fix(ui): point to scvcoder-kpaa.hf.space + ENABLE_PERSISTENT_CONFIG=false

Set SameSite=none+Secure for cross-site iframe cookie (model dropdown fix)

Sync README.md with README_ui.md (sdk=docker, app_port=8080)

Point Open WebUI to renamed backend URL (scvcoder-kpaa.hf.space)

ui: DEFAULT_MODELS → gemma-4-e2b-unsloth-q4

ui: DEFAULT_MODELS → gemma-4-e2b-unsloth (matches new backend default)

docs: rename Space card title/description

ui: update DEFAULT_MODELS to new preset id format

Robust route detection: pushState/replaceState/popstate hooks + console.log

Inject route-change postMessage script into Open WebUI HTML

DEFAULT_MODELS: 개인정보 미니 상담 AI

Fix UID 1000 permissions (chown /app); require WEBUI_SECRET + OAUTH key as secrets

Convert to Open WebUI Docker SDK; delete prior Gradio app files

Diagnostic: count streamer tokens to isolate generate vs streamer issue

Use GenerationConfig (transformers 5.x); add diagnostic prints + streamer.end on exception

Extract input_ids tensor from BatchEncoding (transformers 5.x compat)

Allow huggingface-hub 1.x: required by transformers 5.x; Gradio 5.20+ no longer needs HfFolder

Bump transformers >=5.0 for Gemma 4 tokenizer support

Add module-level @spaces.GPU canary for ZeroGPU startup detector

Disable SSR to avoid Node subprocess shutdown; combine with monkey-patch + show_api=False

Monkey-patch gradio_client to handle bool JSON Schema (additionalProperties:true)

Disable SSR to avoid Gradio /api_info schema bug

Bump Gradio sdk_version 5.0.0 -> 5.20.0 (fix bool schema bug in api_info)

Pin huggingface-hub<1.0 to fix Gradio 5.0.0 HfFolder import

Fix build: drop -e . and add src/ to sys.path

Replace template with KPAA app: src/kpaa, runtime data, requirements

initial commit

Duplicate from gradio-templates/chatbot