Kennethdotse commited on
Commit
891708a
Β·
1 Parent(s): de827b0

made some changes

Browse files
__pycache__/api.cpython-313.pyc ADDED
Binary file (12.7 kB). View file
 
__pycache__/main.cpython-313.pyc ADDED
Binary file (374 Bytes). View file
 
api.py CHANGED
@@ -16,6 +16,7 @@ from fastapi.staticfiles import StaticFiles
16
  from huggingface_hub import HfApi, create_repo, upload_file
17
  from transformers import WhisperForConditionalGeneration, WhisperProcessor
18
 
 
19
  # ── Env ────────────────────────────────────────────────────────────────────
20
  load_dotenv()
21
 
@@ -279,4 +280,5 @@ async def dataset_entries(limit: int = 50, offset: int = 0):
279
 
280
 
281
  # ── Static files ───────────────────────────────────────────────────────────
282
- app.mount("/static", StaticFiles(directory=str(STATIC_DIR)), name="static")
 
 
16
  from huggingface_hub import HfApi, create_repo, upload_file
17
  from transformers import WhisperForConditionalGeneration, WhisperProcessor
18
 
19
+
20
  # ── Env ────────────────────────────────────────────────────────────────────
21
  load_dotenv()
22
 
 
280
 
281
 
282
  # ── Static files ───────────────────────────────────────────────────────────
283
+ app.mount("/static", StaticFiles(directory=str(STATIC_DIR)), name="static")
284
+ app.mount("/assets", StaticFiles(directory="assets"), name="assets")
app.py DELETED
@@ -1,701 +0,0 @@
1
- import os
2
- import uuid
3
- import shutil
4
- import numpy as np
5
- import pandas as pd
6
- import soundfile as sf
7
- import torch
8
- import torchaudio
9
- import gradio as gr
10
-
11
- import threading
12
- from dotenv import load_dotenv
13
- from huggingface_hub import HfApi, create_repo, upload_file
14
- from transformers import WhisperForConditionalGeneration, WhisperProcessor
15
-
16
- # =========================================================
17
- # LOAD ENV VARIABLES
18
- # =========================================================
19
-
20
- load_dotenv()
21
-
22
- HF_TOKEN = os.getenv("HF_TOKEN")
23
- DATASET_REPO = os.getenv("HF_DATASET_REPO")
24
-
25
- # =========================================================
26
- # LOCAL STORAGE
27
- # =========================================================
28
-
29
- LOCAL_DATASET_DIR = "hf_dataset"
30
- LOCAL_AUDIO_DIR = os.path.join(LOCAL_DATASET_DIR, "audio")
31
- LOCAL_METADATA = os.path.join(LOCAL_DATASET_DIR, "metadata.csv")
32
-
33
- os.makedirs(LOCAL_AUDIO_DIR, exist_ok=True)
34
-
35
- # =========================================================
36
- # HUGGING FACE SETUP
37
- # =========================================================
38
-
39
- api = HfApi(token=HF_TOKEN)
40
-
41
- try:
42
- create_repo(
43
- repo_id=DATASET_REPO,
44
- repo_type="dataset",
45
- exist_ok=True,
46
- token=HF_TOKEN
47
- )
48
- except Exception as e:
49
- print("Dataset repo check:", e)
50
-
51
- # =========================================================
52
- # LOAD MODEL
53
- # =========================================================
54
-
55
- model = WhisperForConditionalGeneration.from_pretrained(
56
- "Kennethdot/kasanoma_whisper"
57
- )
58
-
59
- processor = WhisperProcessor.from_pretrained(
60
- "Kennethdot/kasanoma_whisper"
61
- )
62
-
63
- device = torch.device(
64
- "cuda" if torch.cuda.is_available() else "cpu"
65
- )
66
-
67
- model = model.to(device)
68
- model.eval()
69
-
70
- # =========================================================
71
- # TRANSCRIPTION FUNCTION
72
- # =========================================================
73
-
74
- def transcribe_audio(audio_path):
75
-
76
- if audio_path is None:
77
- return "", "", None
78
-
79
- try:
80
- audio_data, sampling_rate = sf.read(audio_path)
81
-
82
- # Stereo β†’ mono
83
- if len(audio_data.shape) > 1:
84
- audio_data = np.mean(audio_data, axis=1)
85
-
86
- audio_data = audio_data.astype(np.float32)
87
-
88
- # Resample
89
- if sampling_rate != 16000:
90
- audio_tensor = torch.tensor(
91
- audio_data,
92
- dtype=torch.float32
93
- )
94
-
95
- resampler = torchaudio.transforms.Resample(
96
- orig_freq=sampling_rate,
97
- new_freq=16000
98
- )
99
-
100
- audio_data = resampler(audio_tensor).numpy()
101
- sampling_rate = 16000
102
-
103
- except Exception as e:
104
- return f"Error reading audio: {e}", "", None
105
-
106
- # Normalize
107
- if np.max(np.abs(audio_data)) > 0:
108
- audio_data = audio_data / np.max(np.abs(audio_data))
109
-
110
- # Feature extraction
111
- input_features = processor.feature_extractor(
112
- audio_data,
113
- sampling_rate=sampling_rate,
