Spaces:
Running on Zero
Running on Zero
Commit History
Yesterday
-
ChatUI Studio v2.0 — FastAPI + Transformers, multi-model / multi-task / agent
Original DeepSeek-Gradio space rewritten to run transformer models locally. Backend (app.py): * Replaced gr.Server + openai/SiliconFlow client with FastAPI + uvicorn * Provider is now transformers.AutoModelForCausalLM (no API key) * 5-model registry: SmolLM2-360M, Qwen2.5-0.5B/1.5B, TinyLlama-1.1B, Phi-3.5-mini * 5-task router: chat, code, summarize, translate, agent (each w/ system prompt) * Agent mode: ReAct loop using local model + 4 dependency-free tools (calculator, web_search stub, file_read sandboxed, now) * Endpoints: /api/models /api/tasks /api/load_model /api/run /healthz * Lazy model cache + chat template support + <think> block extraction Frontend (static/): * Kept original layout (history drawer, studio dashboard, chat thread, settings) * Added header model-picker pill with dropdown + status dot * Added task-bar with task pills under header * Added model-status indicator * Added 'Active Model' card in settings with Pre-load button * Added Sh…d2bc8de
