Spaces:
Runtime error
Runtime error
Commit History
-
Update project video section in README.md to replace embedded YouTube link with a clickable thumbnail for improved user engagement and accessibility.
16689e0 -
Add project video section to README.md with embedded YouTube link for enhanced project visibility and user engagement.
76fb1e7 -
Enhance error handling and logging in subtitle and thumbnail download functions in app.py. Update return values to ensure compatibility with Gradio components, improving user feedback during file download processes. Initialize variables for better error management in video processing.
9eb4184 -
Refactor YouTube research results formatting in app.py for improved clarity and conciseness. Update download functions for subtitles and thumbnails to enhance error handling and ensure compatibility with Gradio. Modify ContentEnhancementAgent to generate only optimized titles using video summaries as descriptions, streamlining the content enhancement process.
8ffd056 -
Enhance ContentAnalysisAgent with improved prompt template and analysis approach. Added detailed tasks for video summary creation, search term optimization, and content categorization. Updated output format for clarity and added validation checks for search terms and summary length to ensure robust analysis.
3a343e7 -
Add video summary placeholder and output in process_video function. Enhance UI with video summary display in create_interface function for improved user experience during video processing.
f303731 -
Implement fallback enhancement method in ContentEnhancementAgent for improved error handling. This addition generates a fallback title and description when LLM enhancement fails, utilizing keywords from the video summary and research data. Enhanced input validation and JSON parsing checks are also included to ensure robustness during content enhancement.
d9a831a -
debug: try - catch
f3768c3 -
Enhance video processing capabilities in app.py and modal_functions.py. Added user warnings for large video files, increased timeout and memory limits for processing, and improved debugging output for large file handling. These changes aim to optimize performance and user experience during video uploads and processing.
f27eaee -
fix UI change
26c1bdf -
some improvment stuff
dadea03 -
Enhance thumbnail generation capabilities in app.py and thumbnail_strategy_agent.py. The generate_thumbnail_images function now accepts additional parameters for search terms, transcript segments, and analysis results, allowing for more context-aware thumbnail strategies. A new extract_key_moments function has been added to identify important segments from transcripts. The ThumbnailStrategyAgent has been updated to generate three distinct thumbnail strategies, incorporating detailed visual psychology principles and competitive insights for improved click-through rates. Modal functions have been adjusted to support the new strategy format, ensuring backward compatibility while utilizing enhanced strategies directly.
ab3342a -
Add architecture documentation for SmartThumbnailer project
8152429 -
Enhance ThumbnailStrategyAgent and modal_functions for improved thumbnail generation. The ThumbnailStrategyAgent now supports weighted prompting and generates a single comprehensive base strategy with enhanced visual styles. The modal_functions have been updated to include new libraries for weighted prompting and to create mixed strategy prompts based on content categories. Additionally, error handling and fallback mechanisms for prompt generation have been improved, ensuring more reliable thumbnail outputs.
046e113 -
Deleting unnecessary code
41ac6c9 -
Revise content analysis prompt and enhance fallback analysis in content_analysis_agent.py. The prompt now includes detailed categorization guidelines for tech, news, comedy, and other categories, along with confidence scoring instructions. A new fallback analysis method has been added to improve handling of LLM failures, utilizing keyword detection for category assignment and generating relevant search terms. This update aims to enhance the accuracy and reliability of video content categorization.
dc12a4e -
Enhance error handling and debugging output in app.py and youtube_research_agent.py. This update includes checks for empty or short transcripts, improved handling of research term results, and the addition of competitive analysis generation. These changes aim to improve reliability and visibility into the content processing workflow.
15d538d -
Remove unused streaming functions and enhance debugging output in app.py and content_analysis_agent.py. Added critical debug statements for transcript content and implemented checks for very short transcripts to improve content analysis reliability.
79dcfa0 -
Enhance debugging output in content_analysis_agent.py by adding detailed logs for LLM calls, response receipt, and JSON parsing. This update improves visibility into the content analysis process, aiding in troubleshooting and ensuring successful JSON parsing.
