fix: remove mock cards in app.py

feat: downloading only needed models

docs: removing nemotron and tildeopen model references

test: verify apkg export works end-to-end

fix: update apkg export import from csv_for_anki to apkg_export

refactor: replace _handle_export_apkg with _handle_export_csv_for_anki

feat: replace APKG stub handler with real implementation

fix: exporting files

fix: export works again

fix: export csv works

feat: wire CSV/APKG export into UI with gr.File download

refactor: extract Gradio UI construction to widgets.build_ui()

fix: more adjustments for images

fix: image prompt and size

refactor: reorder Phase 2 media generation — audio before images

fix: generate cards does not crash

feat: add image generation to Phase 2 pipeline

style: set Audio and Images toggle defaults to ON

fix: cards not displayed on card

fix: audio and images toggles when clicked on and off

fix: disable target language and updating changing language

feat: wire voice dropdown into TTS audio generation

fix: can change languages during session

fix: audio plays in app

fix: audio buttons do not go out of card bounds

feat: wire audio toggle into Phase 2 media generation and card rendering

feat: update to use local langaguage and added drop down for more languages

fix: dynamic cefr levels

fix: translation does work

feat: wire Phase 2 translation via real tiny-aya engine

feat: phase 1 is good enough

feat: adjusting for type safety

fix: using cuda for llama cpp python

refactor: update app.py error messages for MiniCPM5-1B

feat: better feedback for errors

feat: wire Nemotron TextEngine into Phase 1 — English text only, no translation

revert: restore mock data in app.py until engines are ready

feat: wire TextEngine, TTSEngine, ImageGenEngine into app.py two-phase workflow

feat: card fronts adjusted after generating cards

feat: default background

fix: gray out toggles and Generate Cards button until text generation completes

fix: restore Generate Cards button visibility on input change

fix: restore Generate Text button on input change and keep cards visible

fix: keep cards visible and Generate Text button active when changing CEFR level

fix: make _handle_*_generation wrappers generators so Gradio can consume yielded outputs

feat: split generation into two phases — Generate Text then Generate Cards

fix: center layout by merging spacer columns into single row

fix: set audio toggle default to on

feat: add media toggles and streaming progress bar

fix: update Gradio 6 API — move css/theme from Blocks() to launch()