Fix and Improve PDF RAG Streamlit UI: Error Handling, Model Management, and Vector DB Updates

[shreeramdrao]

Summary:

• Fixed various issues in the PDF RAG Streamlit UI application.
• Improved error handling when uploading PDFs and interacting with the vector database.
• Enhanced logging to provide better insights into the process.
• Optimized model handling, ensuring only valid models are used.
• Added a zoom slider for better interaction with PDF images.
• Cleaned up session state variables when deleting the vector database.

Key Changes:

• Refactored the create_vector_db function to improve PDF processing and chunk splitting.
• Updated the process_question function for better model interaction and response formatting.
• Added logic to handle missing or invalid models and user inputs more gracefully.

This update improves the overall functionality and user experience of the application.

[shreeramdrao]

Summary:

• Fixed errors related to vector DB creation and model name extraction.
• Resolved issues where the model names were not being extracted properly from the response.
• Ensured proper handling of missing 'name' keys in model information.

Changes Made:

• Corrected how models are processed and extracted by checking for both 'name' and fallback options ('model', 'parent_model').
• Improved error handling to ensure the code does not crash when encountering missing fields in the response.
• Enhanced logging for better debugging and traceability.

How to Test:

• Upload a PDF file and check if the vector DB is created properly.
• Select a model from the dropdown and ensure the question is processed successfully.
• Test the system with both valid and invalid PDF files to verify error handling.

Additional Notes:

• The code has been thoroughly tested, and logging has been improved to assist in future debugging.
• I believe this addresses the issue mentioned in [issue number] (if applicable).