114
- return_tensors="pt"
115
- ).input_features.to(device)
116
-
117
- # Inference
118
- with torch.no_grad():
119
-
120
- generated_ids = model.generate(
121
- input_features,
122
- task="transcribe",
123
- language="yo",
124
- temperature=0.0
125
- )
126
-
127
- transcription = processor.batch_decode(
128
- generated_ids,
129
- skip_special_tokens=True
130
- )[0].strip()
131
-
132
- return transcription, transcription, audio_path
133
-
134
- # =========================================================
135
- # SAVE FUNCTION
136
- # =========================================================
137
-
138
- # Lock prevents concurrent saves from corrupting the CSV
139
- _csv_lock = threading.Lock()
140
-
141
-
142
- def _upload_in_background(saved_audio_path, relative_audio_path):
143
- """Runs in a daemon thread. Uploads audio then CSV β€” never blocks the UI."""
144
- try:
145
- upload_file(
146
- path_or_fileobj=saved_audio_path,
147
- path_in_repo=relative_audio_path,
148
- repo_id=DATASET_REPO,
149
- repo_type="dataset",
150
- token=HF_TOKEN
151
- )
152
- with _csv_lock:
153
- upload_file(
154
- path_or_fileobj=LOCAL_METADATA,
155
- path_in_repo="metadata.csv",
156
- repo_id=DATASET_REPO,
157
- repo_type="dataset",
158
- token=HF_TOKEN
159
- )
160
- except Exception as e:
161
- print(f"[Background upload error] {e}")
162
-
163
-
164
- def save_sample(audio_path, corrected_text):
165
-
166
- if audio_path is None:
167
- return
168
-
169
- try:
170
- unique_id = str(uuid.uuid4())
171
-
172
- saved_audio_path = os.path.join(
173
- LOCAL_AUDIO_DIR,
174
- f"{unique_id}.wav"
175
- )
176
-
177
- # Copy audio locally β€” fast, no network
178
- shutil.copy(audio_path, saved_audio_path)
179
-
180
- relative_audio_path = f"audio/{unique_id}.wav"
181
-
182
- # Write metadata locally β€” fast, no network
183
- new_row = pd.DataFrame([{
184
- "id": unique_id,
185
- "audio": relative_audio_path,
186
- "transcription": corrected_text,
187
- "language": "twi_en"
188
- }])
189
-
190
- with _csv_lock:
191
- if os.path.exists(LOCAL_METADATA):
192
- existing = pd.read_csv(LOCAL_METADATA)
193
- updated = pd.concat([existing, new_row], ignore_index=True)
194
- else:
195
- updated = new_row
196
- updated.to_csv(LOCAL_METADATA, index=False)
197
-
198
- # Fire-and-forget β€” HuggingFace upload happens in background
199
- threading.Thread(
200
- target=_upload_in_background,
201
- args=(saved_audio_path, relative_audio_path),
202
- daemon=True
203
- ).start()
204
-
205
- return
206
-
207
- except Exception as e:
208
- print(f"[Save error] {e}")
209
- return
210
-
211
- # =========================================================
212
- # CUSTOM CSS (from the styled version)
213
- # =========================================================
214
-
215
- css = """
216
- @import url('https://fonts.googleapis.com/css2?family=Sora:wght@300;400;600;700&family=DM+Mono:wght@400;500&display=swap');
217
-
218
- :root {
219
- --navy: #0d1b4b;
220
- --blue: #1a6fd4;
221
- --sky: #4db8f0;
222
- --violet: #7b4fd4;
223
- --lilac: #b57bee;
224
- --white: #e8f0ff;
225
- --card: rgba(255,255,255,0.07);
226
- --border: rgba(180,160,255,0.2);
227
- --radius: 16px;
228
- }
229
-
230
- /* ── animated mesh background ── */
231
- body, .gradio-container {
232
- background:
233
- radial-gradient(ellipse at 15% 10%, #1a3fa8 0%, transparent 55%),
234
- radial-gradient(ellipse at 85% 5%, #7b3fc4 0%, transparent 45%),
235
- radial-gradient(ellipse at 50% 50%, #d0e8ff 0%, transparent 60%),
236
- radial-gradient(ellipse at 80% 80%, #b57bee 0%, transparent 50%),
237
- radial-gradient(ellipse at 10% 90%, #1a6fd4 0%, transparent 50%),
238
- #0d1b4b !important;
239
- font-family: 'Sora', sans-serif !important;
240
- min-height: 100vh;
241
- }
242
-
243
- /* ── slow drifting orbs ── */
244
- .gradio-container::before,
245
- .gradio-container::after {
246
- content: '';
247
- position: fixed;
248
- border-radius: 50%;
249
- filter: blur(80px);
250
- pointer-events: none;
251
- z-index: 0;
252
- }
253
- .gradio-container::before {
254
- width: 520px; height: 520px;
255
- top: -120px; left: -100px;
256
- background: radial-gradient(circle, rgba(74,130,230,0.45), transparent 70%);
257
- animation: drift1 12s ease-in-out infinite alternate;
258
- }
259
- .gradio-container::after {
260
- width: 480px; height: 480px;
261
- bottom: -100px; right: -80px;
262
- background: radial-gradient(circle, rgba(160,100,240,0.4), transparent 70%);
263
- animation: drift2 15s ease-in-out infinite alternate;
264
