Spaces:
Sleeping
Sleeping
File size: 10,623 Bytes
0cf3992 | 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 | import warnings
from core.state import AgenticState
import json
from loguru import logger
# Configuration
MODEL_NAME = "sshleifer/distilbart-cnn-12-6"
@logger.catch
def load_bart_summarizer():
"""Load BART model with proper configuration"""
from transformers import pipeline, AutoTokenizer, AutoModelForSeq2SeqLM
import torch
warnings.filterwarnings("ignore")
device = 0 if torch.cuda.is_available() else -1
logger.info(f"The model {MODEL_NAME} is loaded", MODEL_NAME=MODEL_NAME)
try:
# Try pipeline first
summarizer = pipeline(
"summarization",
model=MODEL_NAME,
device=device,
truncation=True
)
return summarizer
except:
# Fallback to direct model loading
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)
if device == 0:
model = model.cuda()
class BartSummarizer:
def __init__(self, model, tokenizer, device):
self.model = model
self.tokenizer = tokenizer
self.device = device
def __call__(self, text, max_length=150, min_length=30, **kwargs):
inputs = self.tokenizer(text, return_tensors="pt",
truncation=True, max_length=1024)
if self.device == 0:
inputs = {k: v.cuda() for k, v in inputs.items()}
with torch.no_grad():
summary_ids = self.model.generate(
inputs['input_ids'],
max_length=max_length,
min_length=min_length,
num_beams=4,
early_stopping=True,
no_repeat_ngram_size=3,
**kwargs
)
summary = self.tokenizer.decode(summary_ids[0],
skip_special_tokens=True)
return [{"summary_text": summary}]
return BartSummarizer(model, tokenizer, device)
@logger.catch
def clean_summary_output(result):
"""Extract clean string from various response formats"""
if isinstance(result, list) and len(result) > 0:
if 'summary_text' in result[0]:
return result[0]['summary_text']
elif 'generated_text' in result[0]:
return result[0]['generated_text']
else:
return str(result[0])
elif isinstance(result, dict):
return result.get('summary_text', result.get('generated_text', str(result)))
else:
return str(result)
@logger.catch
def synthesize_executive_summary(sections, summarizer):
"""Create a flowing executive summary from all sections"""
# Collect all section summaries and key points
all_content = []
for section in sections:
all_content.append(section.get('summary', ''))
all_content.extend(section.get('key_points', []))
combined = " ".join(all_content)[:2000]
if not combined:
logger.error("Node 5: No summary available")
return "No summary available."
# Create a prompt-like input for better summaries
prompt = f"Summarize the following interview content: {combined}"
try:
result = summarizer(
prompt,
max_length=180,
min_length=80,
do_sample=False
)
return clean_summary_output(result)
except:
# Fallback to first section summary
return sections[0].get('summary', 'Summary unavailable.')
@logger.catch
def generate_tldr(sections, summarizer):
"""Generate a concise one-sentence TL;DR"""
# Use first section as base
first_section = sections[0].get('summary', '')[:300]
if not first_section:
return "Video summary."
try:
result = summarizer(
first_section,
max_length=40,
min_length=15,
do_sample=False
)
tldr = clean_summary_output(result)
# Ensure it's a complete sentence
if not tldr.endswith(('.', '!', '?')):
tldr += '.'
return tldr
except:
return sections[0].get('title', 'Summary') + '.'
@logger.catch
def format_section_content(section, index, video_id):
"""Format a single section with proper structure"""
title = section.get('title', f'Section {index}')
title = title.replace('-', '').strip()
if not title or len(title) < 5:
title = f"Part {index}"
summary = section.get('summary', 'No summary available.')
if not summary.endswith(('.', '!', '?')):
summary += '.'
key_points = section.get('key_points', [])
if not key_points:
key_points = ["Key insights from this section"]
points_formatted = "\n".join([f"- {p}" for p in key_points if p])
# Explanation
explanation = section.get('explanation', section.get('content', ''))
explanation_section = f"\n**Explanation**\n\n{explanation}\n" if explanation and len(explanation) > len(summary) else ""
return f"""
### {index}. {title}
**Summary**
{summary}
{explanation_section}
**Key Insights**
{points_formatted}
"""
@logger.catch
async def node_5_beautiful_presentation(state: AgenticState) -> AgenticState:
""" Node 5: Beautiful Presentation witg local model
"""
logger.info("🚀 Node 5: Beautiful Presentation (Local) started...")
