File size: 51,074 Bytes
3dca110 b07af54 3dca110 b07af54 3dca110 8eb8af1 3dca110 8eb8af1 3dca110 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 | # -*- coding: utf-8 -*-
"""voc6.ipynb
Automatically generated by Colab.
Original file is located at
https://colab.research.google.com/drive/17WecCovbP3TgYvHDyZ4Yckj77r2q5Nam
"""
# Cell to add FIRST - Your Original WemaRAGSystem
import json
import re
from typing import List, Dict, Tuple
import numpy as np
import faiss
from sentence_transformers import SentenceTransformer
from dataclasses import dataclass
import pickle
import os
import io
from typing import Optional
from spitch import Spitch
import gradio as gr
# ============================================================================
# Wema Bank Voice-Enabled RAG Chatbot with Spitch Integration - CORRECTED
# ============================================================================
import tempfile
import os
import atexit
import glob
import io
from typing import Optional
from spitch import Spitch
import gradio as gr
# ============================================================================
# STEP 1: Initialize Spitch Client
# ============================================================================
class SpitchVoiceHandler:
"""
Handles all voice-related operations using Spitch API.
Supports multilingual speech-to-text and text-to-speech.
"""
def __init__(self, api_key: str):
"""
Initialize Spitch client.
Args:
api_key: Your Spitch API key
"""
self.client = Spitch(api_key=api_key)
def transcribe_audio(
self,
audio_file,
source_language: str = "en",
model: str = "mansa_v1"
) -> str:
"""
Transcribe audio to text using Spitch.
Supports multiple African and international languages.
Args:
audio_file: Audio file path or file-like object
source_language: Language code (e.g., 'en', 'yo', 'ig', 'ha')
model: Spitch model to use (default: mansa_v1)
Returns:
Transcribed text
"""
try:
print(f"π€ Transcribing audio file: {audio_file}")
# If audio_file is a path, open it
if isinstance(audio_file, str):
with open(audio_file, 'rb') as f:
response = self.client.speech.transcribe(
content=f,
language=source_language,
model=model
)
else:
# Assume it's already a file-like object (from Gradio)
response = self.client.speech.transcribe(
content=audio_file,
language=source_language,
model=model
)
print(f"Response type: {type(response)}")
# β
Spitch transcribe returns a response object with .text or json()
if hasattr(response, 'text') and callable(response.text):
# It's a method, not an attribute
transcription_text = response.text()
elif hasattr(response, 'text'):
# It's an attribute
transcription_text = response.text
elif hasattr(response, 'json'):
# Try to parse JSON response
json_data = response.json()
transcription_text = json_data.get('text', str(json_data))
else:
# Try to convert response to string
transcription_text = str(response)
print(f"β
Transcription: {transcription_text}")
return transcription_text
except Exception as e:
print(f"β Transcription error: {e}")
import traceback
traceback.print_exc()
return f"Sorry, I couldn't understand the audio. Error: {str(e)}"
def translate_to_english(self, text: str, source_lang: str = "auto") -> str:
"""
Translate text to English using Spitch translation API.
Args:
text: Text to translate
source_lang: Source language code or 'auto' for auto-detection
Returns:
Translated text in English
"""
try:
# If already in English, return as is
if source_lang == "en":
return text
translation = self.client.text.translate(
text=text,
source=source_lang,
target="en"
)
return translation.text
except Exception as e:
print(f"Translation error: {e}")
return text # Return original if translation fails
def synthesize_speech(
self,
text: str,
target_language: str = "en",
voice: str = "lina"
) -> bytes:
"""
Convert text to speech using Spitch TTS.
Args:
text: Text to convert to speech
target_language: Target language for speech
voice: Voice to use (e.g., 'lina', 'ada', 'kofi')
Returns:
Audio bytes
"""
try:
# Call Spitch TTS API
response = self.client.speech.generate(
text=text,
language=target_language,
voice=voice
)
# β
FIX: Spitch returns BinaryAPIResponse, use .read() to get bytes
if hasattr(response, 'read'):
audio_bytes = response.read()
print(f"β
TTS generated {len(audio_bytes)} bytes of audio")
return audio_bytes
else:
print(f"β Response type: {type(response)}")
print(f"β Response attributes: {dir(response)}")
return None
except Exception as e:
print(f"β TTS error: {e}")
import traceback
traceback.print_exc()
return None
# ============================================================================
# STEP 2: Integrate Voice with Your LangChain RAG System
# ============================================================================
class WemaVoiceAssistant:
"""
Complete voice-enabled assistant combining Spitch voice I/O
with your existing Wema RAG system.
"""
def __init__(
self,
rag_system,
chain,
spitch_api_key: str
):
"""
Initialize the voice assistant.