8a33c2a -
Enhance debugging output in process_video function to include transcript preview, summary preview, and search terms. This update improves visibility into content analysis steps, aiding in troubleshooting and performance monitoring.
78c0399 -
Enhance debugging and performance tracking across audio extraction, transcription, content analysis, YouTube research, content enhancement, and thumbnail generation processes. This update introduces detailed timing logs and status messages to monitor execution times and improve error handling. Additionally, it refines the overall structure of the processing functions for better clarity and maintainability.
7c14cc6 -
Enhance Ollama server readiness checks and improve HTTP API interaction in modal_ollama.py. This update introduces a retry mechanism to verify server responsiveness before making API calls, ensuring more reliable operation. Additionally, it refines error handling for API failures and adjusts timeout settings for better performance during response generation. The fallback to command line interface is maintained for robustness.
b0b0b15 -
Update model configuration to use 'llama3:instruct' across all relevant files, including README.md, agent scripts, and caching mechanisms. Adjust performance settings in modal_ollama.py for the new model, optimizing memory and timeout parameters for improved efficiency. Update documentation to reflect changes in model caching and expected improvements in LLM capabilities.
442567f -
Enhance error handling and server management in modal_ollama.py
This commit improves the robustness of the Ollama server management by initializing the server_process variable to None, preventing UnboundLocalError. It also adds error handling for volume reload failures, allowing the process to continue without critical interruptions. Cleanup procedures for terminating the server process are updated to ensure safe termination, even if the process is already dead. These changes aim to enhance the stability and reliability of the model deployment.69c0ae9 -
Adding streaming to the generated content so it shows on the app as it is generated.
edbba01 -
Add test_query function to modal_ollama.py for real content analysis with cached model
This commit introduces a new function, test_query, which tests the cached Ollama model with a real query. It includes server management, model availability checks, and HTTP API interactions to validate the model's performance. The function aims to ensure the model is ready for deployment by providing structured output and error handling.07ec0f1 -
Update modal_ollama.py with improved Ollama model handling
a1a0e06 -
Update modal_ollama.py to replace min_containers with scaledown_window, allowing containers to remain warm for 5 minutes after use. This change aims to optimize resource management and response times for the deployment.
c6f31f4 -
Refactor caching mechanism in cache_ollama.py to utilize Modal function retrieval for improved clarity and error handling. Update error messages to guide users on deployment verification.
2958bff -
Restructuring documentation for the fine-tuned models documentation.
2ff44ea -
Update README.md and modal deployment scripts to reflect GPU changes for improved performance. Replace A10G with L40S for Whisper transcription and H200 for SDXL and Ollama model caching. Add new function `generate_thumbnails` to README.md. Adjust memory and timeout settings accordingly.
ffff3f7 -
Update model configuration to use 'llama3.1:70b-instruct-q6_K' across all relevant files, including README.md, agent scripts, and caching mechanisms. Enhance performance settings in modal_ollama.py for the new model, adjusting memory and timeout parameters accordingly. Update documentation to reflect changes in model caching and expected improvements in LLM capabilities.
0a5eba6 -
Refactor ContentAnalysisAgent to streamline prompt structure and enhance error handling. Simplify the prompt template for JSON output and implement a fallback analysis mechanism for timeout errors, including basic category detection and search term generation based on transcript content. Update timeout settings in modal_ollama.py to accommodate complex JSON generation.
8b1a737
-
Refactor file handling in app.py to utilize named temporary files for serving subtitles and thumbnails with Gradio. This change simplifies the file creation process, enhances compatibility, and improves logging for file operations. Update return values to reflect the new file handling approach.
2e42a24 -
Refactor file handling in app.py to enhance persistence for downloaded subtitles and thumbnails. Replace temporary file creation with a more durable approach using temporary directories, ensuring proper permissions and improved logging for file operations. Update return values to reflect new file paths.
d82474a -
Enhance file handling in app.py and modal_functions.py by implementing persistent storage for subtitles and thumbnails. Update download and upload endpoints to utilize Modal storage effectively, including improved logging for file operations. Refactor UI components for better user experience with subtitle and thumbnail downloads, ensuring visibility and status updates are clear.
