File size: 38,474 Bytes
55e1fc0
 
169a94e
55e1fc0
 
169a94e
 
55e1fc0
 
169a94e
 
 
 
 
 
 
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
 
 
 
 
169a94e
55e1fc0
169a94e
299c2ce
55e1fc0
169a94e
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
 
 
169a94e
 
 
55e1fc0
169a94e
 
 
 
 
 
 
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
 
169a94e
55e1fc0
169a94e
 
 
 
 
55e1fc0
169a94e
55e1fc0
169a94e
55e1fc0
cc1687a
169a94e
55e1fc0
169a94e
55e1fc0
169a94e
 
 
 
 
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
 
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
55e1fc0
 
169a94e
 
6389c61
299c2ce
169a94e
299c2ce
55e1fc0
169a94e
cc1687a
299c2ce
169a94e
 
 
 
 
 
 
 
 
 
 
fb21a11
55e1fc0
 
fb21a11
169a94e
 
 
55e1fc0
169a94e
299c2ce
6389c61
169a94e
 
 
4b2f370
169a94e
299c2ce
169a94e
 
55e1fc0
4b2f370
169a94e
 
 
 
 
c32d410
169a94e
 
c32d410
169a94e
 
c32d410
299c2ce
169a94e
c32d410
55e1fc0
169a94e
c32d410
299c2ce
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
169a94e
 
 
 
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
 
 
169a94e
 
55e1fc0
 
 
169a94e
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
6389c61
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
55e1fc0
169a94e
 
 
55e1fc0
 
 
 
169a94e
55e1fc0
169a94e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import gradio as gr
import asyncio
from asyncio import Semaphore # Added for concurrency control
from pathlib import Path
import os
import tempfile # Added for temporary chunk files
import traceback # Import traceback for better error logging
import re
import pandas as pd
from dataclasses import dataclass
from typing import Dict, AsyncGenerator, Tuple, Any, List

# Use standard import convention for genai
# Assuming genai is installed and configured elsewhere
from google import genai
from youtube_transcript_api import YouTubeTranscriptApi

# Import pydub for audio manipulation
from pydub import AudioSegment
from pydub.exceptions import CouldntDecodeError


# --- Constants ---
PROMPT_KEYS = ["titles_and_thumbnails", "description", "previews", "clips", "timestamps"]
PROMPT_DISPLAY_NAMES = {
    "titles_and_thumbnails": "Titles and Thumbnails",
    "description": "Twitter Description",
    "previews": "Preview Clips",
    "clips": "Twitter Clips",
    "timestamps": "Timestamps"
}
# --- MODIFIED: Increased chunk size to 30 minutes ---
AUDIO_CHUNK_DURATION_MS = 30 * 60 * 1000 # Process audio in 30-minute chunks
# --- ADDED: Concurrency Limits ---
MAX_CONCURRENT_TRANSCRIPTIONS = 3 # Limit simultaneous transcription API calls
MAX_CONCURRENT_GENERATIONS = 4    # Limit simultaneous content generation API calls

# --- Core Classes (ContentRequest, ContentGenerator) ---
# (ContentRequest and ContentGenerator remain unchanged)
@dataclass
class ContentRequest:
    prompt_key: str

class ContentGenerator:
    def __init__(self):
        self.current_prompts = self._load_default_prompts()
        self.client: genai.Client | None = None

    def _load_default_prompts(self) -> Dict[str, str]:
        # (Implementation identical to previous version)
        prompts = {}
        timestamp_examples, title_examples, description_examples, clip_examples = "", "", "", ""
        try:
            data_dir = Path("data")
            if data_dir.is_dir():
                try: timestamps_df = pd.read_csv(data_dir / "Timestamps.csv"); timestamp_examples = "\n\n".join(timestamps_df['Timestamps'].dropna().tolist())
                except Exception as e: print(f"Warning: Loading Timestamps.csv failed: {e}")
                try: titles_df = pd.read_csv(data_dir / "Titles & Thumbnails.csv"); title_examples = "\n".join([f'Title: "{r.Titles}"\nThumbnail: "{r.Thumbnail}"' for _, r in titles_df.iterrows() if pd.notna(r.Titles) and pd.notna(r.Thumbnail)])
                except Exception as e: print(f"Warning: Loading Titles & Thumbnails.csv failed: {e}")
                try: descriptions_df = pd.read_csv(data_dir / "Viral Episode Descriptions.csv"); description_examples = "\n".join([f'Tweet: "{r["Tweet Text"]}"' for _, r in descriptions_df.iterrows() if pd.notna(r["Tweet Text"])])
                except Exception as e: print(f"Warning: Loading Viral Episode Descriptions.csv failed: {e}")
                try: clips_df = pd.read_csv(data_dir / "Viral Twitter Clips.csv"); clip_examples = "\n\n".join([f'Tweet Text: "{r["Tweet Text"]}"\nClip Transcript: "{r["Clip Transcript"]}"' for _, r in clips_df.iterrows() if pd.notna(r["Tweet Text"]) and pd.notna(r["Clip Transcript"])])
                except Exception as e: print(f"Warning: Loading Viral Twitter Clips.csv failed: {e}")
            else: print("Warning: 'data' directory not found.")
        except Exception as e: print(f"Warning: Error accessing 'data' directory: {e}")