- }
265
- @keyframes drift1 {
266
- from { transform: translate(0, 0); }
267
- to { transform: translate(60px, 80px); }
268
- }
269
- @keyframes drift2 {
270
- from { transform: translate(0, 0); }
271
- to { transform: translate(-50px, -70px); }
272
- }
273
-
274
- /* ── hero ── */
275
- #hero {
276
- text-align: center;
277
- padding: 40px 24px 16px;
278
- position: relative;
279
- z-index: 1;
280
- animation: fadeUp 0.8s ease both;
281
- }
282
- @keyframes fadeUp {
283
- from { opacity: 0; transform: translateY(24px); }
284
- to { opacity: 1; transform: translateY(0); }
285
- }
286
-
287
- /* ── animated orb ── */
288
- .orb-wrap {
289
- display: flex;
290
- justify-content: center;
291
- margin-bottom: 22px;
292
- }
293
- .orb-stage {
294
- position: relative;
295
- width: 110px;
296
- height: 110px;
297
- }
298
- .orb-ring {
299
- position: absolute;
300
- border-radius: 50%;
301
- border: 1.5px solid transparent;
302
- inset: 0;
303
- animation: orbSpin 6s linear infinite;
304
- }
305
- .orb-ring:nth-child(1) {
306
- border-top-color: #4db8f0;
307
- border-right-color: rgba(77,184,240,0.25);
308
- animation-duration: 5s;
309
- }
310
- .orb-ring:nth-child(2) {
311
- inset: 8px;
312
- border-top-color: #b57bee;
313
- border-left-color: rgba(181,123,238,0.25);
314
- animation-direction: reverse;
315
- animation-duration: 7s;
316
- }
317
- .orb-ring:nth-child(3) {
318
- inset: 16px;
319
- border-top-color: #1a6fd4;
320
- border-bottom-color: rgba(26,111,212,0.25);
321
- animation-duration: 9s;
322
- }
323
- @keyframes orbSpin { to { transform: rotate(360deg); } }
324
-
325
- .orb-core {
326
- position: absolute;
327
- inset: 24px;
328
- border-radius: 50%;
329
- background: radial-gradient(circle at 35% 35%,
330
- rgba(200,220,255,0.95),
331
- rgba(100,150,240,0.8) 40%,
332
- rgba(90,55,200,0.9) 75%,
333
- rgba(25,15,75,1)
334
- );
335
- box-shadow:
336
- 0 0 22px 6px rgba(77,184,240,0.45),
337
- 0 0 50px 12px rgba(120,80,220,0.28),
338
- inset 0 0 14px rgba(255,255,255,0.18);
339
- animation: orbBreathe 3.5s ease-in-out infinite;
340
- }
341
- .orb-core::after {
342
- content: '';
343
- position: absolute;
344
- top: 14%; left: 22%;
345
- width: 28%; height: 20%;
346
- border-radius: 50%;
347
- background: rgba(255,255,255,0.5);
348
- filter: blur(3px);
349
- }
350
- @keyframes orbBreathe {
351
- 0%, 100% {
352
- transform: scale(1);
353
- box-shadow: 0 0 22px 6px rgba(77,184,240,0.45), 0 0 50px 12px rgba(120,80,220,0.28), inset 0 0 14px rgba(255,255,255,0.18);
354
- }
355
- 50% {
356
- transform: scale(1.1);
357
- box-shadow: 0 0 34px 12px rgba(77,184,240,0.65), 0 0 70px 20px rgba(120,80,220,0.42), inset 0 0 20px rgba(255,255,255,0.28);
358
- }
359
- }
360
-
361
- .orb-waves {
362
- position: absolute;
363
- inset: 24px;
364
- border-radius: 50%;
365
- display: flex;
366
- align-items: center;
367
- justify-content: center;
368
- gap: 3px;
369
- overflow: hidden;
370
- }
371
- .orb-wave-bar {
372
- width: 3px;
373
- border-radius: 99px;
374
- background: rgba(255,255,255,0.75);
375
- animation: waveBar 1.3s ease-in-out infinite;
376
- }
377
- .orb-wave-bar:nth-child(1) { height: 8px; animation-delay: 0s; }
378
- .orb-wave-bar:nth-child(2) { height: 16px; animation-delay: 0.18s; }
379
- .orb-wave-bar:nth-child(3) { height: 22px; animation-delay: 0.35s; }
380
- .orb-wave-bar:nth-child(4) { height: 16px; animation-delay: 0.52s; }
381
- .orb-wave-bar:nth-child(5) { height: 8px; animation-delay: 0.7s; }
382
- @keyframes waveBar {
383
- 0%, 100% { transform: scaleY(0.3); opacity: 0.45; }
384
- 50% { transform: scaleY(1); opacity: 1; }
385
- }
386
-
387
- /* ── title & subtitle ── */
388
- .kasa-title {
389
- font-size: clamp(2.4rem, 7vw, 4rem);
390
- font-weight: 700;
391
- letter-spacing: -0.03em;
392
- background: linear-gradient(120deg, var(--white) 0%, var(--sky) 40%, var(--lilac) 100%);
393
- -webkit-background-clip: text;
394
- -webkit-text-fill-color: transparent;
395
- background-clip: text;
396
- margin: 0 0 10px;
397
- line-height: 1.1;
398
- }
399
- .kasa-sub {
400
- font-size: 0.97rem;
401
- font-weight: 300;
402
- color: rgba(220,230,255,0.85);
403
- max-width: 460px;
404
- margin: 0 auto 14px;
405
- line-height: 1.65;
406
- }
407
- .kasa-badge {
408
- display: inline-flex;
409
- align-items: center;
410
- gap: 6px;
411
- padding: 5px 16px;
412
- border-radius: 999px;
413
- border: 1px solid rgba(180,160,255,0.3);
414
- background: rgba(120,100,220,0.12);
415
- font-size: 0.7rem;
416
- font-weight: 500;
417
- letter-spacing: 0.13em;