Args:
rag_system: Your initialized WemaRAGSystem
chain: Your LangChain RAG chain (already created)
spitch_api_key: Spitch API key
"""
self.rag_system = rag_system
self.voice_handler = SpitchVoiceHandler(spitch_api_key)
self.chain = chain
def process_voice_query(
self,
audio_input,
input_language: str = "en",
output_language: str = "en",
voice: str = "lina"
):
"""
Complete voice interaction pipeline:
1. Speech to text (any language)
2. Translate to English if needed
3. Query RAG system
4. Generate response
5. Translate response if needed
6. Text to speech
Args:
audio_input: Audio file from user
input_language: User's spoken language
output_language: Desired response language
voice: TTS voice to use
Returns:
tuple: (response_text, response_audio)
"""
try:
# Step 1: Transcribe audio to text
print(f"Transcribing audio in {input_language}...")
transcribed_text = self.voice_handler.transcribe_audio(
audio_input,
source_language=input_language
)
print(f"Transcribed: {transcribed_text}")
# Step 2: Translate to English if not already
if input_language != "en":
print("Translating to English...")
english_query = self.voice_handler.translate_to_english(
transcribed_text,
source_lang=input_language
)
else:
english_query = transcribed_text
print(f"English query: {english_query}")
# Step 3: Get response from RAG system (in English)
print("Querying RAG system...")
response_text = self.chain.invoke({"query": english_query})
print(f"RAG response: {response_text[:100]}...")
# Step 4: Translate response if needed
if output_language != "en":
print(f"Translating response to {output_language}...")
translation = self.voice_handler.client.text.translate(
text=response_text,
source="en",
target=output_language
)
final_text = translation.text
else:
final_text = response_text
# Step 5: Generate speech
print("Generating speech...")
audio_response = self.voice_handler.synthesize_speech(
final_text,
target_language=output_language,
voice=voice
)
return final_text, audio_response
except Exception as e:
error_msg = f"An error occurred: {str(e)}"
print(error_msg)
return error_msg, None
# ============================================================================
# STEP 3: Helper Functions for Audio File Management
# ============================================================================
def save_audio_to_temp_file(audio_bytes):
"""Save audio bytes to a temporary file and return the path."""
if audio_bytes is None:
return None
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.mp3')
temp_file.write(audio_bytes)
temp_file.close()
return temp_file.name
def cleanup_temp_audio_files():
"""Clean up temporary audio files on exit."""
temp_dir = tempfile.gettempdir()
for temp_file in glob.glob(os.path.join(temp_dir, "tmp*.mp3")):
try:
os.remove(temp_file)
except:
pass
# Register cleanup function to run on exit
atexit.register(cleanup_temp_audio_files)
# ============================================================================
# STEP 4: Create Gradio Interface (With Text AND Voice Options)
# ============================================================================
def create_voice_gradio_interface(
rag_system,
chain,
spitch_api_key: str
):
"""
Create a Gradio interface with BOTH text and voice input/output capabilities.
Args:
rag_system: Your initialized WemaRAGSystem
chain: Your LangChain RAG chain (already created)
spitch_api_key: Spitch API key
Returns:
Gradio Interface
"""
# Initialize voice assistant
assistant = WemaVoiceAssistant(rag_system, chain, spitch_api_key)
# β
CORRECT: Exact voice-language mapping from Spitch documentation
LANGUAGE_CONFIG = {
"English": {
"code": "en",
"voices": ["john", "lucy", "lina", "jude", "henry", "kani", "kingsley",
"favour", "comfort", "daniel", "remi"]
},
"Yoruba": {
"code": "yo",
"voices": ["sade", "funmi", "segun", "femi"]
},
"Igbo": {
"code": "ig",
"voices": ["obinna", "ngozi", "amara", "ebuka"]
},
"Hausa": {
"code": "ha",
"voices": ["hasan", "amina", "zainab", "aliyu"]
}
}
# Extract just language names for dropdowns
ALL_LANGUAGES = list(LANGUAGE_CONFIG.keys())
# β
FIXED: Only voices that actually exist in Spitch
# Check Spitch docs for exact voice names
VOICES = ["lina", "ada", "kofi"] # Verify these exist
def handle_text_query(text_input):
"""Handle text-only queries."""
if not text_input or text_input.strip() == "":
return "Please enter a question.", None
try:
response = chain.invoke({"query": text_input})
return response, None
except Exception as e:
return f"Error: {str(e)}", None
def update_voices(language):
"""Update voice dropdown based on selected language."""
voices = LANGUAGE_CONFIG.get(language, {}).get("voices", ["lina"])
return gr.Dropdown(choices=voices, value=voices[0])
def handle_voice_interaction(audio, input_lang, output_lang, voice):
"""Gradio handler function for voice - FIXED VERSION."""
print("="*60)
print("VOICE INTERACTION STARTED")
print(f"Audio input: {audio}")
print(f"Input language: {input_lang}")
print(f"Output language: {output_lang}")
print(f"Voice: {voice}")
print("="*60)
if audio is None:
return "Please record or upload audio.", None
# Get language codes and voices
input_config = LANGUAGE_CONFIG.get(input_lang, LANGUAGE_CONFIG["English"])
output_config = LANGUAGE_CONFIG.get(output_lang, LANGUAGE_CONFIG["English"])
input_code = input_config["code"]
output_code = output_config["code"]
# Validate voice for output language
available_voices = output_config["voices"]
if voice not in available_voices:
voice = available_voices[0]
print(f"β οΈ Voice changed to {voice} for {output_lang}")
try:
# Process voice query
print("\nπ€ Processing voice query...")
