 |
|
129 |
def search_faiss_index(index, query_text, embedding_model, indexed_chunks, k=3):
|
130 |
+
"""Searches the FAISS index and returns top_k relevant chunk texts."""
|
131 |
+
if index is None or embedding_model is None or not query_text:
|
132 |
+
return []
|
133 |
+
try:
|
134 |
+
query_embedding = embedding_model.encode([query_text], convert_to_tensor=False)
|
135 |
+
if query_embedding.ndim == 1:
|
136 |
+
query_embedding = query_embedding.reshape(1, -1)
|
137 |
+
distances, indices = index.search(np.array(query_embedding, dtype=np.float32), k)
|
138 |
+
results = []
|
139 |
+
for i in range(len(indices[0])):
|
140 |
+
idx = indices[0][i]
|
141 |
+
if 0 <= idx < len(indexed_chunks):
|
142 |
+
results.append(indexed_chunks[idx])
|
143 |
+
return results
|
144 |
+
except Exception as e:
|
145 |
+
st.error(f"Error searching FAISS index: {e}")
|
146 |
+
return []
|
147 |
|
148 |
def get_order_details(order_id, customer_orders_data):
|
149 |
+
"""Retrieves order details for a given order_id."""
|
150 |
+
if not customer_orders_data:
|
151 |
+
return "Customer order data is not loaded."
|
152 |
+
for order in customer_orders_data:
|
153 |
+
if order.get("order_id") == order_id:
|
154 |
+
return json.dumps(order, indent=2)
|
155 |
+
return f"No order found with ID: {order_id}."
|
156 |
|
157 |
def get_product_info(query, products_data):
|
158 |
+
"""Retrieves product information based on a query."""
|
159 |
+
if not products_data:
|
160 |
+
return "Product data is not loaded."
|
161 |
+
query_lower = query.lower()
|
162 |
+
found_products = []
|
163 |
+
for product in products_data:
|
164 |
+
if query_lower in (product.get("name", "").lower()) or \
|
165 |
+
query_lower in (product.get("description", "").lower()) or \
|
166 |
+
query_lower == (product.get("product_id", "").lower()):
|
167 |
+
found_products.append(product)
|
168 |
+
if found_products:
|
169 |
+
return json.dumps(found_products, indent=2)
|
170 |
+
return f"No product information found matching your query: '{query}'."
|
171 |
|
172 |
# --- LLM Operations ---
|
173 |
@st.cache_data(show_spinner="Generating response with LLaMA3...")
|
174 |
def generate_response_groq(_groq_client, query, context, model="llama3-8b-8192"):
|
175 |
+
"""Generates a response using GROQ LLaMA3 API."""
|
176 |
+
if not _groq_client:
|
177 |
+
return "GROQ client not initialized. Please check API key."
|
178 |
+
if not query:
|
179 |
+
return "Query is empty."
|
180 |
+
prompt = f"""You are a helpful customer support assistant.
|
181 |
Use the following context to answer the user's question.
|
182 |
If the context doesn't contain the answer, state that you don't have enough information.
|
183 |
Do not make up information. Be concise and polite.
|
|
|
189 |
|
190 |
Assistant Answer:
|
191 |
"""
|
192 |
+
try:
|
193 |
+
chat_completion = _groq_client.chat.completions.create(
|
194 |
+
messages=[
|
195 |
+
{"role": "system", "content": "You are a helpful customer support assistant."},
|
196 |
+
{"role": "user", "content": prompt}
|
197 |
+
],
|
198 |
+
model=model, temperature=0.7, max_tokens=1024, top_p=1
|
199 |
+
)
|
200 |
+
response = chat_completion.choices[0].message.content
|
201 |
+
return response
|
202 |
+
except Exception as e:
|
203 |
+
st.error(f"Error calling GROQ API: {e}")
|
204 |
+
return "Sorry, I encountered an error while trying to generate a response."
|
205 |
|
206 |
def initialize_groq_client(api_key_val):
|
207 |
+
"""Initializes the GROQ client."""
|
208 |
+
if not api_key_val: # Changed parameter name to avoid conflict
|
209 |
+
st.warning("GROQ API Key is missing.")