import asyncio
structured = getattr(state, "structured_script", {})
# Get metadata directly from state attributes
sections = structured.get("sections", [])
if not sections:
state.errors.append({"type": "missing_structure"})
logger.error("Mode 5: Missing structure")
return state
# Get metadata directly from state attributes
video_metadata = getattr(state, "video_metadata", {}) or {}
video_id = (
getattr(state, "video_id", None)
or video_metadata.get("video_id")
or "UNKNOWN"
)
title = (
getattr(state, "title", None)
or video_metadata.get("title")
or (sections[0].get("title") if sections else None)
or "YouTube Video Summary"
)
channel = (
getattr(state, "channel", None)
or video_metadata.get("channel")
or "Unknown Channel"
)
duration_human = (
getattr(state, "duration_human", None)
or video_metadata.get("duration_human")
or None
)
# Upload date
upload_date = (
getattr(state, "upload_date", None)
or video_metadata.get("upload_date")
or None
)
# Fallback to section title if needed
if not title or title == "Unknown Title" or title == "YouTube Video Summary":
if sections and len(sections) > 0:
title = sections[0].get("title", "YouTube Video Summary")
# Load summarizer
loop = asyncio.get_event_loop()
summarizer = await loop.run_in_executor(None, load_bart_summarizer)
# Executive Summary - Synthesize all sections
logger.info(" Generating executive summary...")
exec_summary = await loop.run_in_executor(
None, synthesize_executive_summary, sections, summarizer)
# TL;DR - One sentence
logger.info(" Generating TL;DR...")
tldr = generate_tldr(sections, summarizer)
# Main Topics - From structured data
topics_md = ""
topics = structured.get("main_topics", [])
if topics:
# Clean topics
clean_topics = [t.strip() for t in topics if t and len(t) > 3]
if clean_topics:
topics_md = "## Main Topics\n\n"
topics_md += "\n".join(f"- {t}" for t in clean_topics[:8])
# Table of Contents
toc = "## Table of Contents\n\n"
for i, sec in enumerate(sections, 1):
title_text = sec.get('title', f'Section {i}')
clean_title = title_text.replace('-', '').strip()
if clean_title and not clean_title.startswith('The following'):
toc += f"{i}. {clean_title}\n"
# Sections - Properly formatted
logger.info(" Formatting {sections} sections...", sections=len(sections))
sections_md = []
for i, sec in enumerate(sections, 1):
sections_md.append(format_section_content(sec, i, video_id))
# Quotes - Clean and deduplicate
quotes_md = ""
quotes = structured.get("key_quotes", [])
if quotes:
# Clean quotes
clean_quotes = []
for q in quotes:
if q and len(q) > 10:
# Remove any surrounding quotes
q = q.strip('"').strip()
if q not in clean_quotes:
clean_quotes.append(q)
if clean_quotes:
quotes_md = "## Key Quotes\n\n"
for q in clean_quotes[:6]:
quotes_md += f'> "{q}"\n\n'
# Entities - Clean and sort
entities_md = ""
entities = structured.get("mentioned_entities", [])
if entities:
# Clean and deduplicate
clean_entities = []
for e in entities:
if e and len(e) > 1:
# Remove duplicates and clean
if e not in clean_entities:
clean_entities.append(e)
if clean_entities:
entities_md = "## Key Mentions\n\n"
for ent in sorted(clean_entities)[:20]:
entities_md += f"- {ent}\n"
# Final Markdown
final_md = f"""# {title}
*Channel: {channel}*
*Video ID: {video_id}*
*Duration: {duration_human}*
*Updated: {upload_date}*
---
## Executive Summary
{exec_summary}
## TL;DR
{tldr}
---
{topics_md}
---
{toc}
---
{chr(10).join(sections_md)}
---
{quotes_md}
---
{entities_md}
---
*Generated with YouTubeScriptMaster (Local AI Mode)*
"""
# Clean up any remaining formatting issues
final_md = final_md.replace(' ', ' ').replace('\n\n\n', '\n\n')
state.final_formatted_markdown = final_md
state.presentation_complete = True
logger.info("✅ Node 5 complete")
quotes_count = len(clean_quotes) if 'clean_quotes' in locals() else 0
entities_count = len(clean_entities) if 'clean_entities' in locals() else 0
logger.info(f" Sections: {len(sections_md)}")
logger.info(f" Executive summary: {exec_summary[:100] if exec_summary else ''}...")
logger.info(f" TL;DR: {tldr}")
logger.info(f" Quotes: {quotes_count}")
logger.info(f" Entities: {entities_count}")
return state |