418
- text-transform: uppercase;
419
- color: var(--lilac);
420
- backdrop-filter: blur(8px);
421
- }
422
-
423
- /* ── divider ── */
424
- .kasa-divider {
425
- border: none;
426
- border-top: 1px solid var(--border);
427
- margin: 10px 0 28px;
428
- position: relative;
429
- z-index: 1;
430
- }
431
-
432
- /* ── panel labels ── */
433
- .kasa-label {
434
- font-size: 0.7rem;
435
- font-weight: 600;
436
- letter-spacing: 0.14em;
437
- text-transform: uppercase;
438
- color: #a8d4f8;
439
- margin-bottom: 10px;
440
- }
441
-
442
- /* ── Gradio field labels ── */
443
- label, .gr-form label, .svelte-1gfkn6j {
444
- color: #c8d8f8 !important;
445
- font-family: 'Sora', sans-serif !important;
446
- font-size: 0.82rem !important;
447
- font-weight: 500 !important;
448
- }
449
-
450
- /* ── WHITE audio widget ── */
451
- .gr-audio,
452
- [data-testid="audio"],
453
- .gr-audio > div {
454
- background: #ffffff !important;
455
- border: 1.5px solid rgba(100,140,240,0.4) !important;
456
- border-radius: 14px !important;
457
- box-shadow: 0 4px 28px rgba(26,60,180,0.14) !important;
458
- overflow: hidden !important;
459
- }
460
-
461
- [data-testid="audio"] button,
462
- .gr-audio button {
463
- color: #1a6fd4 !important;
464
- background: transparent !important;
465
- }
466
- [data-testid="audio"] svg,
467
- .gr-audio svg {
468
- stroke: #1a6fd4 !important;
469
- fill: none !important;
470
- }
471
- [data-testid="audio"] span,
472
- [data-testid="audio"] .time,
473
- [data-testid="audio"] .duration,
474
- .gr-audio span {
475
- color: #1a3fa8 !important;
476
- font-family: 'DM Mono', monospace !important;
477
- }
478
- [data-testid="audio"] canvas,
479
- .gr-audio canvas {
480
- filter: hue-rotate(195deg) saturate(2) brightness(0.85) !important;
481
- }
482
- [data-testid="audio"] .tabs button,
483
- .gr-audio .tabs button {
484
- color: #1a6fd4 !important;
485
- font-family: 'Sora', sans-serif !important;
486
- font-weight: 600 !important;
487
- }
488
- [data-testid="audio"] .tabs button.selected,
489
- .gr-audio .tabs button.selected {
490
- border-bottom: 2px solid #1a6fd4 !important;
491
- }
492
-
493
- /* ── WHITE textbox ── */
494
- textarea, .gr-textbox textarea {
495
- background: #ffffff !important;
496
- border: 1.5px solid rgba(100,140,240,0.4) !important;
497
- border-radius: 12px !important;
498
- color: #0d1b6e !important;
499
- font-family: 'Sora', sans-serif !important;
500
- font-size: 0.95rem !important;
501
- font-weight: 400 !important;
502
- padding: 14px !important;
503
- box-shadow: 0 4px 24px rgba(26,60,180,0.1) !important;
504
- transition: border-color 0.2s, box-shadow 0.2s !important;
505
- line-height: 1.75 !important;
506
- letter-spacing: 0.01em !important;
507
- }
508
- textarea::placeholder {
509
- color: #7a9acc !important;
510
- font-style: italic;
511
- font-family: 'Sora', sans-serif !important;
512
- }
513
- textarea:focus {
514
- border-color: #4db8f0 !important;
515
- box-shadow: 0 0 0 3px rgba(77,184,240,0.2) !important;
516
- outline: none !important;
517
- }
518
- .gr-textbox, [data-testid="textbox"] {
519
- background: #ffffff !important;
520
- border: 1.5px solid rgba(100,140,240,0.4) !important;
521
- border-radius: 14px !important;
522
- box-shadow: 0 4px 28px rgba(26,60,180,0.14) !important;
523
- overflow: hidden !important;
524
- }
525
-
526
- /* ── Edit hint text ── */
527
- .edit-hint {
528
- font-size: 0.75rem;
529
- font-weight: 300;
530
- color: rgba(200,215,255,0.6);
531
- margin-top: -6px;
532
- margin-bottom: 8px;
533
- font-style: italic;
534
- letter-spacing: 0.02em;
535
- }
536
-
537
- /* ── Buttons ── */
538
- .gr-button-primary, button.primary {
539
- background: linear-gradient(135deg, #1a6fd4 0%, #7b4fd4 100%) !important;
540
- border: none !important;
541
- border-radius: 10px !important;
542
- font-family: 'Sora', sans-serif !important;
543
- font-weight: 600 !important;
544
- font-size: 0.9rem !important;
545
- padding: 12px 28px !important;
546
- color: #ffffff !important;
547
- transition: transform 0.15s, box-shadow 0.15s !important;
548
- box-shadow: 0 4px 22px rgba(100,80,200,0.4) !important;
549
- position: relative;
550
- z-index: 1;
551
- }
552
- .gr-button-primary:hover, button.primary:hover {
553
- transform: translateY(-2px) !important;
554
- box-shadow: 0 8px 32px rgba(120,80,220,0.55) !important;
555
- }
556
- .gr-button-primary:active, button.primary:active {
557
- transform: translateY(0) !important;
558
- }
559
-
560
- /* ── Save checkmark button ── */