# Step 1: Transcribe (supports more languages)
transcribed_text = assistant.voice_handler.transcribe_audio(
audio,
source_language=input_code
)
print(f"π Transcribed: {transcribed_text}")
# Step 2: Translate to English if needed
if input_code != "en":
print("π Translating to English...")
english_query = assistant.voice_handler.translate_to_english(
transcribed_text,
source_lang=input_code
)
else:
english_query = transcribed_text
print(f"π¬π§ English query: {english_query}")
# Step 3: Get RAG response
print("π Querying RAG system...")
response_text = assistant.chain.invoke({"query": english_query})
print(f"β
RAG response: {response_text[:100]}...")
# Step 4: Translate response text if needed
if output_code != "en":
print(f"π Translating response to {output_lang}...")
try:
translation = assistant.voice_handler.client.text.translate(
text=response_text,
source="en",
target=output_code
)
final_text = translation.text
except Exception as e:
print(f"β οΈ Translation failed: {e}, using English")
final_text = response_text
else:
final_text = response_text
# Step 5: Generate speech in the target language with correct voice
print(f"π Generating speech in {output_lang} with voice {voice}...")
audio_bytes = assistant.voice_handler.synthesize_speech(
final_text,
target_language=output_code,
voice=voice
)
print(f"π Audio bytes type: {type(audio_bytes)}")
print(f"π Audio bytes length: {len(audio_bytes) if audio_bytes else 0}")
# β
FIX: Convert audio bytes to file path
audio_file_path = None
if audio_bytes:
print("\nπΎ Saving audio to temp file...")
audio_file_path = save_audio_to_temp_file(audio_bytes)
print(f"β
Audio saved to: {audio_file_path}")
# Verify file exists and has content
if audio_file_path and os.path.exists(audio_file_path):
file_size = os.path.getsize(audio_file_path)
print(f"β
File size: {file_size} bytes")
else:
print("β File was not created properly!")
else:
print("β No audio bytes received from TTS")
print("="*60)
return final_text, audio_file_path
except Exception as e:
error_msg = f"Error processing voice: {str(e)}"
print(f"\nβ ERROR: {error_msg}")
import traceback
traceback.print_exc()
print("="*60)
return error_msg, None
# Create Gradio interface with BOTH text and voice
with gr.Blocks(theme=gr.themes.Soft()) as demo:
gr.Markdown("""
# π¦ Wema Bank AI Assistant
### Powered by Spitch AI & LangChain RAG
Choose how you want to interact: Type or Speak!
""")
with gr.Tabs():
# TEXT TAB
with gr.Tab("π¬ Text Chat"):
gr.Markdown("### Type your banking questions")
text_input = gr.Textbox(
label="Your Question",
placeholder="Ask me anything about Wema Bank products and services...",
lines=3
)
text_submit_btn = gr.Button("π€ Send", variant="primary", size="lg")
text_output = gr.Textbox(
label="Response",
lines=10,
interactive=False
)
# Examples for text
gr.Examples(
examples=[
["What is ALAT?"],
["How do I open a savings account?"],
["Tell me about Wema Kiddies Account"],
["How can I avoid phishing scams?"],
["What loans does Wema Bank offer?"]
],
inputs=text_input,
label="π‘ Try these questions"
)
text_submit_btn.click(
fn=handle_text_query,
inputs=text_input,
outputs=[text_output, gr.Audio(visible=False)]
)
# Also submit on Enter
text_input.submit(
fn=handle_text_query,
inputs=text_input,
outputs=[text_output, gr.Audio(visible=False)]
)
# VOICE TAB
with gr.Tab("π€ Voice Chat"):
gr.Markdown("""
### Speak your banking questions in your language!
**β
Fully Supported Nigerian Languages:**
- π¬π§ **English** - 11 voices available
- π³π¬ **Yoruba** - 4 voices (Sade, Funmi, Segun, Femi)
- π³π¬ **Igbo** - 4 voices (Obinna, Ngozi, Amara, Ebuka)
- π³π¬ **Hausa** - 4 voices (Hasan, Amina, Zainab, Aliyu)
Speak naturally and get responses in both text and audio in your preferred language!