|
210 |
+
return None
|
211 |
+
try:
|
212 |
+
client = Groq(api_key=api_key_val)
|
213 |
+
return client
|
214 |
+
except Exception as e:
|
215 |
+
st.error(f"Failed to initialize GROQ client: {e}")
|
216 |
+
return None
|
217 |
|
218 |
# --- Twilio Operations ---
|
219 |
def initialize_twilio_client(acc_sid, auth_tkn): # Changed parameter names
|
220 |
+
"""Initializes the Twilio client."""
|
221 |
+
if not acc_sid or not auth_tkn:
|
222 |
+
st.warning("Twilio Account SID or Auth Token is missing.")
|
223 |
+
return None
|
224 |
+
try:
|
225 |
+
client = Client(acc_sid, auth_tkn)
|
226 |
+
return client
|
227 |
+
except Exception as e:
|
228 |
+
st.error(f"Failed to initialize Twilio client: {e}")
|
229 |
+
return None
|
230 |
|
231 |
def get_new_whatsapp_messages(twilio_client, conversation_service_sid_val, bot_start_time_utc, # Renamed
|
232 |
+
processed_message_sids, bot_whatsapp_identity_val): # Renamed
|
233 |
+
"""Fetches new, unanswered WhatsApp messages from Twilio Conversations."""
|
234 |
+
if not twilio_client:
|
235 |
+
st.warning("Twilio client not initialized.")
|
236 |
+
return []
|
237 |
+
if not conversation_service_sid_val:
|
238 |
+
st.warning("Twilio Conversation Service SID not provided.")
|
239 |
+
return []
|
240 |
+
|
241 |
+
new_messages_to_process = []
|
242 |
+
try:
|
243 |
+
conversations = twilio_client.conversations.v1 \
|
244 |
+
.services(conversation_service_sid_val) \
|
245 |
+
.conversations \
|
246 |
+
.list(limit=50)
|
247 |
+
|
248 |
+
for conv in conversations:
|
249 |
+
if conv.date_updated and conv.date_updated > bot_start_time_utc:
|
250 |
+
messages = twilio_client.conversations.v1 \
|
251 |
+
.services(conversation_service_sid_val) \
|
252 |
+
.conversations(conv.sid) \
|
253 |
+
.messages \
|
254 |
+
.list(order='desc', limit=10)
|
255 |
+
|
256 |
+
for msg in messages:
|
257 |
+
if msg.sid in processed_message_sids:
|
258 |
+
continue
|
259 |
+
if msg.author and msg.author.lower() != bot_whatsapp_identity_val.lower() and \
|
260 |
+
msg.date_created and msg.date_created > bot_start_time_utc:
|
261 |
+
new_messages_to_process.append({
|
262 |
+
"conversation_sid": conv.sid, "message_sid": msg.sid,
|
263 |
+
"author_identity": msg.author, "message_body": msg.body,
|
264 |
+
"timestamp_utc": msg.date_created
|
265 |
+
})
|
266 |
+
break
|
267 |
+
except Exception as e:
|
268 |
+
st.error(f"Error fetching Twilio messages: {e}")
|
269 |
+
return sorted(new_messages_to_process, key=lambda m: m['timestamp_utc'])
|
270 |
|
271 |
def send_whatsapp_message(twilio_client, conversation_service_sid_val, conversation_sid, message_body, bot_identity_val): # Renamed
|
272 |
+
"""Sends a message to a Twilio Conversation from the bot's identity."""
|
273 |
+
if not twilio_client:
|
274 |
+
st.error("Twilio client not initialized for sending message.")
|
275 |
+
return False
|
276 |
+
if not conversation_service_sid_val:
|
277 |
+
st.error("Twilio Conversation Service SID not provided for sending message.")
|
278 |
+
return False
|
279 |
+
if not bot_identity_val:
|
280 |
+
st.error("Bot identity not provided for sending message.")