561
- #save-btn button {
562
- width: 52px !important;
563
- height: 52px !important;
564
- min-width: 52px !important;
565
- border-radius: 50% !important;
566
- padding: 0 !important;
567
- font-size: 1.5rem !important;
568
- background: linear-gradient(135deg, #1a6fd4 0%, #7b4fd4 100%) !important;
569
- border: none !important;
570
- box-shadow: 0 4px 22px rgba(100,80,200,0.4) !important;
571
- color: #fff !important;
572
- transition: transform 0.15s, box-shadow 0.15s !important;
573
- line-height: 1 !important;
574
- }
575
- #save-btn button:hover {
576
- transform: scale(1.12) translateY(-2px) !important;
577
- box-shadow: 0 8px 32px rgba(120,80,220,0.55) !important;
578
- }
579
- #save-btn button:active {
580
- transform: scale(0.96) !important;
581
- }
582
-
583
-
584
- /* ── footer ── */
585
- .kasa-footer {
586
- text-align: center;
587
- font-size: 0.72rem;
588
- color: rgba(180,190,255,0.58);
589
- padding: 24px 0 32px;
590
- letter-spacing: 0.05em;
591
- position: relative;
592
- z-index: 1;
593
- }
594
-
595
- /* ── pulse dot ── */
596
- .pulse-dot {
597
- display: inline-block;
598
- width: 9px; height: 9px;
599
- border-radius: 50%;
600
- background: #ffffff;
601
- box-shadow: 0 0 6px 2px rgba(255,255,255,0.6);
602
- animation: pulse 2s ease-in-out infinite;
603
- }
604
- @keyframes pulse {
605
- 0%, 100% { opacity: 1; transform: scale(1); }
606
- 50% { opacity: 0.3; transform: scale(0.65); }
607
- }
608
- """
609
-
610
- # =========================================================
611
- # UI
612
- # =========================================================
613
-
614
- with gr.Blocks(css=css, theme=gr.themes.Base()) as demo:
615
-
616
- # ── Hero ──
617
- gr.HTML("""
618
- <div id="hero">
619
- <div class="orb-wrap">
620
- <div class="orb-stage">
621
- <div class="orb-ring"></div>
622
- <div class="orb-ring"></div>
623
- <div class="orb-ring"></div>
624
- <div class="orb-core"></div>
625
- <div class="orb-waves">
626
- <div class="orb-wave-bar"></div>
627
- <div class="orb-wave-bar"></div>
628
- <div class="orb-wave-bar"></div>
629
- <div class="orb-wave-bar"></div>
630
- <div class="orb-wave-bar"></div>
631
- </div>
632
- </div>
633
- </div>
634
- <div class="kasa-title">Kasanoma ASR</div>
635
- <div class="kasa-sub">
636
- Automatic speech recognition for natural English–Twi
637
- code-switched conversations. Speak the way you actually speak.
638
- </div>
639
- <span class="kasa-badge"><span class="pulse-dot"></span>Live Transcription</span>
640
- </div>
641
- """)
642
-
643
- # ── Main columns ──
644
- with gr.Row(equal_height=True):
645
-
646
- with gr.Column(scale=1):
647
- gr.HTML('<div class="kasa-label">Audio Input</div>')
648
- audio_input = gr.Audio(
649
- sources=["microphone", "upload"],
650
- type="filepath",
651
- label="Record or upload audio"
652
- )
653
- transcribe_btn = gr.Button("Transcribe β†’", variant="primary")
654
-
655
- with gr.Column(scale=1):
656
- gr.HTML('<div class="kasa-label">Transcription</div>')
657
- model_output = gr.Textbox(
658
- label="",
659
- placeholder="Transcription will appear here…",
660
- lines=3,
661
- interactive=False
662
- )
663
- gr.HTML('<div class="kasa-label" style="margin-top:14px;">Edit</div>')
664
- gr.HTML('<div class="edit-hint">Type what was actually said β€” fix any errors the model made, then hit βœ“ to save.</div>')
665
- with gr.Row(equal_height=True):
666
- corrected_output = gr.Textbox(
667
- label="",
668
- placeholder="Correct the transcription if needed…",
669
- lines=3,
670
- scale=9
671
- )
672
- save_btn = gr.Button("βœ“", elem_id="save-btn", scale=1)
673
-
674
-
675
- hidden_audio_path = gr.State()
676
-
677
- # ── Footer ──
678
- gr.HTML("""
679
- <div class="kasa-footer">
680
- Kasanoma &middot; English&ndash;Twi Code-Switching ASR &middot; Project Kasa 2026
681
- </div>
682
- """)
683
-
684
- # ── Event handlers ──
685
- transcribe_btn.click(
686
- fn=transcribe_audio,
687
- inputs=audio_input,
688
- outputs=[model_output, corrected_output, hidden_audio_path]
689
- )
690
-
691
- save_btn.click(
692
- fn=save_sample,
693
- inputs=[hidden_audio_path, corrected_output]
694
- )
695
-
696
- # =========================================================
697
- # LAUNCH
698
- # =========================================================
699
-
700
- if __name__ == "__main__":
701
- demo.launch(share=True)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
assets/gen.png ADDED
assets/logo.png ADDED
main.py CHANGED
@@ -1,3 +1,5 @@
1
  from api import app
 
