feat(fact_image): add Bengali text auto-detection and font support

feat: Ultrafast preset for all modules, Text Story margin fix, Quiz Explain box improvements

perf: Use ultrafast preset for video export - 3-5x faster

perf: Remove normalize_video step - saves 50-90 sec per scene

feat: Major optimizations - Quiz dynamic fonts, TTS 1.2x speed, Video stream copy (10x faster), Single API call, Fact Image dynamic fonts, Text Story position fix

style: Increase fact image text padding by 8px on each side

feat: Add voice dropdown to Quiz section with comprehensive voice options

feat: Update Video Creator voice dropdown with comprehensive voice list

feat: Update voice dropdowns with 60+ voices organized by language groups

fix: Login system - CSS specificity fix and debug logging

feat: Add client-side login system with username/password authentication

docs: Add comprehensive API documentation with HF Cloud Storage section

fix: Text Story UI positioning, Fact Image heading background alignment

feat: Text Story - Facebook Messenger UI with header, avatars, Active Now status

feat: Multiple updates - HTML optimization, dual API search, text story tuning, Messenger UI

fix: Better error handling for fact-image validation errors

perf: Optimize HTML with external CSS/JS - faster loading

chore: Remove YouTube uploader (not working in cloud Docker)

feat: YouTube to HF uploader - folder management, video/audio modes, category naming, text_story UI update

feat: New dating-app style chat UI with gradient header, avatar, Online status

fix: Make message sender field more flexible to fix 422 error

fix: Green screen fallback + auto create gameplay_backgrounds folder in HF Dataset

feat: Natural chat AI, TTS speed 1.3x, HF upload to text_story folder

fix: Sync audio with message timing using CompositeAudioClip

fix: Correct TTS endpoint and payload format, filter .mp4 only

fix: Use GROQ_API instead of GROQ_API_KEY

feat: Add Text Story module with iMessage-style chat video generation

feat: Add Groq GPT-OSS-120B fallback for data formatting with exact CSV spec

trigger: Force HuggingFace rebuild with improved year parsing

fix: Improve year parsing with flexible extraction and better debugging

fix: Convert SystemMessage to HumanMessage for Gemma compatibility

fix: Pin langchain-google-genai version for compatibility

feat: Add rate limiter (29 req/62s) + use gemma-3-27b-it everywhere

fix: Remove kw_only for Python 3.9 compatibility

trigger: Force HuggingFace rebuild

fix: Remove all searxng references from graph.py and configuration.py

fix: Remove SearXNG/langchain-community dependency - use Tavily only

fix: Add langchain-community to requirements.txt

feat: Complete Agentic AI Pipeline for Bar Chart Race

Fix Validator crash on wide data and enable dynamic tavily search

Fix Validator incorrectly renaming entity columns to value

Add AI Knowledge gen and Tavily integration with key

Fix IndentationError in Scout scraping logic

Fix Scout to strictly filter scraped tables for Time Series (valid Bar Race data) only

Fix Surgeon failure by adding Transposed Wide format support in Validator

Fix HF upload bug in router.py

Fix Validator to correctly detect wide format and prevent bad renaming

Add Brain plan validation and merge with fallback defaults

Add topic-based fallback indicator detection in Scout

Fix Validator column detection, smooth racing animation, error handling