|
281 |
+
return False
|
282 |
+
try:
|
283 |
+
twilio_client.conversations.v1 \
|
284 |
+
.services(conversation_service_sid_val) \
|
285 |
+
.conversations(conversation_sid) \
|
286 |
+
.messages \
|
287 |
+
.create(author=bot_identity_val, body=message_body)
|
288 |
+
st.success(f"Sent reply to conversation {conversation_sid}")
|
289 |
+
return True
|
290 |
+
except Exception as e:
|
291 |
+
st.error(f"Error sending Twilio message to {conversation_sid}: {e}")
|
292 |
+
return False
|
293 |
|
294 |
# --- Main Application Logic & UI ---
|
295 |
st.title("🤖 RAG-Based Customer Support Chatbot")
|
|
|
300 |
|
301 |
# Use APP_ prefixed variables for values from secrets, then allow manual input if not found
|
302 |
if APP_TWILIO_ACCOUNT_SID:
|
303 |
+
st.sidebar.text_input("Twilio Account SID (from Secrets)", value="********" + APP_TWILIO_ACCOUNT_SID[-4:] if len(APP_TWILIO_ACCOUNT_SID) > 4 else "********", disabled=True)
|
304 |
+
twilio_account_sid_to_use = APP_TWILIO_ACCOUNT_SID
|
305 |
else:
|
306 |
+
st.sidebar.warning("Secret 'TWILIO_SID' not found.")
|
307 |
+
twilio_account_sid_to_use = st.sidebar.text_input("Twilio Account SID (Enter Manually)", value=DEFAULT_TWILIO_ACCOUNT_SID_FALLBACK, type="password")
|
308 |
|
309 |
if APP_TWILIO_AUTH_TOKEN:
|
310 |
+
st.sidebar.text_input("Twilio Auth Token (from Secrets)", value="********", disabled=True)
|
311 |
+
twilio_auth_token_to_use = APP_TWILIO_AUTH_TOKEN
|
312 |
else:
|
313 |
+
st.sidebar.warning("Secret 'TWILIO_TOKEN' not found.")
|
314 |
+
twilio_auth_token_to_use = st.sidebar.text_input("Twilio Auth Token (Enter Manually)", value=DEFAULT_TWILIO_AUTH_TOKEN_FALLBACK, type="password")
|
315 |
|
316 |
if APP_GROQ_API_KEY:
|
317 |
+
st.sidebar.text_input("GROQ API Key (from Secrets)", value="gsk_********" + APP_GROQ_API_KEY[-4:] if len(APP_GROQ_API_KEY) > 8 else "********", disabled=True)
|
318 |
+
groq_api_key_to_use = APP_GROQ_API_KEY
|
319 |
else:
|
320 |
+
st.sidebar.warning("Secret 'GROQ_API_KEY' not found.")
|
321 |
+
groq_api_key_to_use = st.sidebar.text_input("GROQ API Key (Enter Manually)", value=DEFAULT_GROQ_API_KEY_FALLBACK, type="password")
|
322 |
|
323 |
# For other configurations that can be overridden if secrets not found or for user preference
|
324 |
twilio_conversation_service_sid_to_use = st.sidebar.text_input(
|
325 |
+
"Twilio Conversation Service SID (IS...)",
|
326 |
+
value=APP_TWILIO_CONVERSATION_SERVICE_SID_SECRET or DEFAULT_TWILIO_CONVERSATION_SERVICE_SID,
|
327 |
+
type="password",
|
328 |
+
help="The SID of your Twilio Conversations Service. Can be set by 'TWILIO_CONVERSATION_SERVICE_SID' secret."
|
329 |
)
|
330 |
twilio_bot_whatsapp_identity_to_use = st.sidebar.text_input(
|
331 |
+
"Twilio Bot WhatsApp Identity",
|
332 |
+
value=APP_TWILIO_BOT_WHATSAPP_IDENTITY_SECRET or DEFAULT_TWILIO_BOT_WHATSAPP_IDENTITY,
|
333 |
+
help="e.g., 'whatsapp:+1234567890'. Can be set by 'TWILIO_BOT_WHATSAPP_IDENTITY' secret."