""")
with gr.Row():
with gr.Column():
audio_input = gr.Audio(
sources=["microphone", "upload"],
type="filepath",
label="ποΈ Record or Upload Audio"
)
input_language = gr.Dropdown(
choices=ALL_LANGUAGES,
value="English",
label="Your Language (Speech Input)"
)
with gr.Column():
output_language = gr.Dropdown(
choices=ALL_LANGUAGES,
value="English",
label="Response Language (Audio Output)"
)
voice_selection = gr.Dropdown(
choices=LANGUAGE_CONFIG["English"]["voices"],
value="lina",
label="Voice"
)
# Update voices when output language changes
output_language.change(
fn=update_voices,
inputs=output_language,
outputs=voice_selection
)
voice_submit_btn = gr.Button("π Ask Wema Assist", variant="primary", size="lg")
voice_text_output = gr.Textbox(
label="π Text Response",
lines=8,
interactive=False
)
voice_audio_output = gr.Audio(
label="π Audio Response",
type="filepath" # β
Important: must be filepath
)
voice_submit_btn.click(
fn=handle_voice_interaction,
inputs=[audio_input, input_language, output_language, voice_selection],
outputs=[voice_text_output, voice_audio_output]
)
gr.Markdown("""
---
### π Features
- **Text Chat**: Fast and simple - just type and get instant responses
- **Voice Chat**: Full support for Nigerian languages!
### π³π¬ Supported Nigerian Languages
β
**English** - 11 different voices (male & female)
β
**Yoruba** - E ku α»jα»! (4 authentic Yoruba voices)
β
**Igbo** - Nnα»α»! (4 authentic Igbo voices)
β
**Hausa** - Sannu! (4 authentic Hausa voices)
π‘ **All features work in every language:**
- π€ Speak your question in your language
- π Get text response translated
- π Hear authentic audio response in your language
- π Seamless translation between languages
""")
return demo
# ============================================================================
# ALTERNATIVE: Simpler Hybrid Interface
# ============================================================================
def create_hybrid_interface(
rag_system,
chain,
spitch_api_key: str
):
"""
Creates a simpler interface supporting both text and voice input.
Args:
rag_system: Your initialized WemaRAGSystem
chain: Your LangChain RAG chain (already created)
spitch_api_key: Spitch API key
Returns:
Gradio Interface
"""
assistant = WemaVoiceAssistant(rag_system, chain, spitch_api_key)
def handle_text_query(text_input):
"""Handle text-only query."""
try:
response = chain.invoke({"query": text_input})
return response, None
except Exception as e:
return f"Error: {str(e)}", None
def handle_voice_query(audio, input_lang, output_lang, voice):
"""Handle voice query."""
if audio is None:
return "Please provide audio input.", None
LANGUAGES = {
"English": "en",
"Yoruba": "yo",
"Igbo": "ig",
"Hausa": "ha"
}
input_code = LANGUAGES.get(input_lang, "en")
output_code = LANGUAGES.get(output_lang, "en")
# Process voice query
text_response, audio_bytes = assistant.process_voice_query(
audio,
input_language=input_code,
output_language=output_code,
voice=voice
)
# Convert audio bytes to file path
audio_file_path = None
if audio_bytes:
audio_file_path = save_audio_to_temp_file(audio_bytes)
return text_response, audio_file_path
# Create tabbed interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
gr.Markdown("# π¦ Wema Bank AI Assistant")
with gr.Tabs():
# Text Tab
with gr.Tab("π¬ Text Chat"):
text_input = gr.Textbox(
label="Type your question",
placeholder="Ask about Wema Bank products and services..."
)
text_submit = gr.Button("Send")
text_output = gr.Textbox(label="Response", lines=10)
text_submit.click(
fn=handle_text_query,
inputs=text_input,
outputs=[text_output, gr.Audio(visible=False)]
)
# Voice Tab
with gr.Tab("π€ Voice Chat"):
audio_input = gr.Audio(sources=["microphone", "upload"], type="filepath")
with gr.Row():
input_lang = gr.Dropdown(
["English", "Yoruba", "Igbo", "Hausa"],
value="English",
label="Input Language"
)
output_lang = gr.Dropdown(
["English", "Yoruba", "Igbo", "Hausa"],
value="English",
label="Output Language"
)
voice = gr.Dropdown(
["lina", "ada", "kofi"],
value="lina",
label="Voice"
)
voice_submit = gr.Button("Ask")
voice_text_output = gr.Textbox(label="Response Text", lines=8)
voice_audio_output = gr.Audio(label="Audio Response", type="filepath")
voice_submit.click(
fn=handle_voice_query,
inputs=[audio_input, input_lang, output_lang, voice],
outputs=[voice_text_output, voice_audio_output]
)
return demo
@dataclass
class DocumentChunk:
"""Represents a chunk of text with metadata."""
text: str
metadata: Dict
chunk_id: int
class WemaDocumentChunker:
"""Handles intelligent chunking of Wema Bank documents."""
def __init__(self, chunk_size: int = 800, overlap: int = 150):
"""
Initialize the chunker.
Args:
chunk_size: Target size for each chunk in characters
overlap: Number of characters to overlap between chunks
"""
self.chunk_size = chunk_size
self.overlap = overlap
def identify_sections(self, text: str) -> List[Tuple[str, str]]:
"""
Identify logical sections in the document.