2
 
 
3
  __all__ = ["app"]
 
1
  from api import app
2
+ from fastapi.staticfiles import StaticFiles
3
 
4
+ app.mount("/assets", StaticFiles(directory="assets"), name="assets")
5
  __all__ = ["app"]
requirements.txt CHANGED
@@ -8,4 +8,5 @@ datasets>=3.2.0
8
  torch>=2.0.0
9
  torchaudio>=2.0.0
10
  librosa>=0.9.2
11
- python-dotenv>=0.19.0
 
 
8
  torch>=2.0.0
9
  torchaudio>=2.0.0
10
  librosa>=0.9.2
11
+ python-dotenv>=0.19.0
12
+ gradio>=3.38.0
static/index.html CHANGED
@@ -13,16 +13,16 @@
13
  *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
14
 
15
  :root {
16
- --bg: #f5f3ef;
17
- --surface: #faf9f7;
18
  --card: #ffffff;
19
- --border: rgba(0,0,0,0.08);
20
- --border-md: rgba(0,0,0,0.13);
21
- --text: #1a1714;
22
- --muted: #7a7268;
23
- --hint: #b3aca3;
24
  --accent: #2a5c45;
25
- --accent-bg:#ecf4ef;
26
  --red: #c0392b;
27
  --red-bg: #fdf0ee;
28
  --font-serif: 'Instrument Serif', Georgia, serif;
@@ -37,7 +37,12 @@
37
 
38
  body {
39
  font-family: var(--font-sans);
40
- background: var(--bg);
 
 
 
 
 
41
  color: var(--text);
42
  min-height: 100vh;
43
  display: flex;
@@ -90,6 +95,12 @@
90
  }
91
  /* ── Hero ── */
92
  .hero { margin-bottom: 36px; }
 
 
 
 
 
 
93
  .lang-tag {
94
  display: inline-flex;
95
  align-items: center;
@@ -382,11 +393,34 @@
382
  display: flex;
383
  align-items: center;
384
  justify-content: space-between;
 
 
 
 
 
 
 
385
  }
386
  .footer-copy { font-size: 0.72rem; color: var(--hint); }
387
  .footer-links { display: flex; gap: 18px; }
388
  .footer-links a { font-size: 0.72rem; color: var(--hint); text-decoration: none; transition: color 0.2s; }
389
  .footer-links a:hover { color: var(--muted); }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
390
 
391
  /* ── Toast ── */
392
  .toast {
@@ -413,12 +447,73 @@
413
  @keyframes blink { 0%,100%{opacity:1;transform:scale(1)}50%{opacity:.3;transform:scale(.55)} }
414
  @keyframes rot { to { transform: rotate(360deg); } }
415
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
416
  @media (max-width: 520px) {
417
  body { padding: 28px 14px 60px; }
418
  .stats { grid-template-columns: 1fr 1fr; }
419
  .stats .stat:last-child { grid-column: span 2; }
420
  .nav { display: none; }
421
  .footer { flex-direction: column; gap: 10px; }
 
 
422
  }
423
  </style>
424
  </head>
@@ -426,14 +521,16 @@
426
 
427
  <div class="page">
428
 
429
- <nav class="topbar">
430
- <span class="logo">Kasanoma</span>
431
- <ul class="nav">
432
- <li><a href="#">Docs</a></li>
433
- <li><a href="#">Dataset</a></li>
434
- <li><a href="#">GitHub</a></li>
435
- </ul>
436
- </nav>
 
 
437
 
438
  <div class="hero">
439
  <div class="lang-tag"><i class="ti ti-language" aria-hidden="true"></i>English Β· Twi code-switching</div>
@@ -499,7 +596,8 @@
499
  </button>
500
  </div>
501
  </div>
502
-
 
503
  <div class="sec">
504
  <div class="sec-label">Correct &amp; save</div>
505
  <p class="edit-hint">Fix any errors β€” your correction helps improve the model.</p>
@@ -511,10 +609,11 @@
511
  <button class="disc-btn" id="disc-btn">Discard</button>
512
  </div>
513
  </div>
 
514
 
515
- </div>
516
 
517
- <div class="stats">
518
  <div class="stat">
519
  <div class="stat-v" id="stat-saved">0</div>
520
  <div class="stat-l">Saved</div>
@@ -527,14 +626,20 @@
527
  <div class="stat-v" style="font-size:1.1rem;font-style:normal;font-family:var(--font-sans);font-weight:500;padding-top:4px;">Twi Β· EN</div>
528
  <div class="stat-l">Language</div>
529
  </div>
530
- </div>
531
 
532
  <footer class="footer">
533
- <span class="footer-copy">Project Kasa Β· 2026</span>
534
- <div class="footer-links">
535
- <a href="#">Privacy</a>
536
- <a href="#">Terms</a>
537
- <a href="#">Contact</a>
 
 
 
 
 
 
538
  </div>
539
  </footer>
540
 
@@ -697,38 +802,40 @@
697
  if (t) navigator.clipboard.writeText(t).then(() => toast('Copied', 'ok'));
698
  });
699
 
