File size: 23,105 Bytes
65c1b40 6dde9f1 65c1b40 6dde9f1 65c1b40 dd1bdb6 65c1b40 dd1bdb6 65c1b40 |
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 |
import gradio as gr
import pixeltable as pxt
from pixeltable.iterators import DocumentSplitter, FrameIterator, StringSplitter
from pixeltable.functions.huggingface import sentence_transformer, clip_image, clip_text
from pixeltable.functions.video import extract_audio
from pixeltable.functions.audio import get_metadata
from pixeltable.functions import openai
import numpy as np
import PIL.Image
import os
import getpass
import requests
import tempfile
from datetime import datetime
# Configuration
PIXELTABLE_MEDIA_DIR = os.path.expanduser("~/.pixeltable/media")
MAX_TOKENS_DEFAULT = 300
TEMPERATURE_DEFAULT = 0.7
CHUNK_SIZE_DEFAULT = 300
# Initialize API keys
def init_api_keys():
if 'OPENAI_API_KEY' not in os.environ:
os.environ['OPENAI_API_KEY'] = getpass.getpass('OpenAI API key:')
# Embedding Functions
@pxt.expr_udf
def e5_embed(text: str) -> np.ndarray:
return sentence_transformer(text, model_id='intfloat/e5-large-v2')
@pxt.expr_udf
def embed_image(img: PIL.Image.Image):
return clip_image(img, model_id='openai/clip-vit-base-patch32')
@pxt.expr_udf
def str_embed(s: str):
return clip_text(s, model_id='openai/clip-vit-base-patch32')
# Common Utilities
def initialize_pixeltable(dir_name='unified_app'):
"""Initialize Pixeltable directory"""
pxt.drop_dir(dir_name, force=True)
pxt.create_dir(dir_name)
@pxt.udf
def create_prompt(top_k_list: list[dict], question: str) -> str:
"""Create a standardized prompt format"""
concat_top_k = '\n\n'.join(elt['text'] for elt in reversed(top_k_list))
return f'''
PASSAGES:
{concat_top_k}
QUESTION:
{question}'''
@pxt.udf(return_type=pxt.AudioType())
def generate_audio(script: str, voice: str, api_key: str):
"""Generate audio from text using OpenAI's API"""
if not script or not voice:
return None
try:
response = requests.post(
"https://api.openai.com/v1/audio/speech",
headers={"Authorization": f"Bearer {api_key}"},
json={"model": "tts-1", "input": script, "voice": voice}
)
if response.status_code == 200:
temp_dir = os.path.join(os.getcwd(), "temp")
os.makedirs(temp_dir, exist_ok=True)
temp_file = os.path.join(temp_dir, f"audio_{os.urandom(8).hex()}.mp3")
with open(temp_file, 'wb') as f:
f.write(response.content)
return temp_file
except Exception as e:
print(f"Error in audio synthesis: {e}")
return None
# Document Processing
class DocumentProcessor:
@staticmethod
def process_documents(pdf_files, chunk_limit, chunk_separator):
"""Process uploaded documents for chatbot functionality"""
initialize_pixeltable()
docs = pxt.create_table(
'unified_app.documents',
{'document': pxt.DocumentType(nullable=True)}
)
docs.insert({'document': file.name} for file in pdf_files if file.name.endswith('.pdf'))
chunks = pxt.create_view(
'unified_app.chunks',
docs,
iterator=DocumentSplitter.create(
document=docs.document,
separators=chunk_separator,
limit=chunk_limit if chunk_separator in ["token_limit", "char_limit"] else None
)
)
chunks.add_embedding_index('text', string_embed=e5_embed)
return "Documents processed successfully. You can start asking questions."
@staticmethod
def get_document_answer(question):
"""Get answer from processed documents"""
try:
chunks = pxt.get_table('unified_app.chunks')
sim = chunks.text.similarity(question)
relevant_chunks = chunks.order_by(sim, asc=False).limit(5).select(chunks.text).collect()
context = "\n\n".join(chunk['text'] for chunk in relevant_chunks)
temp_table = pxt.create_table(
'unified_app.temp_response',
{
'question': pxt.StringType(),
'context': pxt.StringType()
}
)
temp_table.insert([{'question': question, 'context': context}])
temp_table['response'] = openai.chat_completions(
messages=[
{
'role': 'system',
'content': 'Answer the question based only on the provided context. If the context doesn\'t contain enough information, say so.'