        prompts_dir = Path("prompts")
        if not prompts_dir.is_dir():
                print("Error: 'prompts' directory not found.")
                return {key: f"ERROR: Prompt directory missing." for key in PROMPT_KEYS}
        for key in PROMPT_KEYS:
            try:
                prompt = (prompts_dir / f"{key}.txt").read_text(encoding='utf-8')
                if key == "timestamps": prompt = prompt.replace("{timestamps_examples}", timestamp_examples)
                elif key == "titles_and_thumbnails": prompt = prompt.replace("{title_examples}", title_examples)
                elif key == "description": prompt = prompt.replace("{description_examples}", description_examples)
                elif key == "clips": prompt = prompt.replace("{clip_examples}", clip_examples)
                prompts[key] = prompt
            except Exception as e:
                print(f"Warning: Loading prompt prompts/{key}.txt failed: {e}")
                prompts[key] = f"Generate {key} based on the transcript. Do not use markdown formatting." # Fallback
        for key in PROMPT_KEYS: prompts.setdefault(key, f"Generate {key} based on the transcript. Do not use markdown formatting.")
        return prompts

    async def generate_content(self, request: ContentRequest, transcript: str) -> str:
        # (Implementation identical to previous version)
        if not self.client: return "ERROR_CONFIGURATION: Gemini Client not initialized."
        if not transcript: return "ERROR_INTERNAL: Empty transcript provided for content generation."
        try:
            system_prompt = self.current_prompts.get(request.prompt_key)
            if not system_prompt: return f"ERROR_INTERNAL: System prompt for '{request.prompt_key}' missing."
            contents_for_api = [system_prompt, transcript]
            # --- IMPORTANT: Model kept as gemini-1.5-flash ---
            model_name = "gemini-2.5-pro-preview-03-25"
            response = await asyncio.to_thread(
                self.client.models.generate_content, model=model_name, contents=contents_for_api
            )
            if not response: return f"ERROR_API: No response received for {request.prompt_key}."
            try:
                    if response.text:
                            try:
                                    if hasattr(response, 'prompt_feedback') and response.prompt_feedback.block_reason:
                                            reason = response.prompt_feedback.block_reason.name; return f"ERROR_BLOCKED: Blocked by API. Reason: {reason}"
                            except AttributeError: pass
                            return str(response.text.strip())
                    else:
                            if response.candidates and response.candidates[0].content and response.candidates[0].content.parts:
                                    full_text = "".join(part.text for part in response.candidates[0].content.parts if hasattr(part, 'text')).strip()
                                    if full_text:
                                        print(f"Warning: Used fallback text extraction via candidates for {request.prompt_key}")
                                        return str(full_text)
                            return f"ERROR_NO_TEXT: Could not extract text from response for {request.prompt_key}."
            except (ValueError, AttributeError) as e:
                    print(f"Error accessing response text/feedback for {request.prompt_key} (potentially blocked): {e}")
                    reason = "Unknown"
                    try:
                        if hasattr(response, 'prompt_feedback') and response.prompt_feedback.block_reason: reason = response.prompt_feedback.block_reason.name
                    except AttributeError: pass
                    return f"ERROR_BLOCKED: Content generation failed (possibly blocked). Reason: {reason}"
        except Exception as e:
            print(f"Error generating content for {request.prompt_key}: {traceback.format_exc()}")
            error_str = str(e).lower()
            # Add specific check for rate limit errors if the API provides clear indicators
            if "rate limit exceeded" in error_str or "quota exceeded" in error_str or "429" in error_str:
                 return f"ERROR_RATE_LIMIT: API limit likely exceeded. Details: {str(e)}"
            elif "permission denied" in error_str or "api key not valid" in error_str: return f"ERROR_PERMISSION_DENIED: API Error (Permission Denied?). Check Key. Details: {str(e)}"
            # elif "quota" in error_str: return f"ERROR_QUOTA: API Quota Error. Details: {str(e)}" # Covered by rate limit check above
            elif "model" in error_str and "not found" in error_str: return f"ERROR_MODEL_NOT_FOUND: Model name likely incorrect. Details: {str(e)}"
            else: return f"ERROR_API_GENERAL: API Error during generation. Details: {str(e)}"

