fix: robust PaddleX 3.x OCRResult text extraction and multi-route alias

fix: support multipart form data and improve paddleX text extraction

paddlepaddle ocr v6

decorator

Update app.py to PaddleX 3.x text-only format

Trigger rebuild to pull latest main.py from GitHub

Fix: convert PIL.Image to numpy array before passing to PaddleOCR

Fix: modify demo.app.routes via clear and extend instead of reassignment

Fix: add hasattr(path) check to routes to prevent AttributeError

Fix: elevate /ocr route priority to avoid 405 routing conflicts

Fix: revert custom FastAPI launch and register route on demo.app

Feat: mount gradio on custom_app to expose /ocr endpoint

Feat: register /ocr JSON endpoint for FastAPI

Fix: resolve syntax error and restore gradio blocks

Update ocr endpoint to accept base64 json

Update ocr endpoint to accept base64 json

Fix: Revert demo override and use native Gradio API to expose /api/ocr

Fix: Mount Gradio app to FastAPI to preserve /ocr post endpoint

Update ocr endpoint to accept base64 json

api change due to the webapp alignment

Update route endpoint to /api/ocr to pass proxy whitelist

Directly register post route on demo.app to bypass mounting reset

Change FastAPI mount path to /ocr_api to avoid Gradio conflict

Fix port conflict by mounting FastAPI under /api and using demo.launch

Fix FastAPI mounting to resolve 405 error

fix: enforce CPU runtime with dummy GPU trigger to support ZeroGPU space

fix: bump paddlepaddle to 2.6.2 to resolve PyPI version mismatch

fix: revert to stable CPU dependency in app.py

fix: restrict numpy version to <2.0.0 to solve ABI incompatibility

fix: lock paddlepaddle-gpu and paddleocr to stable versions

fix: downgrade Python to 3.10 and install paddlepaddle-gpu

opt: set OMP_NUM_THREADS=1 at entrypoint

fix: disable PIR compiler API at entrypoint

fix: disable oneDNN at entrypoint first line

feat: enforce file size validation in Gradio via filepath

fix: call demo.launch() to prevent process from exiting

fix: integrate fastapi routes into gradio to avoid port collision

fix: add dummy gpu trigger for ZeroGPU startup check

fix: add dummy gpu trigger for ZeroGPU startup check

python 3.12

Initialize Hugging Face Space deployment shell