},
{
'role': 'user',
'content': f"Context:\n{context}\n\nQuestion: {question}"
}
],
model='gpt-4o-mini-2024-07-18'
)
answer = temp_table.select(
answer=temp_table.response.choices[0].message.content
).tail(1)['answer'][0]
pxt.drop_table('unified_app.temp_response', force=True)
return answer
except Exception as e:
return f"Error: {str(e)}"
# Call Analysis
class CallAnalyzer:
@staticmethod
def process_call(video_file):
"""Process and analyze call recordings"""
try:
initialize_pixeltable()
calls = pxt.create_table(
'unified_app.calls',
{"video": pxt.VideoType(nullable=True)}
)
calls['audio'] = extract_audio(calls.video, format='mp3')
calls['transcription'] = openai.transcriptions(audio=calls.audio, model='whisper-1')
calls['text'] = calls.transcription.text
sentences = pxt.create_view(
'unified_app.sentences',
calls,
iterator=StringSplitter.create(text=calls.text, separators='sentence')
)
sentences.add_embedding_index('text', string_embed=e5_embed)
@pxt.udf
def generate_insights(text: str) -> list[dict]:
return [
{'role': 'system', 'content': 'Analyze this call transcript and provide key insights:'},
{'role': 'user', 'content': text}
]
calls['insights_prompt'] = generate_insights(calls.text)
calls['insights'] = openai.chat_completions(
messages=calls.insights_prompt,
model='gpt-4o-mini-2024-07-18'
).choices[0].message.content
calls.insert([{"video": video_file}])
result = calls.select(calls.text, calls.audio, calls.insights).tail(1)
return result['text'][0], result['audio'][0], result['insights'][0]
except Exception as e:
return f"Error processing call: {str(e)}", None, None
# Video Search
class VideoSearcher:
@staticmethod
def process_video(video_file):
"""Process video for searching"""
try:
initialize_pixeltable()
videos = pxt.create_table('unified_app.videos', {'video': pxt.VideoType()})
frames = pxt.create_view(
'unified_app.frames',
videos,
iterator=FrameIterator.create(video=videos.video, fps=1)
)
frames.add_embedding_index('frame', string_embed=str_embed, image_embed=embed_image)
videos.insert([{'video': video_file.name}])
return "Video processed and indexed for search."
except Exception as e:
return f"Error processing video: {str(e)}"
@staticmethod
def search_video(search_type, text_query=None, image_query=None):
"""Search processed video frames"""
try:
frames = pxt.get_table('unified_app.frames')
if search_type == "Text" and text_query:
sim = frames.frame.similarity(text_query)
elif search_type == "Image" and image_query is not None:
sim = frames.frame.similarity(image_query)
else:
return []
results = frames.order_by(sim, asc=False).limit(5).select(frames.frame).collect()
return [row['frame'] for row in results]
except Exception as e:
print(f"Search error: {str(e)}")
return []
# Gradio Interface
def create_interface():
with gr.Blocks(theme=gr.themes.Base()) as demo:
# Header
gr.HTML(
"""
<div style="text-align: left; margin-bottom: 1rem;">
<img src="https://raw.githubusercontent.com/pixeltable/pixeltable/main/docs/source/data/pixeltable-logo-large.png" alt="Pixeltable" style="max-width: 150px;" />
</div>
"""
)
gr.Markdown(
"""
# Multimodal Powerhouse
"""
)
gr.HTML(
"""
<p>
<a href="https://github.com/pixeltable/pixeltable" target="_blank" style="color: #F25022; text-decoration: none; font-weight: bold;">Pixeltable</a>
is a declarative interface for working with text, images, embeddings, and video, enabling you to store, transform, index, and iterate on data.
</p>
<div style="background-color: #E5DDD4; border: 1px solid #e9ecef; border-radius: 8px; padding: 15px; margin: 15px 0;">
<strong>⚠️ Note:</strong> This app runs best with GPU. For optimal performance, consider
<a href="https://huggingface.co/spaces/Pixeltable/Multimodal-Processing-Suite?duplicate=true" target="_blank" style="color: #F25022; text-decoration: none; font-weight: bold;">duplicating this space</a>
to run locally or with better computing resources.