    def update_prompts(self, *values):
        # (Implementation identical to previous version)
        updated_keys = []
        for key, value in zip(PROMPT_KEYS, values):
            if isinstance(value, str): self.current_prompts[key] = value; updated_keys.append(key)
        return f"Prompts updated: {', '.join(updated_keys)}" if updated_keys else "No prompts updated."

# (extract_video_id and get_transcript remain unchanged)
def extract_video_id(url: str) -> str | None:
    patterns = [r"(?:v=|\/)([0-9A-Za-z_-]{11}).*", r"youtu\.be\/([0-9A-Za-z_-]{11})"]
    for pattern in patterns:
        match = re.search(pattern, url);
        if match: return match.group(1)
    return None

def get_transcript(video_id: str) -> str:
    if not video_id: raise ValueError("Invalid Video ID")
    try:
        t_list = YouTubeTranscriptApi.list_transcripts(video_id)
        transcript = t_list.find_transcript(['en', 'en-US'])
        fetched = transcript.fetch()
        if not fetched: raise ValueError("Fetched transcript empty")
        return " ".join(entry.get("text", "") for entry in fetched).strip()
    except Exception as e:
        return f"ERROR_TRANSCRIPT_FETCH: Failed for ID '{video_id}'. Reason: {e}"

# --- TranscriptProcessor Class (Refactored for Concurrency Control) ---
class TranscriptProcessor:
    def __init__(self):
        self.generator = ContentGenerator()

    # (Helper _get_youtube_transcript remains unchanged)
    def _get_youtube_transcript(self, url: str) -> str:
       # ... (identical implementation)
        print(f"Extracting Video ID from: {url}")
        video_id = extract_video_id(url)
        if not video_id: raise ValueError(f"Invalid YouTube URL/ID: {url}")
        print(f"Video ID: {video_id}. Fetching transcript...")
        try:
                transcript = get_transcript(video_id)
                if transcript.startswith("ERROR_TRANSCRIPT_FETCH"): raise Exception(transcript)
                if not transcript: raise ValueError(f"Empty transcript for ID: {video_id}")
                print(f"Transcript fetched (length: {len(transcript)}).")
                return transcript
        except Exception as e: print(f"Error fetching YouTube transcript: {e}"); raise Exception(f"Failed to get YouTube transcript: {str(e)}")


    # --- MODIFIED: Added Semaphore argument ---
    async def _transcribe_chunk(self, client: genai.Client, chunk_path: Path, chunk_index: int, total_chunks: int, semaphore: Semaphore) -> str:
        """Transcribes a single audio chunk using Gemini API, respecting the semaphore."""
        # Acquire semaphore before proceeding
        async with semaphore:
            print(f"Semaphore acquired for chunk {chunk_index + 1}/{total_chunks}. Processing...")
            gemini_audio_file_ref = None
            try:
                print(f"Uploading chunk {chunk_index + 1}/{total_chunks}: {chunk_path.name}")
                gemini_audio_file_ref = await asyncio.to_thread(client.files.upload, file=chunk_path)
                print(f"Chunk {chunk_index + 1} uploaded. File Ref: {gemini_audio_file_ref.name}")

                prompt_for_transcription = "Transcribe the following audio file accurately."
                contents = [prompt_for_transcription, gemini_audio_file_ref]
                # --- IMPORTANT: Model kept as gemini-1.5-flash ---
                model_name = "gemini-2.5-pro-preview-03-25"

                print(f"Requesting transcription for chunk {chunk_index + 1}...")
                # Make the API call *within* the semaphore lock
                transcription_response = await asyncio.to_thread(
                    client.models.generate_content, model=model_name, contents=contents
                )
                print(f"Transcription response received for chunk {chunk_index + 1}.")

