removed commit message

Fix: catch Firebase init errors so chat works even if credentials are invalid

Re-apply Firebase Firestore lead storage integration

Add Firebase Firestore integration for persistent lead storage

Add full contact details (address) to bot responses and contact rule

Groq model update and environment variable override

Fix: check error type before printing traceback — no noise for known errors

Fix: handle network/connection errors gracefully instead of 500 crash

Update widget.js to match new UI + add /leads endpoint for captured contacts

Add multi-key Groq rotation, session cleanup, smarter FAISS chunking

Fix: correct local port in docstring to 8500

Fix: use system+human message structure to stop prompt leaking in responses

Fix: handle greetings naturally instead of treating them as out-of-scope

Fix: stop greeting user on every reply

Remove office hours from all bot responses

Fix host: bind to 0.0.0.0 so HuggingFace can reach the app

Simplify to single Groq API key

Fix UnboundLocalError: declare _retriever,_prompt global in chat()

Speed up startup: lazy-load embedding model on first request

Fix SyntaxError: move global declaration to top of chat function

Add multi-key Groq fallback + embeddable widget.js for vibhusolutions.com

url update

Revert to Groq: OpenAI quota exceeded, use llama-3.1-8b-instant free tier

Switch LLM: Groq -> OpenAI gpt-4o-mini

Fix rate limit: switch to llama-3.1-8b-instant (500K TPD), handle 429 gracefully, clean hero page

Fix agent: correct business name, stop history leakage, smart lead capture

Add Vibhu Solutions AI Agent

ui update

fist update