</div>
"""
)
# Documentation Sections
with gr.Row():
with gr.Column():
with gr.Accordion("🎯 What This App Does", open=False):
gr.Markdown("""
1. 📚 **Document Processing**
* Chat with your documents using RAG
* Process multiple document formats
* Extract key insights
2. 🎥 **Video Analysis**
* Text and image-based video search
* Frame extraction and indexing
* Visual content discovery
3. 🎙️ **Call Analysis**
* Automatic transcription
* Key insight extraction
* Audio processing
""")
with gr.Column():
with gr.Accordion("⚙️ How It Works", open=False):
gr.Markdown("""
1. 🔄 **Data Processing**
* Chunking and indexing documents
* Embedding generation for search
* Multi-modal data handling
2. 🤖 **AI Integration**
* LLM-powered analysis
* Speech-to-text conversion
* Semantic search capabilities
3. 📊 **Storage & Retrieval**
* Efficient data organization
* Quick content retrieval
* Structured data management
""")
with gr.Tabs():
# Document Chat Tab
with gr.TabItem("📚 Document Chat"):
with gr.Row():
with gr.Column():
doc_files = gr.File(label="Upload Documents", file_count="multiple")
chunk_size = gr.Slider(
minimum=100,
maximum=500,
value=CHUNK_SIZE_DEFAULT,
label="Chunk Size"
)
chunk_type = gr.Dropdown(
choices=["token_limit", "char_limit", "sentence", "paragraph"],
value="token_limit",
label="Chunking Method"
)
process_docs_btn = gr.Button("Process Documents")
process_status = gr.Textbox(label="Status")
with gr.Column():
chatbot = gr.Chatbot(label="Document Chat")
msg = gr.Textbox(label="Ask a question")
send_btn = gr.Button("Send")
# Call Analysis Tab
with gr.TabItem("🎙️ Call Analysis"):
with gr.Row():
with gr.Column():
call_upload = gr.Video(label="Upload Call Recording")
analyze_btn = gr.Button("Analyze Call")
with gr.Column():
with gr.Tabs():
with gr.TabItem("📝 Transcript"):
transcript = gr.Textbox(label="Transcript", lines=10)
with gr.TabItem("💡 Insights"):
insights = gr.Textbox(label="Key Insights", lines=10)
with gr.TabItem("🔊 Audio"):
audio_output = gr.Audio(label="Extracted Audio")
# Video Search Tab
with gr.TabItem("🎥 Video Search"):
with gr.Row():
with gr.Column():
video_upload = gr.File(label="Upload Video")
process_video_btn = gr.Button("Process Video")
video_status = gr.Textbox(label="Processing Status")
search_type = gr.Radio(
choices=["Text", "Image"],
label="Search Type",
value="Text"
)
text_input = gr.Textbox(label="Text Query")
image_input = gr.Image(label="Image Query", type="pil", visible=False)
search_btn = gr.Button("Search")
with gr.Column():
results_gallery = gr.Gallery(label="Search Results")
# Event Handlers
def document_chat(message, chat_history):
bot_message = DocumentProcessor.get_document_answer(message)
chat_history.append((message, bot_message))
return "", chat_history
def update_search_type(choice):
return {
text_input: gr.update(visible=choice=="Text"),
image_input: gr.update(visible=choice=="Image")
}
# Connect Events
process_docs_btn.click(
DocumentProcessor.process_documents,
inputs=[doc_files, chunk_size, chunk_type],
outputs=[process_status]
)
send_btn.click(
document_chat,
inputs=[msg, chatbot],
outputs=[msg, chatbot]
)
analyze_btn.click(
CallAnalyzer.process_call,
inputs=[call_upload],
outputs=[transcript, audio_output, insights]
)
process_video_btn.click(
VideoSearcher.process_video,
inputs=[video_upload],
outputs=[video_status]
)
search_type.change(
update_search_type,
search_type,
[text_input, image_input]
)
search_btn.click(
VideoSearcher.search_video,
inputs=[search_type, text_input, image_input],
outputs=[results_gallery]
)
# Related Pixeltable Spaces
gr.Markdown("## 🌟 Explore More Pixeltable Apps")
with gr.Row():
with gr.Column():
gr.HTML(