                # Extract transcript text (identical logic)
                transcript_piece = ""
                try:
                    if transcription_response.text:
                        transcript_piece = transcription_response.text.strip()
                    elif transcription_response.candidates and transcription_response.candidates[0].content and transcription_response.candidates[0].content.parts:
                        transcript_piece = "".join(part.text for part in transcription_response.candidates[0].content.parts if hasattr(part, 'text')).strip()

                    if not transcript_piece and hasattr(transcription_response, 'prompt_feedback') and transcription_response.prompt_feedback.block_reason:
                         reason = transcription_response.prompt_feedback.block_reason.name
                         print(f"Warning: Transcription blocked for chunk {chunk_index + 1}. Reason: {reason}")
                         return f"[CHUNK_ERROR: Blocked - {reason}]"

                    print(f"Chunk {chunk_index + 1} transcript length: {len(transcript_piece)}")
                    return str(transcript_piece)

                except (ValueError, AttributeError, Exception) as extraction_err:
                    print(f"Error extracting transcript for chunk {chunk_index + 1}: {extraction_err}. Response: {transcription_response}")
                    return f"[CHUNK_ERROR: Extraction Failed - {str(extraction_err)}]"

            except Exception as e:
                print(f"Error processing chunk {chunk_index + 1} (within semaphore): {traceback.format_exc()}")
                error_str = str(e).lower()
                # Add specific check for rate limit errors
                if "rate limit exceeded" in error_str or "quota exceeded" in error_str or "429" in error_str:
                    return f"[CHUNK_ERROR: API Rate Limit Exceeded - {str(e)}]"
                elif "permission denied" in error_str or "api key not valid" in error_str:
                     return f"[CHUNK_ERROR: API Permission Denied - {str(e)}]"
                elif "file size" in error_str:
                     return f"[CHUNK_ERROR: File Size Limit Exceeded - {str(e)}]"
                else:
                     return f"[CHUNK_ERROR: General API/Processing Error - {str(e)}]"
            finally:
                # Cleanup happens *before* semaphore is released automatically by 'async with'
                if gemini_audio_file_ref:
                    # Run cleanup in background to avoid blocking semaphore release if deletion is slow
                    asyncio.create_task(self.delete_uploaded_file(client, gemini_audio_file_ref.name, f"chunk {chunk_index + 1} cleanup"))
                if chunk_path.exists():
                    try:
                        os.remove(chunk_path)
                    except OSError as e:
                        print(f"Warning: Could not delete local temp chunk file {chunk_path}: {e}")
                print(f"Semaphore released for chunk {chunk_index + 1}/{total_chunks}.")
        # Semaphore is automatically released when exiting 'async with' block


    async def process_transcript(self, client: genai.Client, audio_file: Any) -> AsyncGenerator[Tuple[str, Any], None]:
        """
        Processes audio with larger chunks and controlled concurrency using Semaphores.
        """
        if AudioSegment is None:
            yield "error", "Audio processing library (pydub) not loaded. Cannot proceed."
            return
        if not client:
            yield "error", "Gemini Client object was not provided."
            return
        self.generator.client = client
        if not audio_file:
            yield "error", "No audio file provided."
            return

        audio_path_str = getattr(audio_file, 'name', None)
        if not audio_path_str:
            yield "error", "Invalid audio file object."
            return
        original_audio_path = Path(audio_path_str)
        if not original_audio_path.exists():
            yield "error", f"Audio file not found: {original_audio_path}"
            return

        # --- ADDED: Initialize Semaphores ---
        transcription_semaphore = Semaphore(MAX_CONCURRENT_TRANSCRIPTIONS)
        generation_semaphore = Semaphore(MAX_CONCURRENT_GENERATIONS)

        try:
            yield "status", f"Loading audio file: {original_audio_path.name}..."
            print(f"Loading audio file with pydub: {original_audio_path}")
            try:
                 file_format = original_audio_path.suffix.lower().replace('.', '')
                 audio = AudioSegment.from_file(original_audio_path, format=file_format if file_format else None)
            except CouldntDecodeError as decode_error:
                 print(f"pydub decode error: {decode_error}. Make sure ffmpeg is installed.")
                 yield "error", f"Failed to load/decode audio file: {original_audio_path.name}. Ensure valid format and ffmpeg."
                 return
            except Exception as load_err:
                 print(f"Error loading audio with pydub: {traceback.format_exc()}")
                 yield "error", f"Error loading audio file {original_audio_path.name}: {load_err}"
                 return

            duration_ms = len(audio)
            # --- MODIFIED: Chunk duration increased ---
            total_chunks = (duration_ms + AUDIO_CHUNK_DURATION_MS - 1) // AUDIO_CHUNK_DURATION_MS
            print(f"Audio loaded. Duration: {duration_ms / 1000:.2f}s. Splitting into {total_chunks} x {AUDIO_CHUNK_DURATION_MS / 60000:.1f}min chunks.")
            yield "status", f"Audio loaded ({duration_ms / 1000:.2f}s). Transcribing in {total_chunks} chunks (max {MAX_CONCURRENT_TRANSCRIPTIONS} concurrent)..."

