feat: add gTTS fallback to guarantee text-to-speech audio playback under rate limit

feat: optimize latency, implement global MutationObserver for instant audio play, and add user interrupt handlers for click and spacebar

feat: implement real-time continuous hands-free voice chat mode using Web Speech API VAD, global audio capture, and autoplay response

feat: add microphone voice input with automatic Groq Whisper STT transcription and TTS audio autoplay response

fix: change stream_to_file to write_to_file for new Groq SDK, sanitize all PDF text for Helvetica font

feat: add PDF export button, conversational welcome message, green theme, and verbose debug logging

debug: add verbose step-by-step logging and fallback for plain text LLM responses

fix: remove GEMINI_API_KEY guard that blocked all messages on HF Spaces

fix: change app.py to run standalone Gradio server, resolving styling issues on HF Spaces

fix: revert LLM backend to use the new active NVIDIA NIM API key

fix: switch to gemini-2.0-flash LLM model as primary backend to resolve unauthorized NVIDIA credentials

feat: simplify interface to a clean, simple chatbot layout

fix: restore button sizing to resolve layout breakage and non-clickable elements

feat: implement premium recruiter dashboard UI in Gradio

fix: apply global SVG styling constraints to prevent giant icons

fix chatbot avatar icon styling sizes

Implement Gradio Voice Input transcription via Gemini Audio upload API

Update system prompt to adopt consultative, human-like advisor tone explaining why each assessment is chosen

Mount Gradio UI onto FastAPI server in app.py to expose health and chat API endpoints

Switch agent model execution to NVIDIA NIM API running diffusiongemma-26b-a4b-it

Fix catalog json parser error: allow unescaped control characters in catalog_data.json

Revert agent to use Gemini API model execution to fix Transformers import crash and resource limitations on HF

Remove starlette version restriction to fix gradio 5 dependency clash

Upgrade Gradio SDK version to 5.12.0 in Space metadata for Python 3.13 support

Commit cleaned requirements.txt resolving faiss-cpu compatibility on Python 3.13

Downgrade starlette to <0.39.0 to match FastAPI dependencies and fix HF build pip install crash

Trigger rebuild with stable pinned Gradio SDK 4.44.0

Pin sdk_version: 4.44.0 in README.md to ensure stable Gradio build

Remove gradio from requirements.txt to avoid conflict with Hugging Face's pre-installed Gradio environment

Remove custom Dockerfile to let Hugging Face build native Gradio SDK space

Add app.py Gradio entrypoint at root level for Hugging Face Spaces

Replace Streamlit with Gradio UI and update HF Spaces metadata

Fix JSON parsing error: set strict=False to allow control characters like newlines in JSON strings

Force transformers upgrade in Dockerfile build steps to bypass caching

Force rebuild on Hugging Face to apply transformers upgrade

Upgrade transformers dependency in requirements.txt to >= 4.49.0 for Gemma 3 MultimodalLM support

Integrate Gemma 3 (AutoModelForMultimodalLM & Processor) into agent pipeline

Configure Dockerfile to run Streamlit UI on Hugging Face Spaces

Add approach document and local UI

SHL Assessment Recommender - Full implementation with hybrid retrieval, slot-based conversation analysis, safety guards, and evaluation harness