Returns:
List of tuples (section_title, section_content)
"""
sections = []
# Common section headers in banking documents
section_patterns = [
r'(Avoiding Financial and Phishing Scams)',
r'(Keeping Your Card.*?Safe)',
r'(E-mails and calls from.*?)',
r'(Scam Alert Tips)',
r'(Guard Yourself)',
r'(Bank Verification Number)',
r'(Personal Banking)',
r'(Business Banking)',
r'(Corporate Banking)',
r'(.*?Account)',
r'(.*?Loan.*?)',
]
# Try to split by recognizable headers
combined_pattern = '|'.join(section_patterns)
matches = list(re.finditer(combined_pattern, text, re.IGNORECASE))
if matches:
for i, match in enumerate(matches):
start = match.start()
end = matches[i + 1].start() if i + 1 < len(matches) else len(text)
section_title = match.group(0).strip()
section_content = text[start:end].strip()
sections.append((section_title, section_content))
else:
# If no clear sections, treat as one section
sections.append(("General Content", text))
return sections
def chunk_text(self, text: str, metadata: Dict) -> List[DocumentChunk]:
"""
Chunk text with semantic awareness and overlap.
Args:
text: Text to chunk
metadata: Metadata to attach to chunks
Returns:
List of DocumentChunk objects
"""
chunks = []
# First, try to identify sections
sections = self.identify_sections(text)
chunk_id = 0
for section_title, section_content in sections:
# If section is smaller than chunk_size, keep it whole
if len(section_content) <= self.chunk_size:
chunk_metadata = metadata.copy()
chunk_metadata['section'] = section_title
chunks.append(DocumentChunk(
text=section_content,
metadata=chunk_metadata,
chunk_id=chunk_id
))
chunk_id += 1
else:
# Split section into smaller chunks with overlap
sentences = self._split_into_sentences(section_content)
current_chunk = []
current_length = 0
for sentence in sentences:
sentence_length = len(sentence)
if current_length + sentence_length > self.chunk_size and current_chunk:
# Save current chunk
chunk_text = ' '.join(current_chunk)
chunk_metadata = metadata.copy()
chunk_metadata['section'] = section_title
chunks.append(DocumentChunk(
text=chunk_text,
metadata=chunk_metadata,
chunk_id=chunk_id
))
chunk_id += 1
# Keep overlap sentences for next chunk
overlap_text = chunk_text[-self.overlap:] if len(chunk_text) > self.overlap else chunk_text
overlap_sentences = self._split_into_sentences(overlap_text)
current_chunk = overlap_sentences
current_length = sum(len(s) for s in current_chunk)
current_chunk.append(sentence)
current_length += sentence_length
# Add remaining chunk
if current_chunk:
chunk_metadata = metadata.copy()
chunk_metadata['section'] = section_title
chunks.append(DocumentChunk(
text=' '.join(current_chunk),
metadata=chunk_metadata,
chunk_id=chunk_id
))
chunk_id += 1
return chunks
def _split_into_sentences(self, text: str) -> List[str]:
"""Split text into sentences."""
# Simple sentence splitter
sentences = re.split(r'(?<=[.!?])\s+', text)
return [s.strip() for s in sentences if s.strip()]
class WemaRAGSystem:
"""Complete RAG system for Wema Bank documents."""
def __init__(self, model_name: str = 'sentence-transformers/all-MiniLM-L6-v2'):
"""
Initialize the RAG system.
Args:
model_name: Name of the sentence transformer model to use
"""
print(f"Loading embedding model: {model_name}")
self.model = SentenceTransformer(model_name)
self.dimension = self.model.get_sentence_embedding_dimension()
self.index = None
self.chunks = []
self.chunker = WemaDocumentChunker()
def load_and_process_document(self, json_path: str):
"""
Load JSON document, chunk it, and create embeddings.
Args:
json_path: Path to the JSON file
"""
print(f"Loading document from: {json_path}")
with open(json_path, 'r', encoding='utf-8') as f:
data = json.load(f)
# Process each document in the JSON
all_chunks = []
if isinstance(data, list):
documents = data
elif isinstance(data, dict):
documents = [data]
else:
raise ValueError("JSON must contain a document object or list of documents")
for doc in documents:
text = doc.get('text', '')
metadata = {
'url': doc.get('url', ''),
'title': doc.get('title', ''),
'meta_description': doc.get('meta_description', '')
}
# Chunk the document
chunks = self.chunker.chunk_text(text, metadata)
all_chunks.extend(chunks)
print(f"Created {len(chunks)} chunks from document: {metadata['title'][:50]}...")
self.chunks = all_chunks
print(f"Total chunks created: {len(self.chunks)}")
# Generate embeddings
self._create_embeddings()
def _create_embeddings(self):
"""Generate embeddings for all chunks and create FAISS index."""
print("Generating embeddings...")
texts = [chunk.text for chunk in self.chunks]
embeddings = self.model.encode(texts, show_progress_bar=True)
# Create FAISS index
print("Creating FAISS index...")
self.index = faiss.IndexFlatL2(self.dimension)
self.index.add(embeddings.astype('float32'))
print(f"FAISS index created with {self.index.ntotal} vectors")
def save(self, index_path: str = 'wema_faiss.index',
chunks_path: str = 'wema_chunks.pkl'):
"""
Save FAISS index and chunks to disk.
