fix: add 90s per-chunk TTS timeout to prevent CPU audio hangs

fix: improve audio concat, dynamic context limits, thread-safe jobstore, TTS base64 decoding, video stream-loop sync

fix: better error handling for network failures during upload

fix: persist job state across page refresh with localStorage

fix: rewrite chapter detector with AI agent classification + RAG fallback

fix: chunk text for TTS (500 char segments + concat), strip #hex from Manim scripts

fix: stronger sanitization for Manim #CCC>L color errors (force redeploy)

fix: make stop/pause buttons work - use url-encoded POST, add pre-chapter control check, interruptible sleep

fix: improve audio generation (keep code context, raise min size) and fix Manim color parsing

fix: run generation in thread executor so SSE progress streaming works

fix: remove duplicate 'saved' declaration breaking JS execution, wrap password fields in form

fix: answer key timeout, frontend error handling, SonarLint issues

fix(engine): faster model switching on empty responses and BaseException handling

fix(audio): strip markdown before TTS synthesis, validate output size

fix: resolve HuggingFace deployment issues (video API, sanitization, prompt engineering, section detection)

feat: PDF bookmark injection via pypdf; print-optimized HTML for xhtml2pdf (rgba->hex, no fixed sidebar, inline TOC page, @page CSS)

fix: add xhtml2pdf PDF fallback (emoji/Unicode support, no GTK); sanitize chapter_text in generate_chapter_video to fix Manim color parse errors

fix: strip HTML tags/entities from Manim scene text to prevent color parse errors

fix: zip audio/video dirs on download (500 FileResponse on dir); fix Manim IndentationError (double-indent in scene_body)

chore: ruff clean — fix E402/F841/E741/E501; suppress intentional S/B patterns in pyproject.toml

fix: SyntaxError in video_generator.py line 372 (nested f-string with backslash); suppress LoRACompatibleLinear FutureWarning

fix: suppress pkg_resources deprecation warning from perth/chatterbox in web and CLI entrypoints

fix: treat JSON decode errors as retryable; switch model after exhausting empty-response retries

feat: Manim animated video (code write-on, flowcharts, concept cards); slideshow fallback; fix video None model; cache TTS engine; strip emoji from PDF; Manim+ffmpeg system deps in Dockerfile

fix: video None model + slideshow fallback (Pillow+ffmpeg); cache TTS engine; strip emoji from PDF filenames; pin setuptools<81; pass audio to video

fix: LLM timeout 90s; empty response retries with backoff; total validation failure switches model; skip failed chapters in answer key

fix: model switching probes before switching; 401 raises immediately; truncation limit; network errors don't fail models; video auto-selects free model; polling_url used; checkbox bool fix; no global env mutation; flatten audio/video paths

fix: use from_pretrained() for Chatterbox engines; fix infinite TTS retry loop

fix: remove duplicate audio/video checkboxes and form.append calls in index.html

feat: full Docker extras (pdf+tts+video); frontend audio/video controls; updated docs and meta tags

chore: bump version to 0.9.2; update frontend Usage Guide with audio/video/multimedia features

fix: 3 issues — test patch target, test filesystem isolation, PLAN validate done; videopython import guard

fix: syntax error - indent contextlib import; shorten test comment

fix: 6 issues — tts-kitten optional, video model free warning, lifespan context, engine test isolation, PLAN updates, chromadb lock refresh

chore: bump version to 0.9.1

style: final lint cleanups - line breaks, import ordering, test client context

feat: comprehensive improvements (Groups 1-6)

Fix download filenames, add answer key retry

Fix surrogate emoji encoding in HTML

Add BYOK support: users provide their own API keys via browser

Add Apache 2.0 license, HF Spaces config, configurable port

feat(export): add Pygments syntax highlighting and copy button support

feat(deps): replace WeasyPrint with Playwright for PDF generation

fix: review step crash - wrong attribute name + add try/except safety

feat: playwright HTML-to-PDF with emoji support, fpdf2 fallback

fix: improve subject detection, skip junk lines (page numbers, copyright, dates)

release: v0.9.0

feat: auto model switching + model health testing

feat(job): add pause/stop control for generation jobs and enhance error handling

v0.8.0: theme system, TOC navigation, docs update- Bump version to 0.8.0 (pyproject.toml, web.py)- CHANGELOG: add 0.8.0 entry- README: add theme feature, --theme CLI example- PLAN.md: update status to v0.8.0, fix known limitations, update file map- AGENTS.md: add --theme flag, themes.py, EPUB loader- TASK_TRACKER.md: add v0.7.0 and v0.8.0 completed sections