Update image input type in app.py from PIL to numpy for improved compatibility. Modify Dockerfile to install specific versions of Gradio and Gradio-client to prevent schema TypeError crashes, ensuring a stable environment for the application.

Refactor Dockerfile to optimize llama-cpp-python installation by using a pre-built CPU wheel, reducing build time significantly. Removed unnecessary system packages and streamlined app requirements for clarity.

Add Dockerfile for containerized deployment, update README and requirements.txt

Update llama-cpp-python version in requirements.txt to 0.3.16 for improved compatibility with Qwen3.5 VL support.

Integrate vision-language (VL) support for Qwen3.5 models by adding optional VL filenames and paths in model configurations. Update model loading logic to handle VL models, enabling image input processing. Adjust requirements.txt to specify llama-cpp-python version for VL compatibility. Enhance logging for model loading status and VL capability.

Enhance model loading in app.py to support vision projector integration, allowing for image input handling. Update add_message and respond functions to accommodate optional image parameters and improve message formatting. Refactor logging for better clarity on message processing and duplicate detection.

Update requirements.txt to include global pip options for llama-cpp-python, ensuring it is installed from a pre-built CPU wheel index to enhance startup efficiency and eliminate source compilation.

Refactor app.py to remove subprocess-based package installations for llama-cpp-python and duckduckgo-search, improving startup efficiency. Introduce background model preloading for faster initial user queries. Update README.md to reflect a reduced startup duration timeout. Adjust requirements.txt to streamline dependencies and remove unused packages.

Refactor model loading and display name generation in app.py; update model repository mapping in download_models.py to include new Qwen3.5 models with enhanced display names. Enhance domain mapping configuration with keyword weights and negative keywords for improved context handling across various domains.

Update CSS for input containers and typing dots to improve layout and visual consistency. Adjust padding values and enhance typing dot dimensions for better display in chatbot interface.

Enhance typing dots animation and improve chatbot auto-scroll functionality. Update message formatting for better visual consistency. Inject current date into prompts for accurate context in responses. Add new market-related keywords to agent configuration.

Update Gradio launch config for HF Spaces

Fix typing dots animation - use inline-block for transform support

Enable share button and improve typing dots animation

Fix typing dots to display inline with meeTARA text

Fix typing indicator - show animated dots while processing

Fix hallucinated follow-up questions in responses

Fix Gradio/huggingface-hub compatibility issue

Enhance agent response handling by implementing tool detection for calculator and web search functionalities. Streamline response generation with improved prompt instructions for news queries. Upgrade transformers library for better performance and compatibility. Adjust token streaming limits to allow for more comprehensive responses while ensuring cleaner output.

Reduce max tokens for streaming responses to improve performance and user experience. Enhance control token cleaning in response generation to prevent leakage and ensure cleaner output. Update CSS for better text visibility and padding adjustments in the chat interface.

Implement tool detection and streaming response for improved agent interaction. Refactor response handling to check for tool needs before executing, and stream responses for better UI responsiveness. Enhance error handling with streaming feedback when agent mode fails.

Enhance input styling by removing borders and outlines for a cleaner look. Adjust message width to utilize full available space and ensure text wraps naturally. Disable typing dots for a more streamlined chat interface.

Reverse message order in chat interface to display bot responses above user requests. Update styling to ensure all message elements have transparent backgrounds and no borders, enhancing visual clarity and consistency.

Refactor message styling to remove visual containers and enhance text clarity. All message elements now have transparent backgrounds, no borders, and consistent padding, improving the overall chat interface appearance.

Add modern auto-hide scrollbar styling for various components, enhancing UI consistency and aesthetics. Adjust spacing in message pair containers for a tighter layout.

Fix indentation in web search logging for clarity in MeeTARAAgent class

Remove QUICK_START.md and reorganize documentation structure; add detailed deployment guide for Hugging Face Spaces and performance optimization documentation for agent mode.

Wider model selector to show meeTARA fully

Increase model selector width, bigger preprompt icons with more spacing

Fix: remove spacer completely, model selector left-aligned after icons, smaller icons

Fix: model selector moved left (remove flex spacer), mobile icons more compact

Mobile fix: reduce padding, move icons up, model selector closer to icons

Move send button next to input box, preprompt icons + model on bottom row

Fix mobile layout - shift input up, ensure send button and model selector visible

Fix popup close - clear choices on selection to force hide

Fix popup auto-close - more aggressive hiding after selection

Fix popup issues - hide empty gray box and auto-close popup after selection

Fix popup visibility - use position:fixed and higher z-index to avoid chatbot blocking

Fix preprompt popup - correct indentation level (outside input-box-gemini, not inside)

Restore preprompt popup with multiple questions like commit 0321f68

Simplify preprompt - click icon fills input with random prompt

Fix preprompt popup - move inside input-box-gemini for proper absolute positioning

Fix preprompt popup - use Radio component with proper event handlers

Fix preprompt popup - use native Gradio buttons for reliable popup

Fix preprompt popup - replace Gradio Radio with pure JavaScript popup for reliability

Fix preprompt popup CSS - increase z-index and ensure visibility

Fix Chatbot label - add show_label=False and container=False to hide label completely

Fix Chatbot label - use JavaScript to hide it since CSS selectors don't work with Gradio dynamic classes

Fix preprompt popup - initialize Radio with placeholder, hide chatbot label more aggressively

Fix mobile UI - hide chatbot label/icon, ensure send button and model selector visible