Neura AI v0.1.9 - RAG Similarity | Initial Query Triggers Added | FE Improvements

Our latest update brings significant enhancements to our RAG, Image Generation and frontend.

BACKEND

Sequential Image Generation: The prompt clearly specifies that if multiple images are requested, each will be generated individually. This ensures that each image receives focused attention and resources, improving the quality and relevance of each generated image.
Avoidance of Composite Images: By instructing the system to prevent the creation of composite images, the prompt guarantees that each request results in distinct and separate visuals. This avoids confusion and maintains the integrity of the user's individual creative visions.
Feedback Mechanism: After each image is generated, the system provides feedback on how the user's prompt was interpreted. This educational feature helps users refine their future prompts for better outcomes and enhances their understanding of how the AI interprets input.
No Text in Images: The directive to exclude text from the images ensures that the visuals remain clean and focused on the imagery alone, enhancing the visual impact and artistic quality of the generated content.

Cosine Similarity Calculation: We have refactored the cosine similarity calculation to improve its accuracy and performance. The new implementation utilizes the NumPy library to optimize vector operations, resulting in faster and more reliable similarity calculations.
Content Retrieval: We have enhanced the content retrieval functionality to retrieve the most similar content based on the calculated cosine similarity. This update enables more accurate and relevant content suggestions, improving the overall user experience.
Code Organization: We have reorganized the code to improve maintainability and readability. The Embedding Module now follows a more modular architecture, making it easier to update and extend its functionality.
Error Handling: We have implemented robust error-handling mechanisms to ensure that the Embedding Module can handle unexpected errors and exceptions gracefully.

Initial Prompt Triggers: New prompt triggers added to FANA LLM Frontend
Implemented 3-Dot Loading Animation: Replaced the static "Thinking..." message with a dynamic 3-dot loading animation, enhancing the visual appeal and user engagement.
CSS Animation: Utilized CSS keyframe animation to create a smooth, infinite animation cycle for the loading dots, ensuring a consistent and responsive user experience.
Dot Spacing and Alignment: Adjusted the margin and padding properties to optimize the spacing between the loading dots, creating a visually appealing and harmonious animation.
Animation Synchronization: Implemented animation delays to synchronize the movement of each dot, creating a staggered, one-by-one animation effect that enhances the overall visual appeal.
Container Alignment: Applied text-align and float properties to ensure the loading dots are aligned to the left side of the container, maintaining a clean and organized visual layout.

Last updated 7 months ago