Args:
index_path: Path to save FAISS index
chunks_path: Path to save chunks metadata
"""
if self.index is None:
raise ValueError("No index to save. Process documents first.")
print(f"Saving FAISS index to: {index_path}")
faiss.write_index(self.index, index_path)
print(f"Saving chunks metadata to: {chunks_path}")
with open(chunks_path, 'wb') as f:
pickle.dump(self.chunks, f)
print("Save complete!")
def load(self, index_path: str = 'wema_faiss.index',
chunks_path: str = 'wema_chunks.pkl'):
"""
Load FAISS index and chunks from disk.
Args:
index_path: Path to FAISS index
chunks_path: Path to chunks metadata
"""
print(f"Loading FAISS index from: {index_path}")
self.index = faiss.read_index(index_path)
print(f"Loading chunks metadata from: {chunks_path}")
with open(chunks_path, 'rb') as f:
self.chunks = pickle.load(f)
print(f"Loaded {len(self.chunks)} chunks with index size {self.index.ntotal}")
def search(self, query: str, top_k: int = 5) -> List[Dict]:
"""
Search for relevant chunks given a query.
Args:
query: Search query
top_k: Number of results to return
Returns:
List of dictionaries containing chunk text, metadata, and similarity score
"""
if self.index is None:
raise ValueError("No index loaded. Load or create an index first.")
# Encode query
query_embedding = self.model.encode([query])[0].astype('float32').reshape(1, -1)
# Search
distances, indices = self.index.search(query_embedding, top_k)
# Prepare results
results = []
for i, idx in enumerate(indices[0]):
chunk = self.chunks[idx]
results.append({
'text': chunk.text,
'metadata': chunk.metadata,
'score': float(distances[0][i]),
'chunk_id': chunk.chunk_id
})
return results
def get_context_for_rag(self, query: str, top_k: int = 3,
max_context_length: int = 2000) -> str:
"""
Get formatted context for RAG applications.
Args:
query: Search query
top_k: Number of chunks to retrieve
max_context_length: Maximum length of context to return
Returns:
Formatted context string
"""
results = self.search(query, top_k)
context_parts = []
current_length = 0
for i, result in enumerate(results, 1):
chunk_text = result['text']
section = result['metadata'].get('section', 'N/A')
# Format context with source information
formatted = f"[Source {i} - {section}]\n{chunk_text}\n"
if current_length + len(formatted) > max_context_length:
break
context_parts.append(formatted)
current_length += len(formatted)
return "\n".join(context_parts)
from langchain_core.runnables import RunnablePassthrough, RunnableParallel, RunnableLambda
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_google_genai import ChatGoogleGenerativeAI
import gradio as gr
from typing import Dict, Any, List
import json
class WemaDocumentProcessorRunnable:
"""
Wraps the document loading, chunking, embedding, and storing as a LangChain Runnable.
This preserves ALL the original WemaRAGSystem functionality.
"""
def __init__(self, rag_system):
"""
Initialize with a WemaRAGSystem instance.
Args:
rag_system: An initialized WemaRAGSystem object
"""
self.rag = rag_system
# Create runnables for each step
self.document_loader = RunnableLambda(self._load_document)
self.chunker = RunnableLambda(self._chunk_documents)
self.embedder = RunnableLambda(self._create_embeddings)
self.storer = RunnableLambda(self._store_index)
# Complete pipeline runnable
self.full_pipeline = (
self.document_loader
| self.chunker
| self.embedder
| self.storer
)
def _load_document(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Loads JSON document(s).
Args:
inputs: Dictionary with 'json_path' key
Returns:
Dictionary with loaded documents
"""
json_path = inputs.get("json_path", inputs) if isinstance(inputs, dict) else inputs
print(f"Loading document from: {json_path}")
with open(json_path, 'r', encoding='utf-8') as f:
data = json.load(f)
# Process documents
if isinstance(data, list):
documents = data
elif isinstance(data, dict):
documents = [data]
else:
raise ValueError("JSON must contain a document object or list of documents")
return {
"json_path": json_path,
"documents": documents,
"status": "loaded"
}
def _chunk_documents(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Chunks documents using WemaDocumentChunker.
Args:
inputs: Dictionary with 'documents' key
Returns:
Dictionary with chunked documents
"""
documents = inputs["documents"]
print("Chunking documents...")
all_chunks = []
for doc in documents:
text = doc.get('text', '')
metadata = {
'url': doc.get('url', ''),
'title': doc.get('title', ''),
'meta_description': doc.get('meta_description', '')
}
# Use the original chunker from WemaRAGSystem
chunks = self.rag.chunker.chunk_text(text, metadata)
all_chunks.extend(chunks)
print(f"Created {len(chunks)} chunks from document: {metadata['title'][:50]}...")
self.rag.chunks = all_chunks
print(f"Total chunks created: {len(self.rag.chunks)}")
return {
"json_path": inputs.get("json_path"),
"documents": documents,
"chunks": all_chunks,
"chunk_count": len(all_chunks),
"status": "chunked"
}
def _create_embeddings(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Creates embeddings and FAISS index using the original method.
