Add guarded model selector (Diba-Base / Diba-mini)

Direct-link: fall back to CPU on any GPU failure (text-only)

Welcome text + voice progress + favicon logo

Welcome text + voice progress + favicon logo

Welcome text + voice progress + favicon logo

Welcome text + voice progress + favicon logo

Load Diba-Base with trust_remote_code (custom code)

UI v0.2 label

load Dibachain/Diba-Base (v0.2)

auth: parse username

whoami returns username string

Hugging Face sign-in (own ZeroGPU quota)

Hugging Face sign-in (own ZeroGPU quota)

Hugging Face sign-in (own ZeroGPU quota)

Hugging Face sign-in (own ZeroGPU quota)

UI: clarify random/unclear input

UI: fall back to CPU endpoint when ZeroGPU quota is exhausted

GPU time sized to reply length; CPU fallback endpoint

surface errors; code font + table fixes

surface errors; code font + table fixes

show real errors, disable SSR

UI: ZeroGPU user-token handshake + show server error details

UI: ZeroGPU user-token handshake + show server error details

Self-hosted UI assets, polished markdown, load public DibaAi/Diba-Base

UI: direct Gradio SSE adapter, bidi lists, no page overflow

ChatGPT-style Diba UI over ZeroGPU, streaming /chat_stream

never reveal base model

remove hint that leaked into replies; greedy default for accuracy

fix string literal; register reminder; lower default temperature

register reminder on user turn, lower default temperature

Persian replies in written register

Professional chat UI, reply in user's language, professional tone

ZeroGPU fast chat for Diba v0.1

initial commit