"""
<div style="border: 1px solid #ddd; padding: 15px; border-radius: 8px; margin-bottom: 10px;">
<h3>📚 Document & Text Processing</h3>
<ul style="list-style-type: none; padding-left: 0;">
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/Multi-LLM-RAG-with-Groundtruth-Comparison" target="_blank" style="color: #F25022; text-decoration: none;">
🤖 Multi-LLM RAG Comparison
</a>
</li>
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/Document-to-Audio-Synthesis" target="_blank" style="color: #F25022; text-decoration: none;">
🔊 Document to Audio Synthesis
</a>
</li>
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/Prompt-Engineering-and-LLM-Studio" target="_blank" style="color: #F25022; text-decoration: none;">
💡 Prompt Engineering Studio
</a>
</li>
</ul>
</div>
"""
)
with gr.Column():
gr.HTML(
"""
<div style="border: 1px solid #ddd; padding: 15px; border-radius: 8px; margin-bottom: 10px;">
<h3>🎥 Video & Audio Processing</h3>
<ul style="list-style-type: none; padding-left: 0;">
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/AI-Video-Analyzer-GTP4-Vision-TTS-Narration" target="_blank" style="color: #F25022; text-decoration: none;">
🎥 Video GPT Vision & TTS Narration
</a>
</li>
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/Call-Analysis-AI-Tool" target="_blank" style="color: #F25022; text-decoration: none;">
🎙️ Call Analysis Tool
</a>
</li>
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/object-detection-in-videos-with-yolox" target="_blank" style="color: #F25022; text-decoration: none;">
🔍 Video Object Detection
</a>
</li>
</ul>
</div>
"""
)
with gr.Column():
gr.HTML(
"""
<div style="border: 1px solid #ddd; padding: 15px; border-radius: 8px; margin-bottom: 10px;">
<h3>🎮 Interactive Applications</h3>
<ul style="list-style-type: none; padding-left: 0;">
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/AI-RPG-Adventure" target="_blank" style="color: #F25022; text-decoration: none;">
🎲 AI RPG Adventure
</a>
</li>
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/AI-Financial-Analysis-Platform" target="_blank" style="color: #F25022; text-decoration: none;">
📈 Financial Analysis Platform
</a>
</li>
<li style="margin-bottom: 10px;">
<a href="https://huggingface.co/spaces/Pixeltable/video-to-social-media-post-generator" target="_blank" style="color: #F25022; text-decoration: none;">
📱 Social Media Post Generator
</a>
</li>
</ul>
</div>
"""
)
gr.HTML(
"""
<div style="margin-top: 2rem; padding-top: 1rem; border-top: 1px solid #e5e7eb;">
<div style="display: flex; justify-content: space-between; align-items: center; flex-wrap: wrap; gap: 1rem;">
<div style="flex: 1;">
<h4 style="margin: 0; color: #374151;">🚀 Built with Pixeltable</h4>
<p style="margin: 0.5rem 0; color: #6b7280;">
Open Source AI Data infrastructure.
</p>
</div>
<div style="flex: 1;">
<h4 style="margin: 0; color: #374151;">🔗 Resources</h4>
<div style="display: flex; gap: 1.5rem; margin-top: 0.5rem;">
<a href="https://github.com/pixeltable/pixeltable" target="_blank" style="color: #4F46E5; text-decoration: none;">
💻 GitHub
</a>
<a href="https://docs.pixeltable.com" target="_blank" style="color: #4F46E5; text-decoration: none;">
📚 Documentation
</a>
<a href="https://huggingface.co/Pixeltable" target="_blank" style="color: #4F46E5; text-decoration: none;">
🤗 Hugging Face
</a>
</div>
</div>
</div>
<p style="margin: 1rem 0 0; text-align: center; color: #9CA3AF; font-size: 0.875rem;">
© 2024 Pixeltable | Apache License 2.0
</p>
</div>
"""
)
return demo
if __name__ == "__main__":
init_api_keys()
demo = create_interface()
demo.launch(
allowed_paths=[PIXELTABLE_MEDIA_DIR],
show_api=False
) |