            transcript_pieces = [""] * total_chunks # Pre-allocate list to store pieces in order
            transcription_tasks = []

            # --- MODIFIED: Create tasks with semaphore ---
            for i in range(total_chunks):
                start_ms = i * AUDIO_CHUNK_DURATION_MS
                end_ms = min((i + 1) * AUDIO_CHUNK_DURATION_MS, duration_ms)
                chunk = audio[start_ms:end_ms]

                with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_chunk_file:
                    chunk_path = Path(temp_chunk_file.name)
                try:
                    chunk.export(chunk_path, format="wav")
                except Exception as export_err:
                     print(f"Error exporting chunk {i+1}: {traceback.format_exc()}")
                     yield "error", f"Failed to create temporary audio chunk file: {export_err}"
                     if chunk_path.exists(): os.remove(chunk_path)
                     return

                # Pass semaphore to the chunk transcription function
                task = asyncio.create_task(self._transcribe_chunk(client, chunk_path, i, total_chunks, transcription_semaphore))
                # Store task along with its index to place result correctly
                transcription_tasks.append((i, task))

            # Process transcription results as they complete, maintaining order
            processed_chunks = 0
            # Wait for all tasks using gather, but process results as they come in via callbacks or checking task states?
            # Using asyncio.gather might be simpler here if we need all results before proceeding. Let's try gather.
            # results = await asyncio.gather(*(task for _, task in transcription_tasks), return_exceptions=True)

            # Alternative: Process as completed, but store in correct order
            temp_results = {}
            for index, task in transcription_tasks:
                 try:
                     result = await task
                     temp_results[index] = result
                     processed_chunks += 1
                     yield "status", f"Transcribed chunk {processed_chunks}/{total_chunks}..."
                     # Check for critical chunk errors immediately if needed
                     if isinstance(result, str) and ("[CHUNK_ERROR: API Rate Limit Exceeded" in result or \
                                                      "[CHUNK_ERROR: API Permission Denied" in result or \
                                                      "[CHUNK_ERROR: API Quota Exceeded" in result):
                         print(f"Critical API error in chunk {index + 1}, stopping transcription. Error: {result}")
                         yield "error", f"Transcription stopped. Critical API error in chunk {index + 1}: {result.split('-', 1)[-1].strip()}"
                         # Cancel remaining tasks (important!)
                         for j, other_task in transcription_tasks:
                            if not other_task.done():
                                other_task.cancel()
                         return # Stop processing
                 except asyncio.CancelledError:
                     print(f"Transcription task for chunk {index + 1} was cancelled.")
                     temp_results[index] = "[CHUNK_ERROR: Cancelled]"
                     # If one task is cancelled due to an error in another, we might stop everything
                     if processed_chunks < total_chunks: # Avoid double error message if already stopped
                         yield "error", "Transcription process was cancelled."
                         return
                 except Exception as e:
                     print(f"Error waiting for transcription task {index + 1}: {traceback.format_exc()}")
                     temp_results[index] = f"[CHUNK_ERROR: Task Processing Failed - {str(e)}]"
                     processed_chunks += 1 # Count as processed even though it failed

            # Reconstruct the transcript in the correct order
            transcript_pieces = [temp_results.get(i, "[CHUNK_ERROR: Missing Result]") for i in range(total_chunks)]
            full_transcript = " ".join(transcript_pieces).strip()

            # Improved check for transcription failure
            if not full_transcript or full_transcript.isspace() or all(s.startswith("[CHUNK_ERROR") for s in transcript_pieces if s):
                 error_summary = " ".join(p for p in transcript_pieces if p.startswith("[CHUNK_ERROR"))
                 print(f"Transcription failed or resulted in only errors. Summary: {error_summary}")
                 yield "error", f"Failed to transcribe audio or all chunks failed. Errors: {error_summary[:200]}"
                 return

            print(f"Full transcript concatenated (length: {len(full_transcript)}).")
            yield "status", "Transcription complete. Generating content..."