|
334 |
)
|
335 |
embedding_model_name_to_use = st.sidebar.text_input( # Renamed
|
336 |
+
"Embedding Model Name",
|
337 |
+
value=DEFAULT_EMBEDDING_MODEL_NAME
|
338 |
)
|
339 |
polling_interval_to_use = st.sidebar.number_input( # Renamed
|
340 |
+
"Twilio Polling Interval (seconds)",
|
341 |
+
min_value=10, max_value=300,
|
342 |
+
value=DEFAULT_POLLING_INTERVAL_S,
|
343 |
+
step=5
|
344 |
)
|
345 |
|
346 |
# --- Initialize Session State ---
|
|
|
354 |
|
355 |
# --- Helper: Simple Intent Classifier ---
|
356 |
def simple_intent_classifier(query):
|
357 |
+
query_lower = query.lower()
|
358 |
+
if any(k in query_lower for k in ["order", "status", "track", "delivery"]):
|
359 |
+
# More specific regex to find 'ORD' followed by digits (assuming order IDs are like ORD1001)
|
360 |
+
match = re.search(r'\b(ord\d{3,})\b', query_lower) # Matches 'ord' followed by at least 3 digits, as a whole word
|
361 |
+
if match:
|
362 |
+
return "ORDER_STATUS", match.group(1).upper() # Return intent and extracted ID
|
363 |
+
# Fallback if specific order ID not found but still an order-related query
|
364 |
+
return "ORDER_STATUS", None # Indicate order status intent but no specific ID found yet
|
365 |
+
|
366 |
+
if any(k in query_lower for k in ["product", "item", "buy", "price", "feature", "stock"]): return "PRODUCT_INFO", None
|
367 |
+
if any(k in query_lower for k in ["return", "policy", "refund", "exchange", "faq", "question", "how to", "support"]): return "GENERAL_POLICY_FAQ", None
|
368 |
+
return "UNKNOWN", None # Return intent and None for ID if unknown
|
369 |
|
370 |
# --- Main Application Controls ---
|
371 |
col1, col2, col3, col4 = st.columns(4)
|
372 |
with col1:
|
373 |
+
if st.button("🚀 Start App", disabled=st.session_state.app_started, use_container_width=True):
|
374 |
+
if not groq_api_key_to_use: # Use the correct variable
|
375 |
+
st.error("GROQ API Key is required.")
|
376 |
+
else:
|
377 |
+
with st.spinner("Initializing RAG pipeline..."):
|
378 |
+
st.session_state.embedding_model = initialize_embedding_model(embedding_model_name_to_use) # Use correct var
|
379 |
+
st.session_state.customer_orders_data = load_json_data(CUSTOMER_ORDERS_FILE)
|
380 |
+
st.session_state.products_data = load_json_data(PRODUCTS_FILE)
|
381 |
+
policy_pdf_pages = load_pdf_data(POLICY_PDF_FILE)
|
382 |
+
faq_pdf_pages = load_pdf_data(FAQ_PDF_FILE)
|
383 |
+
all_pdf_text_pages = policy_pdf_pages + faq_pdf_pages
|
384 |
+
st.session_state.pdf_text_chunks_raw = chunk_text(all_pdf_text_pages)
|
385 |
+
|
386 |
+
if st.session_state.embedding_model and st.session_state.pdf_text_chunks_raw:
|
387 |
+
st.session_state.faiss_index_pdfs, st.session_state.indexed_pdf_chunks = \
|
388 |
+
create_faiss_index(st.session_state.pdf_text_chunks_raw, st.session_state.embedding_model)
|
389 |
+
else:
|
390 |
+
st.session_state.faiss_index_pdfs, st.session_state.indexed_pdf_chunks = None, []
|
391 |
+
st.warning("FAISS index for PDFs could not be created.")
|
392 |
+
|
393 |
+
st.session_state.groq_client = initialize_groq_client(groq_api_key_to_use) # Use correct var
|
394 |
+
|
395 |
+
if st.session_state.embedding_model and st.session_state.groq_client and \
|
396 |
+
st.session_state.customer_orders_data and st.session_state.products_data:
|
397 |
+
st.session_state.rag_pipeline_ready = True
|
398 |
+
st.session_state.app_started = True
|
399 |
+
st.success("RAG Application Started!")
|
400 |
+
st.rerun()
|
401 |
+
else:
|
402 |
+
st.error("Failed to initialize RAG pipeline. Check configurations and ensure all data files are present in 'docs/'.")