Args:
inputs: Dictionary with 'chunks' key
Returns:
Dictionary with embedding info
"""
print("Generating embeddings...")
# Use the original _create_embeddings method
self.rag._create_embeddings()
return {
"json_path": inputs.get("json_path"),
"documents": inputs["documents"],
"chunks": inputs["chunks"],
"chunk_count": inputs["chunk_count"],
"index_size": self.rag.index.ntotal,
"status": "embedded"
}
def _store_index(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Saves FAISS index and chunks to disk.
Args:
inputs: Dictionary with processing results
Returns:
Dictionary with save status
"""
index_path = inputs.get("index_path", "wema_faiss.index")
chunks_path = inputs.get("chunks_path", "wema_chunks.pkl")
# Use the original save method
self.rag.save(index_path=index_path, chunks_path=chunks_path)
return {
"json_path": inputs.get("json_path"),
"chunk_count": inputs["chunk_count"],
"index_size": inputs["index_size"],
"index_path": index_path,
"chunks_path": chunks_path,
"status": "saved"
}
def get_full_pipeline(self):
"""Returns the complete processing pipeline as a LangChain Runnable."""
return self.full_pipeline
def get_loader_runnable(self):
"""Returns just the document loader."""
return self.document_loader
def get_chunker_runnable(self):
"""Returns just the chunker."""
return self.chunker
def get_embedder_runnable(self):
"""Returns just the embedder."""
return self.embedder
def get_storer_runnable(self):
"""Returns just the storer."""
return self.storer
class WemaRAGRetrieverRunnable:
"""
Wraps the retrieval functionality as a LangChain Runnable.
"""
def __init__(self, rag_system):
"""
Initialize with an existing WemaRAGSystem instance.
Args:
rag_system: An initialized WemaRAGSystem object
"""
self.rag = rag_system
self.retriever = RunnableLambda(self._retrieve_context)
def _retrieve_context(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Retrieves context from the RAG system using the original search method.
Args:
inputs: Dictionary containing 'query' key
Returns:
Dictionary with query and context
"""
query = inputs.get("query", inputs) if isinstance(inputs, dict) else inputs
# Use the original get_context_for_rag method
context = self.rag.get_context_for_rag(query, top_k=3)
return {
"query": query,
"context": context
}
def get_retriever_runnable(self):
"""Returns the retriever as a LangChain Runnable."""
return self.retriever
class WemaRAGLoaderRunnable:
"""
Wraps the loading functionality as a LangChain Runnable.
"""
def __init__(self, rag_system):
"""
Initialize with a WemaRAGSystem instance.
Args:
rag_system: An initialized WemaRAGSystem object
"""
self.rag = rag_system
self.loader = RunnableLambda(self._load_index)
def _load_index(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
"""
Loads FAISS index and chunks from disk using the original method.
Args:
inputs: Dictionary with 'index_path' and 'chunks_path' keys
Returns:
Dictionary with load status
"""
index_path = inputs.get("index_path", "wema_faiss.index")
chunks_path = inputs.get("chunks_path", "wema_chunks.pkl")
# Use the original load method
self.rag.load(index_path=index_path, chunks_path=chunks_path)
return {
"index_path": index_path,
"chunks_path": chunks_path,
"chunk_count": len(self.rag.chunks),
"index_size": self.rag.index.ntotal,
"status": "loaded"
}
def get_loader_runnable(self):
"""Returns the loader as a LangChain Runnable."""
return self.loader
def create_wema_rag_chain(rag_system, google_api_key: str):
"""
Creates a complete LangChain RAG chain using the WemaRAGSystem.
Args:
rag_system: An initialized WemaRAGSystem object
google_api_key: Google API key for Gemini
Returns:
A LangChain runnable chain
"""
# Wrap the RAG retriever as a runnable
wema_retriever = WemaRAGRetrieverRunnable(rag_system)
# Initialize the LLM
llm = ChatGoogleGenerativeAI(
model="gemini-2.0-flash-exp",
google_api_key=google_api_key,
temperature=0.7
)
# Create the prompt template
prompt = ChatPromptTemplate.from_template("""
You are **Wema Assist**, an intelligent, reliable, and customer-friendly virtual assistant for **Wema Bank**.
Your goal is to provide accurate, thoughtful, and helpful responses to users β just like a top-tier banking representative β but with the warmth and empathy of a human advisor.
You have access to verified Wema Bank documentation provided below as *Context*.
Your tone should be:
- professional yet conversational,
- confident but not pushy,
- and focused on solving the user's need while representing Wema Bank's products effectively.
