dineshupadhya/docqa-chatbot

Update frontend/requirements.txt

Update frontend/app.py

docs: add YAML frontmatter to README

docs: update README with v3.0 features

fix: move voice_query handler to tab level outside c1 column

fix: move voice_query handler outside VOICE_AVAILABLE block, show suggestions in chat when empty

feat: streaming answers, query suggestions after upload, OpenAI key input in sidebar

v3.0: multi-model, streaming, similarity viz, doc compare, chat persistence, voice output, highlighted sources

fix: simplify prompt for flan-t5-small - minimal format works better with small models

fix: replace broken pipeline(text2text-generation) with custom FlanT5LLM class using model.generate() directly

fix: comprehensive error handling - backend returns JSON on all errors, LLM status endpoint, frontend handles all failure modes

fix: increase timeout to 300s, add JSON decode error handling, show warming up message

fix: return friendly message when no docs uploaded, add status indicator, clean dead code

fix: rewrite voice input - pass file-like object directly to sr.AudioFile, use session_state.pop

fix: remove st.rerun() from inside chat_message block that caused errors

Auto-send voice input to chat after transcription

Add Send Voice Question button to actually ask the transcribed question

Fix: remove pyaudio (not needed for st.audio_input)

Force Docker rebuild: pin streamlit>=1.33, add SpeechRecognition

Force rebuild: voice input fix

Fix: server-side voice input with SpeechRecognition + st.audio_input

Fix: standalone voice input component with Web Speech API

Fix: implement real working voice input with Web Speech API

Deploy DocQA Chatbot with all features

initial commit