700
- $('save-btn').addEventListener('click', async () => {
701
- const text = $('edit-ta').value.trim();
702
- if (!text) { toast('Nothing to save.', 'err'); return; }
703
- if (!audioBlob) { toast('No audio attached.', 'err'); return; }
704
- const btn = $('save-btn'); btn.disabled = true;
705
- try {
706
- const fd = new FormData();
707
- fd.append('audio', audioBlob, audioBlob.name || 'audio.webm');
708
- fd.append('transcription', text);
709
- const res = await fetch('/save', { method: 'POST', body: fd });
710
- if (!res.ok) throw new Error(`Server error ${res.status}`);
711
- saved++; $('stat-saved').textContent = saved;
712
- btn.classList.add('saved');
713
- setTimeout(() => btn.classList.remove('saved'), 2000);
714
- toast('Saved to dataset', 'ok');
715
- } catch (err) {
716
- toast('Save failed: ' + err.message, 'err');
717
- } finally { btn.disabled = false; }
718
- });
719
-
720
- $('disc-btn').addEventListener('click', () => {
721
- $('edit-ta').value = '';
722
- $('out').classList.remove('has');
723
- $('out-txt').textContent = '';
724
- toast('Cleared');
725
- });
726
-
727
- // ── Load real saved count from backend on page load ──
728
- fetch('/dataset/stats')
729
- .then(r => r.json())
730
- .then(d => { if (d.total) { saved = d.total; $('stat-saved').textContent = saved; } })
731
- .catch(() => {});
 
 
732
  </script>
733
  </body>
734
  </html>
 
13
  *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
14
 
15
  :root {
16
+ --bg: #f4faf6;
17
+ --surface: #fbfdfc;
18
  --card: #ffffff;
19
+ --border: rgba(42,92,69,0.10);
20
+ --border-md: rgba(42,92,69,0.18);
21
+ --text: #16261f;
22
+ --muted: #5f7a6c;
23
+ --hint: #9db3a6;
24
  --accent: #2a5c45;
25
+ --accent-bg:#e8f5ee;
26
  --red: #c0392b;
27
  --red-bg: #fdf0ee;
28
  --font-serif: 'Instrument Serif', Georgia, serif;
 
37
 
38
  body {
39
  font-family: var(--font-sans);
40
+ background:
41
+ radial-gradient(ellipse 900px 600px at 12% 0%, rgba(154,211,181,0.35), transparent 60%),
42
+ radial-gradient(ellipse 900px 650px at 100% 15%, rgba(210,235,220,0.55), transparent 65%),
43
+ radial-gradient(ellipse 800px 700px at 50% 100%, rgba(180,220,197,0.30), transparent 60%),
44
+ linear-gradient(180deg, #ffffff 0%, #f2f9f5 45%, #eaf6ef 100%);
45
+ background-attachment: fixed;
46
  color: var(--text);
47
  min-height: 100vh;
48
  display: flex;
 
95
  }
96
  /* ── Hero ── */
97
  .hero { margin-bottom: 36px; }
98
+ .hero-logo {
99
+ height: 44px;
100
+ width: auto;
101
+ margin-bottom: 16px;
102
+ display: block;
103
+ }
104
  .lang-tag {
105
  display: inline-flex;
106
  align-items: center;
 
393
  display: flex;
394
  align-items: center;
395
  justify-content: space-between;
396
+ margin-top: 8px;
397
+ padding-top: 8px;
398
+ }
399
+ .footer-content {
400
+ display: flex;
401
+ align-items: center;
402
+ gap: 24px;
403
  }
404
  .footer-copy { font-size: 0.72rem; color: var(--hint); }
405
  .footer-links { display: flex; gap: 18px; }
406
  .footer-links a { font-size: 0.72rem; color: var(--hint); text-decoration: none; transition: color 0.2s; }
407
  .footer-links a:hover { color: var(--muted); }
408
+ .footer-logos {
409
+ display: flex;
410
+ gap: 12px;
411
+ align-items: center;
412
+ }
413
+ .footer-logo {
414
+ height: 72px;
415
+ width: 72px;
416
+ border-radius: 10%;
417
+ background: var(--surface);
418
+ transition: transform 0.2s ease, border-color 0.2s ease;
419
+ }
420
+ .footer-logo:hover {
421
+ transform: scale(1.08);
422
+ border-color: var(--accent);
423
+ }
424
 
425
  /* ── Toast ── */
426
  .toast {
 
447
  @keyframes blink { 0%,100%{opacity:1;transform:scale(1)}50%{opacity:.3;transform:scale(.55)} }
448
  @keyframes rot { to { transform: rotate(360deg); } }
449
 
450
+ /* ── Header Logo Section ── */
451
+ .header-logo-section {
452
+ display: flex;
453
+ align-items: center;
454
+ justify-content: space-between;
455
+ padding: 20px 0 24px;
456
+ margin-bottom: 28px;
457
+ animation: slideDown 0.5s ease;
458
+ }
459
+ .header-logo-container {
460
+ display: flex;
461
+ align-items: center;
462
+ gap: 16px;
463
+ }
464
+ .header-logo-container img {
465
+ height: 64px;
466
+ width: auto;
467
+ transition: transform 0.3s ease;
468
+ }
469
+ .header-logo-container img:hover {
470
+ transform: scale(1.06);
471
+ }
472
+ .header-title {
473
+ font-family: var(--font-serif);
474
+ font-size: 1.15rem;
475
+ font-weight: 400;
476
+ color: var(--text);
477
+ letter-spacing: -0.01em;
478
+ }
479
+ .header-nav {
480
+ display: flex;
481
+ gap: 16px;
482
+ align-items: center;
483
+ }
484
+ .header-nav-link {
485
+ display: inline-flex;
486
+ align-items: center;
487
+ gap: 5px;
488
+ padding: 7px 12px;
489
+ border-radius: 6px;
490
+ border: 1px solid var(--border);
491
+ background: var(--accent-bg);
492
+ color: var(--accent);
493
+ text-decoration: none;
494
+ font-size: 0.8rem;
495
+ font-weight: 500;
496
+ transition: all 0.2s ease;
497
+ font-family: var(--font-sans);
498
+ }
499
+ .header-nav-link:hover {
500
+ border-color: var(--accent);
501
+ background: rgba(42, 92, 69, 0.15);
502
+ color: var(--text);
503
+ }
504
+ @keyframes slideDown {
505
+ from { opacity: 0; transform: translateY(-12px); }
506
+ to { opacity: 1; transform: translateY(0); }
507
+ }
508
+
509
  @media (max-width: 520px) {
510
  body { padding: 28px 14px 60px; }
511
  .stats { grid-template-columns: 1fr 1fr; }
512
  .stats .stat:last-child { grid-column: span 2; }
513
  .nav { display: none; }
514
  .footer { flex-direction: column; gap: 10px; }
515
+ .header-nav { gap: 10px; }
516
+ .header-nav-link { font-size: 0.75rem; padding: 6px 10px; }
517
  }
518
  </style>
519
  </head>
 
