bug fix

Increase read timeout and max retries in API call

Reduce max_chun_len from 128k to 32k

chore: update Space title to DataFlow-VQA

feat: embed extracted images as base64 data URIs in preview cards

fix: post-process img tags for badges; add MathJax for LaTeX rendering in preview

feat: render QA preview as markdown with image placeholders

feat: bundle extracted images + JSONL into vqa_output.zip for download

ux: remove PDF preview entirely

fix: replace data-URI iframe with /gradio_api/file= URL for cross-browser PDF preview

ux: clear default api_url, leave as placeholder only

ux: default en, move examples section, clean up preview label and API table

fix: Markdown does not accept scale kwarg, wrap in Column

feat: bilingual toggle, PDF preview, stop btn, JSONL card preview

feat: add example PDFs via LFS and gr.Examples in UI

fix: upgrade to gradio 5.25 to fix bool-schema TypeError in gradio_client

fix: remove generator/queue, add api_name, add startup logs to diagnose No API Found

fix: rewrite as generator for streaming progress, fix connection timeout

fix: add demo.queue() and remove gr.Progress to fix No API found error in Gradio 5

improve: clearer error messages for invalid API keys

fix: use Python 3.11 to avoid whylogs-sketching build failure on 3.13

refactor: demo now wraps Stage1 VQA extraction pipeline

fix: upgrade to gradio 5.x to match open-dataflow requirement

fix: correct package name open-dataflow

fix: use PEP 440 direct URL syntax for DataFlow git dependency

initial hf spaces demo