---
### π― Core Guidelines:
1. **Use the provided "Context" first.**
- It contains factual information from official Wema Bank documents.
- Do not invent product details β rely on the context where possible.
2. **Always connect answers to Wema Bank offerings.**
- When relevant, mention products or services such as:
- Wema Savings Account
- ALAT Digital Bank or ALAT Savings Goals
- Wema Kiddies Account
- Business or SME Banking
- Wema Loans
- Wema Security Tips or Scam Alerts
- Even if the user query seems general, highlight any Wema product that could help.
3. **Be natural and practical.**
- Offer useful, step-by-step guidance.
- Use phrasing like:
- "At Wema Bank, you can..."
- "A good option through Wema is..."
- "Wema's ALAT platform allows you to..."
4. **If the context isn't related to the query:**
- Simply give a general, thoughtful answer β *without apologizing or saying the context is irrelevant.*
---
### π Information You Have:
**Context:**
{context}
**User Query:**
{query}
---
### π§ Task:
Answer the query in a complete, natural, and customer-friendly way β integrating Wema Bank products or services wherever relevant.
If the RAG and context are not related, just give a general answer and don't complain.
### π¬ Final Answer:
""")
# Build the chain using LCEL (LangChain Expression Language)
chain = (
RunnablePassthrough()
| wema_retriever.get_retriever_runnable()
| prompt
| llm
| StrOutputParser()
)
return chain
def create_gradio_interface(rag_system, google_api_key: str):
"""
Creates a Gradio interface using the LangChain RAG chain.
Args:
rag_system: An initialized WemaRAGSystem object
google_api_key: Google API key for Gemini
Returns:
Gradio Interface object
"""
# Create the LangChain chain
chain = create_wema_rag_chain(rag_system, google_api_key)
def chat_function(query: str) -> str:
"""Wrapper function for Gradio."""
try:
response = chain.invoke({"query": query})
return response
except Exception as e:
return f"An error occurred: {str(e)}"
# Create Gradio interface
iface = gr.Interface(
fn=chat_function,
inputs=gr.Textbox(
label="Enter your query about Wema Bank:",
placeholder="Ask me anything about Wema Bank products and services..."
),
outputs=gr.Textbox(
label="Wema Assist Response:",
lines=10
),
title="π¦ Wema Bank RAG Chatbot (LangChain Edition)",
description="Powered by LangChain and your custom Wema RAG System",
theme="soft"
)
return iface
# Initialize RAG system
rag = WemaRAGSystem()
# Wrap it as a LangChain runnable
processor = WemaDocumentProcessorRunnable(rag)
# Cell 3: Run the complete pipeline (load β chunk β embed β store)
result = processor.get_full_pipeline().invoke({
"json_path": "wema_cleaned.json",
"index_path": "wema_faiss.index",
"chunks_path": "wema_chunks.pkl"
})
print(f"Processing complete!")
print(f"Chunks created: {result['chunk_count']}")
print(f"Index size: {result['index_size']}")
print(f"Saved to: {result['index_path']}")
# Assuming you have an instance of WemaRAGSystem called 'rag'
#rag = WemaRAGSystem()
# Replace 'your_document.json' with the actual path to your file
#rag.load_and_process_document("your_document.json")
"""
# Cell 4: Create and launch Gradio interface
from google.colab import userdata
GOOGLE_API_KEY = userdata.get('GOOGLE_API_KEY')
iface = create_gradio_interface(rag, GOOGLE_API_KEY)
iface.launch()
"""
'''
# Cell 2: Set up your RAG system (your existing code)
rag = WemaRAGSystem()
rag.load() # Load your existing index
# Cell 3: Initialize API keys
from google.colab import userdata
GOOGLE_API_KEY = userdata.get('GOOGLE_API_KEY')
SPITCH_API_KEY = userdata.get('SPITCH_API_KEY') # Add this to your Colab secrets
# Cell 4: Launch voice interface
iface = create_voice_gradio_interface(
rag_system=rag,
google_api_key=GOOGLE_API_KEY,
spitch_api_key=SPITCH_API_KEY
)
iface.launch(share=True)
'''
# Cell 2: Set up your RAG system (your existing code)
rag = WemaRAGSystem()
rag.load() # Load your existing index
# Cell 3: Initialize API keys
import os
GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
SPITCH_API_KEY = os.getenv("SPITCH_API_KEY")
if not GOOGLE_API_KEY or not SPITCH_API_KEY:
raise ValueError("Missing one or more API keys. Make sure they are added as secrets in your Space.")
# Cell 4: Launch voice interface
# The create_voice_gradio_interface function needs the chain, not the google_api_key directly.
# We need to create the chain first.
chain = create_wema_rag_chain(rag, GOOGLE_API_KEY)
iface = create_voice_gradio_interface(
rag_system=rag,
chain=chain, # Pass the created chain
spitch_api_key=SPITCH_API_KEY
)
iface.launch(share=True, debug=True)
|