            # --- Generate other content using the FULL transcript with Semaphore ---
            generation_tasks = []
            for key in PROMPT_KEYS:
                # Pass generation semaphore to the item generation function
                task = asyncio.create_task(self._generate_single_item(key, full_transcript, generation_semaphore))
                generation_tasks.append(task)

            generated_items = 0
            total_items = len(PROMPT_KEYS)
            # Process generation results as they complete
            for future in asyncio.as_completed(generation_tasks):
                try:
                    key, result = await future # Result from _generate_single_item
                    yield "progress", (key, result)
                    generated_items += 1
                    # More granular status for generation
                    yield "status", f"Generating content ({key} done, {generated_items}/{total_items} total)..."
                except asyncio.CancelledError:
                     # Should not happen unless transcription failed and cancelled tasks
                     print("Content generation task was cancelled.")
                     yield "error", "Content generation cancelled."
                     return
                except Exception as e:
                    print(f"Error processing completed generation task: {traceback.format_exc()}")
                    yield "status", f"Error during content generation phase: {str(e)}"
                    # Optionally yield an error for the specific item?
                    # key_if_possible = "unknown_key" # How to get key here? Task doesn't easily pass it back on exception
                    # yield "progress", (key_if_possible, f"ERROR_GENERATION: {str(e)}")

            yield "status", "All content generation tasks complete."

        except FileNotFoundError as e:
            yield "error", f"File Error: {str(e)}"
            return
        except Exception as e: # Catch-all for transcription setup phase
            print(f"Error during transcription setup/chunking phase: {traceback.format_exc()}")
            yield "error", f"System Error during transcription setup: {str(e)}"
            return


    async def delete_uploaded_file(self, client: genai.Client, file_name: str, context: str):
        # (Implementation identical - called in background now)
        if not client or not file_name:
            # print(f"Skipping deletion: Invalid client or file name ({context}).") # Reduce noise
            return
        try:
            # print(f"Attempting background cleanup: {file_name} ({context})")
            await asyncio.to_thread(client.files.delete, name=file_name)
            print(f"Successfully cleaned up Gemini file: {file_name} ({context})")
        except Exception as cleanup_err:
            if "not found" in str(cleanup_err).lower() or "404" in str(cleanup_err):
                 pass # Ignore file not found during cleanup
                # print(f"Info: File {file_name} likely already deleted ({context}).")
            else:
                print(f"Warning: Failed Gemini file cleanup for {file_name} ({context}): {cleanup_err}")


    # --- MODIFIED: Added Semaphore argument ---
    async def _generate_single_item(self, key: str, transcript: str, semaphore: Semaphore) -> Tuple[str, str]:
        """Helper to generate one piece of content, respecting the semaphore."""
        # Acquire semaphore before calling the API
        async with semaphore:
            print(f"Semaphore acquired for generating: {key}. Calling API...")
            result = await self.generator.generate_content(ContentRequest(key), transcript)
            print(f"Finished generation task for: {key}. Semaphore released.")
            # Semaphore is released automatically by 'async with'
            return key, result

    def update_prompts(self, *values) -> str:
        # (Implementation identical to previous version)
        return self.generator.update_prompts(*values)


# --- Gradio Interface Creation (UI remains unchanged from previous version) ---

def create_interface():
    """Create the Gradio interface (UI definition identical to last version)."""
    processor = TranscriptProcessor()

    key_titles = "titles_and_thumbnails"
    key_desc = "description"
    key_previews = "previews"
    key_clips = "clips"
    key_timestamps = "timestamps"
    display_titles = PROMPT_DISPLAY_NAMES[key_titles]
    display_desc = PROMPT_DISPLAY_NAMES[key_desc]
    display_previews = PROMPT_DISPLAY_NAMES[key_previews]
    display_clips = PROMPT_DISPLAY_NAMES[key_clips]
    display_timestamps = PROMPT_DISPLAY_NAMES[key_timestamps]

    with gr.Blocks(title="Gemini Podcast Content Generator") as app:
        gr.Markdown(
            """
            # Gemini Podcast Content Generator
            Generate social media content from podcast audio using Gemini.
            Enter your Google API key below and upload an audio file.
            Audio will be processed in larger (~30min) chunks with controlled concurrency.
            """
        ) # Updated description slightly

