Upload 2 files

app.py

@@ -0,0 +1,301 @@
+import os
+import gradio as gr
+import tempfile
+from src.mainFunctions import process_video_file, process_audio_document, process_youtube_video
+from src.video_processing import extract_audio_from_video
+from src.documentProcessing import process_document
+
+ELEVENLABS_API_KEY = os.environ.get("ELEVENLABS_API_KEY", None)
+GOOGLE_API_KEY = os.environ.get('GOOGLE_API_KEY', None)
+
+with gr.Blocks(title="Document & Media Processing Tool") as app:
+    gr.Markdown("# Document & Media Processor")
+    gr.Markdown("Upload a document, video, audio, or provide a YouTube link to generate summaries and quizzes.")
+    
+    with gr.Row():
+        with gr.Column():
+            elevenlabs_api_key = gr.Textbox(
+                placeholder="Enter your ElevenLabs API key",
+                label="ElevenLabs API Key (for transcription)",
+                type="password",
+                value=ELEVENLABS_API_KEY
+            )
+            model_id = gr.Dropdown(
+                choices=["scribe_v1"],
+                value="scribe_v1",
+                label="Transcription Model"
+            )
+            gemini_api_key = gr.Textbox(
+                placeholder="Enter your Google Gemini API key",
+                label="Google Gemini API Key",
+                type="password",
+                value=GOOGLE_API_KEY
+            )
+
+    with gr.Row():
+        with gr.Column():
+            language_selector = gr.Radio(
+                choices=["Uzbek", "English", "Russian"],
+                value="English",
+                label="Content Language"
+            )
+
+    with gr.Tabs():
+        with gr.TabItem("Upload Document"):
+            with gr.Row():
+                with gr.Column():
+                    document_input = gr.File(label="Upload Document", file_types=[".pdf", ".docx", ".txt"])
+                    with gr.Row():
+                        generate_summary_button = gr.Button("Generate Summary")
+                        generate_quiz_button = gr.Button("Generate Quiz")
+                with gr.Column():
+                    document_status_output = gr.Textbox(label="Document Processing Status")
+                    document_text_file_output = gr.File(label="Extracted Text File")
+            with gr.Row():
+                with gr.Column():
+                    document_output = gr.Textbox(
+                        label="Generated Content",
+                        lines=15
+                    )
+                    with gr.Row():
+                        document_file_output = gr.File(label="Download Text File")
+                        document_json_file_output = gr.File(label="Download JSON File")
+        
+        # Video Upload Tab
+        with gr.TabItem("Upload Video"):
+            with gr.Row():
+                with gr.Column():
+                    video_input = gr.Video(label="Upload Video")
+                    format_choice_file = gr.Radio(["mp3", "wav"], value="mp3", label="Audio Format")
+                    with gr.Row():
+                        video_summary_button = gr.Button("Generate Summary")
+                        video_quiz_button = gr.Button("Generate Quiz")
+                with gr.Column():
+                    audio_output_file = gr.Audio(label="Extracted Audio", type="filepath")
+                    status_output_file = gr.Textbox(label="Audio Extraction Status")
+                    transcript_file_output = gr.File(label="Transcription Text File")
+                    transcript_status_output = gr.Textbox(label="Transcription Status")
+            with gr.Row():
+                with gr.Column():
+                    video_output = gr.Textbox(
+                        label="Generated Content",
+                        lines=15
+                    )
+                    with gr.Row():
+                        video_text_file_output = gr.File(label="Download Text File")
+                        video_json_file_output = gr.File(label="Download JSON File")
+        
+        # YouTube Tab
+        with gr.TabItem("YouTube Video"):
+            with gr.Row():
+                with gr.Column():
+                    youtube_url = gr.Textbox(
+                        placeholder="Enter YouTube URL",
+                        label="YouTube URL"
+                    )
+                    yt_format_choice = gr.Radio(["mp3", "wav"], value="mp3", label="Audio Format")
+                    with gr.Row():
+                        youtube_summary_button = gr.Button("Generate Summary")
+                        youtube_quiz_button = gr.Button("Generate Quiz")
+                with gr.Column():
+                    yt_audio_output = gr.Audio(label="Extracted Audio", type="filepath")
+                    yt_status_output = gr.Textbox(label="YouTube Processing Status")
+                    yt_transcript_file_output = gr.File(label="Transcription Text File")
+                    yt_transcript_status_output = gr.Textbox(label="Transcription Status")
+            with gr.Row():
+                with gr.Column():
+                    youtube_output = gr.Textbox(
+                        label="Generated Content",
+                        lines=15
+                    )
+                    with gr.Row():
+                        youtube_text_file_output = gr.File(label="Download Text File")
+                        youtube_json_file_output = gr.File(label="Download JSON File")
+        
+        # Audio Tab
+        with gr.TabItem("Upload Audio"):
+            with gr.Row():
+                with gr.Column():