|
403 |
+
st.session_state.app_started = False
|
404 |
with col2:
|
405 |
+
if st.button("🛑 Stop App", disabled=not st.session_state.app_started, use_container_width=True):
|
406 |
+
keys_to_reset = ["app_started", "bot_started", "rag_pipeline_ready", "embedding_model",
|
407 |
+
"customer_orders_data", "products_data", "pdf_text_chunks_raw",
|
408 |
+
"faiss_index_pdfs", "indexed_pdf_chunks", "groq_client", "twilio_client",
|
409 |
+
"bot_start_time_utc", "processed_message_sids", "manual_chat_history"]
|
410 |
+
for key in keys_to_reset:
|
411 |
+
if key in st.session_state: del st.session_state[key]
|
412 |
+
st.session_state.app_started = False
|
413 |
+
st.session_state.bot_started = False
|
414 |
+
st.session_state.rag_pipeline_ready = False
|
415 |
+
st.session_state.processed_message_sids = set()
|
416 |
+
st.session_state.manual_chat_history = []
|
417 |
+
st.success("Application Stopped.")
|
418 |
+
st.rerun()
|
419 |
with col3:
|
420 |
+
if st.button("💬 Start WhatsApp Bot", disabled=not st.session_state.app_started or st.session_state.bot_started, use_container_width=True):
|
421 |
+
if not all([twilio_account_sid_to_use, twilio_auth_token_to_use, twilio_conversation_service_sid_to_use, twilio_bot_whatsapp_identity_to_use]): # Use correct vars
|
422 |
+
st.error("Twilio credentials, Service SID, and Bot Identity are required.")
|
423 |
+
else:
|
424 |
+
st.session_state.twilio_client = initialize_twilio_client(twilio_account_sid_to_use, twilio_auth_token_to_use) # Use correct vars
|
425 |
+
if st.session_state.twilio_client:
|
426 |
+
st.session_state.bot_started = True
|
427 |
+
st.session_state.bot_start_time_utc = datetime.now(timezone.utc)
|
428 |
+
st.session_state.processed_message_sids = set()
|
429 |
+
st.session_state.last_twilio_poll_time = time.time() - polling_interval_to_use -1 # Use correct var
|
430 |
+
st.success("WhatsApp Bot Started!")
|
431 |
+
st.rerun()
|
432 |
+
else:
|
433 |
+
st.error("Failed to initialize Twilio client.")
|
434 |
with col4:
|
435 |
+
if st.button("🔕 Stop WhatsApp Bot", disabled=not st.session_state.bot_started, use_container_width=True):
|
436 |
+
st.session_state.bot_started = False
|
437 |
+
st.info("WhatsApp Bot Stopped.")
|
438 |
+
st.rerun()
|
439 |
st.divider()
|
440 |
|
441 |
# --- Manual Query Interface ---
|
442 |
if st.session_state.get("app_started") and st.session_state.get("rag_pipeline_ready"):
|
443 |
+
st.subheader("💬 Manual Query")
|
444 |
+
for chat_entry in st.session_state.manual_chat_history:
|
445 |
+
with st.chat_message(chat_entry["role"]):
|
446 |
+
st.markdown(chat_entry["content"])
|
447 |
+
if "context" in chat_entry and chat_entry["context"]:
|
448 |
+
with st.expander("Retrieved Context"):
|
449 |
+
try:
|
450 |
+
# Attempt to parse as JSON only if it looks like a JSON string
|
451 |
+
if isinstance(chat_entry["context"], str) and (chat_entry["context"].strip().startswith('{') or chat_entry["context"].strip().startswith('[')):
|
452 |
+
st.json(json.loads(chat_entry["context"]))
|
453 |
+
else:
|
454 |
+
# Otherwise, display as plain text
|
455 |
+
st.text(str(chat_entry["context"]))
|
456 |
+
except (json.JSONDecodeError, TypeError):
|
457 |
+
# Fallback for any other parsing errors
|
458 |
+
st.text(str(chat_entry["context"]))
|
459 |
+
|
460 |
+
user_query_manual = st.chat_input("Ask a question:")
|
461 |
+
if user_query_manual:
|
462 |
+
st.session_state.manual_chat_history.append({"role": "user", "content": user_query_manual})
|
463 |
+
with st.chat_message("user"): st.markdown(user_query_manual)
|
464 |
+
|
465 |
+
with st.spinner("Thinking..."):
|
466 |
+
intent_result = simple_intent_classifier(user_query_manual) # Get both intent and potential_id
|
467 |
+
intent = intent_result[0]
|
468 |
+
potential_oid_from_intent = intent_result[1] # This is the extracted ID if any
|
469 |
+
|
470 |
+
context_for_llm, raw_context_data = "No specific context.", None
|
471 |
+
|
472 |
+
if intent == "ORDER_STATUS":
|
473 |
+
order_id_to_check = None
|
474 |
+
if potential_oid_from_intent:
|
475 |
+
order_id_to_check = potential_oid_from_intent
|
476 |
+
else:
|
477 |
+
# Fallback for edge cases, though the regex should catch most
|
478 |
+
words = user_query_manual.upper().split()
|
479 |
+
# This regex specifically looks for 'ORD' followed by digits
|
480 |
+
possible_match = next((w for w in words if re.match(r'ORD\d+', w)), None)
|
481 |
+
if possible_match:
|
482 |
+
order_id_to_check = possible_match
|
483 |
+
|
484 |
+
|
485 |
+
if order_id_to_check:
|
486 |
+
raw_context_data = get_order_details(order_id_to_check.upper(), st.session_state.customer_orders_data)
|
487 |
+
context_for_llm = f"Order Details: {raw_context_data}"
|
488 |
+
else:
|
489 |
+
context_for_llm = "Please provide a valid Order ID (e.g., ORD1234)."