        with gr.Tab("Generate Content"):
            google_api_key_input = gr.Textbox(
                label="Google API Key", type="password",
                placeholder="Enter your Google API Key here",
                info="Your GCP account needs to have billing enabled to use the 2.5 pro model."
            )
            input_audio = gr.File(
                label="Upload Audio File", file_count="single",
                file_types=["audio", ".mp3", ".wav", ".ogg", ".flac", ".m4a", ".aac"]
            )
            submit_btn = gr.Button("Generate with Gemini", variant="huggingface")

            gr.Markdown("### Processing Status")
            output_status = gr.Textbox(label="Current Status", value="Idle.", interactive=False, lines=1, max_lines=5)

            gr.Markdown(f"### {display_titles}")
            output_titles = gr.Textbox(label="", value="...", interactive=False, lines=3, max_lines=10) 

            gr.Markdown(f"### {display_desc}")
            output_desc = gr.Textbox(label="", value="...", interactive=False, lines=3, max_lines=10) 

            gr.Markdown(f"### {display_previews}")
            output_previews = gr.Textbox(label="", value="...", interactive=False, lines=3, max_lines=10) 

            gr.Markdown(f"### {display_clips}")
            output_clips = gr.Textbox(label="", value="...", interactive=False, lines=3, max_lines=10) 

            gr.Markdown(f"### {display_timestamps}")
            output_timestamps = gr.Textbox(label="", value="...", interactive=False, lines=3, max_lines=10) 

            outputs_list = [
                output_status,
                output_titles, output_desc, output_previews,
                output_clips, output_timestamps
            ]
            results_component_map = {
                key_titles: output_titles, key_desc: output_desc, key_previews: output_previews,
                key_clips: output_clips, key_timestamps: output_timestamps
            }

            # --- process_wrapper (UI Update Logic - largely unchanged) ---
            async def process_wrapper(google_key, audio_file_obj, progress=gr.Progress(track_tqdm=True)):
                print("Started Processing...")
                initial_updates = {
                    output_status: gr.update(value="Initiating..."),
                    output_titles: gr.update(value="⏳ Pending..."),
                    output_desc: gr.update(value="⏳ Pending..."),
                    output_previews: gr.update(value="⏳ Pending..."),
                    output_clips: gr.update(value="⏳ Pending..."),
                    output_timestamps: gr.update(value="⏳ Pending..."),
                }
                yield initial_updates

                if not google_key:
                    yield {output_status: gr.update(value="πŸ›‘ Error: Missing Google API Key.")}
                    return
                if not audio_file_obj:
                    yield {output_status: gr.update(value="πŸ›‘ Error: No audio file uploaded.")}
                    return

                masked_key = f"{'*'*(len(google_key)-4)}{google_key[-4:]}" if len(google_key) > 4 else "****"
                print(f"Using Google Key: {masked_key}")
                print(f"Audio file: Name='{getattr(audio_file_obj, 'name', 'N/A')}'")
                client: genai.Client | None = None
                try:
                    yield {output_status: gr.update(value="⏳ Initializing Gemini Client...")}
                    client = await asyncio.to_thread(genai.Client, api_key=google_key)
                    print("Gemini Client initialized successfully.")
                    yield {output_status: gr.update(value="βœ… Client Initialized.")}
                except Exception as e:
                    error_msg = f"πŸ›‘ Error: Failed Client Initialization: {e}"
                    print(f"Client Init Error: {traceback.format_exc()}")
                    yield {output_status: gr.update(value=error_msg)}
                    return

                updates_to_yield = {}
                try:
                    # Call the refactored processor
                    async for update_type, data in processor.process_transcript(client, audio_file_obj):
                        updates_to_yield = {}
                        if update_type == "status":
                            updates_to_yield[output_status] = gr.update(value=f"⏳ {data}")
                        elif update_type == "progress":
                            key, result = data
                            component_to_update = results_component_map.get(key)
                            if component_to_update:
                                ui_result = ""
                                if isinstance(result, str) and result.startswith("ERROR_"):
                                    # Handle specific rate limit error display
                                    if result.startswith("ERROR_RATE_LIMIT"):
                                         ui_result = f"❌ Error (Rate Limit):\n{result.split(':', 1)[-1].strip()}"
                                    else:
                                        try:
                                            error_type, error_detail = result.split(':', 1)
                                            error_type_display = error_type.replace('ERROR_', '').replace('_', ' ').title()
                                            ui_result = f"❌ Error ({error_type_display}):\n{error_detail.strip()}"
                                        except ValueError:
                                            ui_result = f"❌ Error:\n{result}"
                                else:
                                    ui_result = str(result)
                                updates_to_yield[component_to_update] = gr.update(value=ui_result)
                            else:
                                print(f"Warning: No UI component mapped for result key '{key}'")
                        elif update_type == "error":
                            error_message = f"πŸ›‘ Processing Error: {data}"
                            updates_to_yield[output_status] = gr.update(value=error_message)
                            yield updates_to_yield
                            return