+                    audio_input = gr.Audio(label="Upload Audio", type="filepath")
+                    with gr.Row():
+                        audio_summary_button = gr.Button("Generate Summary")
+                        audio_quiz_button = gr.Button("Generate Quiz")
+                with gr.Column():
+                    audio_status_output = gr.Textbox(label="Audio Processing Status")
+                    audio_transcript_file_output = gr.File(label="Transcription Text File")
+            with gr.Row():
+                with gr.Column():
+                    audio_output = gr.Textbox(
+                        label="Generated Content",
+                        lines=15
+                    )
+                    with gr.Row():
+                        audio_text_file_output = gr.File(label="Download Text File")
+                        audio_json_file_output = gr.File(label="Download JSON File")
+
+    # Document processing
+    generate_summary_button.click(
+        fn=process_document,
+        inputs=[
+            document_input,
+            gemini_api_key,
+            language_selector,
+            gr.State("summary") 
+        ],
+        outputs=[
+            document_status_output,
+            document_text_file_output,
+            document_output,
+            document_file_output,
+            document_json_file_output
+        ]
+    )
+    
+    generate_quiz_button.click(
+        fn=process_document,
+        inputs=[
+            document_input,
+            gemini_api_key,
+            language_selector,
+            gr.State("quiz") 
+        ],
+        outputs=[
+            document_status_output,
+            document_text_file_output,
+            document_output,
+            document_file_output,
+            document_json_file_output
+        ]
+    )
+    
+    # Video processing
+    video_summary_button.click(
+        fn=process_video_file,
+        inputs=[
+            video_input, 
+            format_choice_file, 
+            elevenlabs_api_key, 
+            model_id, 
+            gemini_api_key,
+            language_selector,
+            gr.State("summary") 
+        ],
+        outputs=[
+            audio_output_file, 
+            status_output_file, 
+            transcript_file_output,
+            transcript_status_output, 
+            video_output, 
+            video_text_file_output, 
+            video_json_file_output
+        ]
+    )
+    
+    video_quiz_button.click(
+        fn=process_video_file,
+        inputs=[
+            video_input, 
+            format_choice_file, 
+            elevenlabs_api_key, 
+            model_id, 
+            gemini_api_key,
+            language_selector,
+            gr.State("quiz")  
+        ],
+        outputs=[
+            audio_output_file, 
+            status_output_file, 
+            transcript_file_output,
+            transcript_status_output, 
+            video_output, 
+            video_text_file_output, 
+            video_json_file_output
+        ]
+    )
+    
+    # YouTube processing
+    youtube_summary_button.click(
+        fn=process_youtube_video,
+        inputs=[
+            youtube_url,
+            yt_format_choice,
+            elevenlabs_api_key,
+            model_id,
+            gemini_api_key,
+            language_selector,
+            gr.State("summary")  
+        ],
+        outputs=[
+            yt_audio_output,
+            yt_status_output,
+            yt_transcript_file_output,
+            yt_transcript_status_output,
+            youtube_output,
+            youtube_text_file_output,
+            youtube_json_file_output
+        ]
+    )
+    
+    youtube_quiz_button.click(
+        fn=process_youtube_video,
+        inputs=[
+            youtube_url,
+            yt_format_choice,
+            elevenlabs_api_key,
+            model_id,
+            gemini_api_key,
+            language_selector,
+            gr.State("quiz")  
+        ],
+        outputs=[
+            yt_audio_output,
+            yt_status_output,
+            yt_transcript_file_output,
+            yt_transcript_status_output,
+            youtube_output,
+            youtube_text_file_output,
+            youtube_json_file_output
+        ]
+    )
+    
+    # Audio processing
+    audio_summary_button.click(
+        fn=process_audio_document,
+        inputs=[
+            audio_input,
+            elevenlabs_api_key,
+            model_id,
+            gemini_api_key,
+            language_selector,
+            gr.State("summary") 
+        ],
+        outputs=[
+            audio_status_output,
+            audio_transcript_file_output,
+            audio_output,
+            audio_text_file_output,
+            audio_json_file_output
+        ]
+    )
+    
+    audio_quiz_button.click(
+        fn=process_audio_document,
+        inputs=[
+            audio_input,
+            elevenlabs_api_key,
+            model_id,
+            gemini_api_key,
+            language_selector,
+            gr.State("quiz") 
+        ],
+        outputs=[
+            audio_status_output,
+            audio_transcript_file_output,
+            audio_output,
+            audio_text_file_output,
+            audio_json_file_output
+        ]
+    )
+
+if __name__ == "__main__":
+    app.launch(share=True, debug=True)

requirements.txt

@@ -0,0 +1,15 @@
+gradio
+uuid
+yt-dlp
+python-dotenv
+requests
+sentence-transformers
+langchain-google-genai
+anthropic
+pydantic
+transformers==4.50.3
+langchain-openai
+PyMuPDF
+python-docx
+pytube
+PyPDF2