|
490 |
+
raw_context_data = {"message": "Order ID needed."}
|
491 |
+
elif intent == "PRODUCT_INFO":
|
492 |
+
raw_context_data = get_product_info(user_query_manual, st.session_state.products_data)
|
493 |
+
context_for_llm = f"Product Information: {raw_context_data}"
|
494 |
+
elif intent == "GENERAL_POLICY_FAQ" or intent == "UNKNOWN":
|
495 |
+
# ... (rest of your existing logic for these intents) ...
|
496 |
+
if st.session_state.faiss_index_pdfs and st.session_state.embedding_model:
|
497 |
+
k_val = 2 if intent == "GENERAL_POLICY_FAQ" else 1
|
498 |
+
retrieved_chunks = search_faiss_index(st.session_state.faiss_index_pdfs, user_query_manual,
|
499 |
+
st.session_state.embedding_model, st.session_state.indexed_pdf_chunks, k=k_val)
|
500 |
+
if retrieved_chunks:
|
501 |
+
context_for_llm = "\n\n".join(retrieved_chunks)
|
502 |
+
raw_context_data = retrieved_chunks
|
503 |
+
else:
|
504 |
+
context_for_llm = "No specific policy/FAQ info found." if intent == "GENERAL_POLICY_FAQ" else "Could not find relevant info."
|
505 |
+
raw_context_data = {"message": "No relevant PDF chunks found."}
|
506 |
+
else:
|
507 |
+
context_for_llm = "Policy/FAQ documents unavailable."
|
508 |
+
raw_context_data = {"message": "PDF index not ready."}
|
509 |
+
|
510 |
+
llm_response = generate_response_groq(st.session_state.groq_client, user_query_manual, context_for_llm)
|
511 |
+
with st.chat_message("assistant"):
|
512 |
+
st.markdown(llm_response)
|
513 |
+
if raw_context_data:
|
514 |
+
with st.expander("Retrieved Context"):
|
515 |
+
try:
|
516 |
+
if isinstance(raw_context_data, str) and (raw_context_data.strip().startswith('{') or raw_context_data.strip().startswith('[')):
|
517 |
+
st.json(json.loads(raw_context_data))
|
518 |
+
else:
|
519 |
+
st.text(str(raw_context_data))
|
520 |
+
except (json.JSONDecodeError, TypeError):
|
521 |
+
st.text(str(raw_context_data))
|
522 |
+
st.session_state.manual_chat_history.append({"role": "assistant", "content": llm_response, "context": raw_context_data})
|
523 |
|
524 |
# --- Twilio Bot Polling Logic ---
|
525 |
if st.session_state.get("bot_started") and st.session_state.get("rag_pipeline_ready"):
|
526 |
+
current_time = time.time()
|
527 |
+
if (current_time - st.session_state.get("last_twilio_poll_time", 0)) > polling_interval_to_use: # Use correct var
|
528 |
+
st.session_state.last_twilio_poll_time = current_time
|
529 |
+
with st.spinner("Checking WhatsApp messages..."):
|
530 |
+
if not st.session_state.get("twilio_client") or not twilio_conversation_service_sid_to_use or not twilio_bot_whatsapp_identity_to_use: # Use correct vars
|
531 |
+
st.warning("Twilio client/config missing for polling.")