521
 
522
  <div class="page">
523
 
524
+ <div class="header-logo-section">
525
+ <div class="header-logo-container">
526
+ <img src="/assets/logo.png" alt="Kasanoma logo" />
527
+ <div class="header-title">Kasanoma</div>
528
+ </div>
529
+ <div class="header-nav">
530
+ <a href="https://huggingface.co/datasets/Kennethdot/Ghana_English-Twi_Code-switching_ASR" class="header-nav-link">Docs</a>
531
+ <a href="https://github.com/Kennethdotse/project-kasa" class="header-nav-link">GitHub</a>
532
+ </div>
533
+ </div>
534
 
535
  <div class="hero">
536
  <div class="lang-tag"><i class="ti ti-language" aria-hidden="true"></i>English Β· Twi code-switching</div>
 
596
  </button>
597
  </div>
598
  </div>
599
+
600
+ <!-- COMMENTED OUT: Correct & save section
601
  <div class="sec">
602
  <div class="sec-label">Correct &amp; save</div>
603
  <p class="edit-hint">Fix any errors β€” your correction helps improve the model.</p>
 
609
  <button class="disc-btn" id="disc-btn">Discard</button>
610
  </div>
611
  </div>
612
+ -->
613
 
614
+ </div>
615
 
616
+ <!-- <div class="stats">
617
  <div class="stat">
618
  <div class="stat-v" id="stat-saved">0</div>
619
  <div class="stat-l">Saved</div>
 
626
  <div class="stat-v" style="font-size:1.1rem;font-style:normal;font-family:var(--font-sans);font-weight:500;padding-top:4px;">Twi Β· EN</div>
627
  <div class="stat-l">Language</div>
628
  </div>
629
+ </div> -->
630
 
631
  <footer class="footer">
632
+ <div class="footer-content">
633
+ <span class="footer-copy">Project Kasa Β· 2026</span>
634
+ <div class="footer-links">
635
+ <a href="#">Privacy</a>
636
+ <a href="#">Terms</a>
637
+ <a href="#">Contact</a>
638
+ </div>
639
+ </div>
640
+ <div class="footer-logos">
641
+ <img src="/assets/logo.png" alt="Kasanoma logo" class="footer-logo" />
642
+ <img src="/assets/gen.png" alt="Gen logo" class="footer-logo" />
643
  </div>
644
  </footer>
645
 
 
802
  if (t) navigator.clipboard.writeText(t).then(() => toast('Copied', 'ok'));
803
  });
804
 
805
+ // COMMENTED OUT: Save to dataset functionality
806
+ // $('save-btn').addEventListener('click', async () => {
807
+ // const text = $('edit-ta').value.trim();
808
+ // if (!text) { toast('Nothing to save.', 'err'); return; }
809
+ // if (!audioBlob) { toast('No audio attached.', 'err'); return; }
810
+ // const btn = $('save-btn'); btn.disabled = true;
811
+ // try {
812
+ // const fd = new FormData();
813
+ // fd.append('audio', audioBlob, audioBlob.name || 'audio.webm');
814
+ // fd.append('transcription', text);
815
+ // const res = await fetch('/save', { method: 'POST', body: fd });
816
+ // if (!res.ok) throw new Error(`Server error ${res.status}`);
817
+ // saved++; $('stat-saved').textContent = saved;
818
+ // btn.classList.add('saved');
819
+ // setTimeout(() => btn.classList.remove('saved'), 2000);
820
+ // toast('Saved to dataset', 'ok');
821
+ // } catch (err) {
822
+ // toast('Save failed: ' + err.message, 'err');
823
+ // } finally { btn.disabled = false; }
824
+ // });
825
+
826
+ // COMMENTED OUT: Discard transcription
827
+ // $('disc-btn').addEventListener('click', () => {
828
+ // $('edit-ta').value = '';
829
+ // $('out').classList.remove('has');
830
+ // $('out-txt').textContent = '';
831
+ // toast('Cleared');
832
+ // });
833
+
834
+ // COMMENTED OUT: Load real saved count from backend on page load
835
+ // fetch('/dataset/stats')
836
+ // .then(r => r.json())
837
+ // .then(d => { if (d.total) { saved = d.total; $('stat-saved').textContent = saved; } })
838
+ // .catch(() => {});
839
  </script>
840
  </body>
841
  </html>