8264910 -
Refactor thumbnail generation logic to implement STRICT MODE for content analysis. Update function signatures and internal logic to enforce category validation and confidence thresholds. Introduce new documentation for fine-tuned model loading and operational constraints, ensuring only premium models are utilized without fallbacks.
1342ba7 -
Update README.md and agent files to integrate Modal Ollama for content analysis and enhancement. Change model configuration to use 'llama3:instruct' and add caching instructions for improved performance. Refactor agent classes to utilize ModalOllamaLLM, enhancing LlamaIndex settings and streamlining model initialization.
86d8dd3 -
Refactor model loading logic in modal_functions.py to improve caching mechanism for Whisper turbo model. Enhance error handling and provide clearer feedback during model loading and caching processes. Ensure fallback to downloading fresh models when no cached files are available.
b21d6d6 -
Enhance performance by implementing caching for the Whisper turbo model to optimize audio transcription. Update README.md with instructions for one-time model caching setup and document performance benefits. Add caching functions in modal_functions.py to streamline model loading and improve user experience during transcriptions.
0e6ddab -
Update audio transcription to use OpenAI Whisper turbo model for improved performance and accuracy. Adjust comments to reflect the model change.
f6775ec -
Add prompt truncation for CLIP compatibility and enhance error handling in upload endpoints
- Introduced `truncate_prompt_for_clip` function to ensure prompts fit within CLIP's token limit, improving compatibility. - Updated `upload_subtitle_endpoint` and `upload_thumbnail_endpoint` to validate input requests and handle errors more gracefully, raising appropriate HTTP exceptions for invalid data. - Enhanced security checks in `download_file_endpoint` to prevent unauthorized file access, ensuring the file path is within the designated storage directory.06bb31d -
Remove cache_model.py script and update references in README.md and performance-optimization.md to reflect the new path for model caching. This change ensures consistency in instructions for initializing the SDXL model cache using Modal storage.
1a92033 -
Implement caching for the SDXL model to optimize thumbnail generation performance. Update README.md with instructions for one-time model caching setup. Enhance modal_functions.py with functions to download and cache the SDXL model, ensuring faster subsequent loads. Adjust thumbnail generation logic to utilize cached model, improving efficiency.
abf90a0 -
Refactor app.py to replace Supabase storage with Modal storage for file handling. Update functions for subtitle and thumbnail uploads, downloads, and storage checks. Revise README.md to reflect changes in storage requirements and setup instructions. Remove SupabaseStorage class and related documentation, ensuring a streamlined integration with Modal for improved performance and user experience.
e058213 -
Enhance SupabaseStorage class with improved error handling and user feedback during client initialization and file uploads. Add detailed logging for bucket existence checks, subtitle, and thumbnail uploads, including success and error messages. Ensure graceful handling of bucket creation failures and provide informative output for upload processes.
4372cfe -
Enhance Supabase integration in app.py and modal_functions.py by implementing thumbnail and subtitle uploads. Add error handling for storage initialization and upload processes, improving user feedback on upload status. Remove deprecated Supabase client initialization code and streamline thumbnail processing logic.
6851496 -
Remove unused clear_user_data function from modal_functions.py and update app.py to reflect changes in data clearing process, focusing on clearing old files from Supabase storage. Enhance comments for clarity.
bf5a8b5 -
Update Supabase dependency in modal_functions.py from version 2.9.1 to 2.15.2 to ensure compatibility and access to the latest features.
f03ff44 -
Refactor app.py to remove Modal storage fallback for subtitle and thumbnail downloads, ensuring reliance on Supabase for file retrieval. Update README.md and documentation to emphasize the necessity of Supabase configuration for proper functionality. Enhance user feedback in the interface regarding storage status and file availability.
4983d68