|
532 |
+
else:
|
533 |
+
new_messages = get_new_whatsapp_messages(st.session_state.twilio_client, twilio_conversation_service_sid_to_use,
|
534 |
+
st.session_state.bot_start_time_utc, st.session_state.processed_message_sids,
|
535 |
+
twilio_bot_whatsapp_identity_to_use) # Use correct vars
|
536 |
+
if new_messages:
|
537 |
+
st.info(f"Found {len(new_messages)} new WhatsApp message(s).")
|
538 |
+
for msg_data in new_messages:
|
539 |
+
user_query_whatsapp, conv_sid, msg_sid, author_id = msg_data["message_body"], msg_data["conversation_sid"], msg_data["message_sid"], msg_data["author_identity"]
|
540 |
+
st.write(f"Processing from {author_id} in {conv_sid}: '{user_query_whatsapp}'")
|
541 |
+
|
542 |
+
intent_result_whatsapp = simple_intent_classifier(user_query_whatsapp) # Use the updated classifier
|
543 |
+
intent_whatsapp = intent_result_whatsapp[0]
|
544 |
+
potential_oid_whatsapp = intent_result_whatsapp[1] # Extracted ID from intent classifier
|
545 |
+
|
546 |
+
context_whatsapp = "No specific context."
|
547 |
+
if intent_whatsapp == "ORDER_STATUS":
|
548 |
order_id_to_check_whatsapp = None
|
549 |
if potential_oid_whatsapp:
|
550 |
order_id_to_check_whatsapp = potential_oid_whatsapp
|
|
|
555 |
order_id_to_check_whatsapp = possible_match_whatsapp
|
556 |
|
557 |
if order_id_to_check_whatsapp:
|
558 |
+
context_whatsapp = f"Order Details: {get_order_details(order_id_to_check_whatsapp.upper(), st.session_state.customer_orders_data)}"
|
559 |
else:
|
560 |
context_whatsapp = "Please provide a valid Order ID."
|
561 |
+
elif intent_whatsapp == "PRODUCT_INFO":
|
562 |
+
context_whatsapp = f"Product Info: {get_product_info(user_query_whatsapp, st.session_state.products_data)}"
|
563 |
+
elif intent_whatsapp == "GENERAL_POLICY_FAQ" or intent_whatsapp == "UNKNOWN":
|
564 |
+
if st.session_state.faiss_index_pdfs and st.session_state.embedding_model:
|
565 |
+
k_val = 2 if intent_whatsapp == "GENERAL_POLICY_FAQ" else 1
|
566 |
+
chunks = search_faiss_index(st.session_state.faiss_index_pdfs, user_query_whatsapp, st.session_state.embedding_model, st.session_state.indexed_pdf_chunks, k=k_val)
|
567 |
+
context_whatsapp = "\n\n".join(chunks) if chunks else ("No policy/FAQ info." if intent_whatsapp == "GENERAL_POLICY_FAQ" else "No relevant info.")
|
568 |
+
else: context_whatsapp = "Policy/FAQ docs unavailable."
|
569 |
+
|
570 |
+
response_whatsapp = generate_response_groq(st.session_state.groq_client, user_query_whatsapp, context_whatsapp)
|
571 |
+
if send_whatsapp_message(st.session_state.twilio_client, twilio_conversation_service_sid_to_use, conv_sid, response_whatsapp, twilio_bot_whatsapp_identity_to_use): # Use correct vars
|
572 |
+
st.session_state.processed_message_sids.add(msg_sid)
|
573 |
+
st.success(f"Responded to {msg_sid} from {author_id}")
|
574 |
+
else: st.error(f"Failed to send response for {msg_sid}")
|
575 |
+
st.experimental_rerun()
|
576 |
|
577 |
# --- Footer & Status ---
|
578 |
st.sidebar.markdown("---")
|
579 |
st.sidebar.info("Ensure all keys and SIDs are correctly configured. Primary API keys (Twilio SID/Token, GROQ Key) are loaded from secrets if available.")
|
580 |
if st.session_state.get("app_started"):
|
581 |
+
st.sidebar.success(f"App RUNNING. WhatsApp Bot {'RUNNING' if st.session_state.get('bot_started') else 'STOPPED'}.")
|
582 |
else:
|
583 |
+
st.sidebar.warning("App is STOPPED.")
|