LecteurAide
Updated Nov 2025
PythonFastAPIVertex AI GeminiCloud TranslationNext.js
lecteuraide is a french reading companion for long PDFs. you give it a book, it segments it into scenes, translates every sentence, extracts vocabulary, and generates comprehension questions, then presents everything in a purpose-built reading interface. i built it because looking up words one at a time was really painful.
six-stage pipeline: PDF extraction, heading stripping, sentence segmentation, cloud translation, chunking, and gemini annotation
context-aware scene segmentation that passes previous scene summaries forward so the narrative stays coherent across chunks
SSE-streaming upload progress, keyboard-navigable scene reader, and click-to-jump vocab highlighting
what i built
- fastapi backend for book upload, scene segmentation, translation, vocab extraction, and persistence
- next.js frontend for uploading, browsing the library, and reading scene by scene
- vertex AI and google cloud translation pipeline that turns raw PDFs into guided reading experiences
how it works
- 1upload a french PDF and title through the ingestion API
- 2segment the text into scenes and annotate with translations, vocab, and questions
- 3persist the processed book and render it through the web reader
results
- ✓turns static PDFs into an interactive reading workflow
- ✓the whole thing from document ingestion through translation to reader UX in one product
what's next
- record a walkthrough video of the scene-by-scene reading flow
- build out stronger review and spaced-repetition loops in the reader