                        if updates_to_yield:
                           yield updates_to_yield

                    final_success_update = {output_status: gr.update(value="βœ… Processing Complete.")}
                    final_success_update.update(updates_to_yield) # Include any final progress updates
                    yield final_success_update
                    print("Process wrapper finished successfully.")

                except Exception as e:
                    print(f"Error in process_wrapper async loop: {traceback.format_exc()}")
                    error_msg = f"πŸ›‘ Unexpected wrapper error: {e}"
                    yield {output_status: gr.update(value=error_msg)}

            submit_btn.click(
                fn=process_wrapper,
                inputs=[google_api_key_input, input_audio],
                outputs=outputs_list
            )

        with gr.Tab("Customize Prompts"):
            # (Customize Prompts tab UI remains unchanged)
            gr.Markdown("## Customize Generation Prompts")
            prompt_inputs = []
            default_prompts = processor.generator.current_prompts
            for key in PROMPT_KEYS:
                display_name = PROMPT_DISPLAY_NAMES.get(key, key.replace('_', ' ').title())
                default_value = default_prompts.get(key, "")
                prompt_inputs.append(gr.Textbox(label=f"{display_name} Prompt", lines=10, value=default_value or ""))

            status_prompt_tab = gr.Textbox(label="Status", interactive=False)
            update_btn = gr.Button("Update Session Prompts")
            update_btn.click(fn=processor.update_prompts, inputs=prompt_inputs, outputs=[status_prompt_tab])

            reset_btn = gr.Button("Reset to Default Prompts")
            def reset_prompts_ui():
                try:
                    defaults = processor.generator._load_default_prompts()
                    if any(isinstance(v, str) and v.startswith("ERROR:") for v in defaults.values()): raise ValueError("Failed to load one or more default prompts.")
                    processor.generator.current_prompts = defaults
                    updates = {status_prompt_tab: gr.update(value="Prompts reset to defaults!")}
                    for i, key in enumerate(PROMPT_KEYS):
                        updates[prompt_inputs[i]] = gr.update(value=defaults.get(key, ""))
                    return updates
                except Exception as e:
                    print(f"Error during prompt reset: {e}")
                    return {status_prompt_tab: gr.update(value=f"Error resetting prompts: {str(e)}")}

            reset_btn.click(
                 fn=reset_prompts_ui,
                 inputs=None,
                 outputs=[status_prompt_tab] + prompt_inputs
            )

    return app

# --- Main Execution Block (Unchanged) ---
if __name__ == "__main__":
    if AudioSegment is None:
        print("\nFATAL ERROR: pydub is required but could not be imported.")
        print("Please install it ('pip install pydub') and ensure ffmpeg is available.")
        print("Application cannot start correctly.")
        exit(1)

    Path("prompts").mkdir(exist_ok=True)
    Path("data").mkdir(exist_ok=True)
    _prompt_dir = Path("prompts")
    for key in PROMPT_KEYS:
        prompt_file = _prompt_dir / f"{key}.txt"
        if not prompt_file.exists():
            # Ensure default prompts advise against markdown
            default_content = f"This is the default placeholder prompt for {PROMPT_DISPLAY_NAMES[key]}. Process the transcript provided. Important: Generate the response as plain text only. Do not use any Markdown formatting (no '#', '*', '_', list formatting, bolding, etc.)."
            if key == "titles_and_thumbnails": default_content += "\n\nExamples:\n{title_examples}"
            elif key == "description": default_content += "\n\nExamples:\n{description_examples}"
            elif key == "clips": default_content += "\n\nExamples:\n{clip_examples}"
            elif key == "timestamps": default_content += "\n\nExamples:\n{timestamps_examples}"
            prompt_file.write_text(default_content, encoding='utf-8')
            print(f"Created dummy prompt file: {prompt_file}")

    print("Starting Gradio application...")
    app = create_interface()
    app.launch()
    print("Gradio application stopped.")