PageLM Is a Self-Hostable, NotebookLM-Style Study Tool With Pluggable AI Providers

PageLM is an open-source education platform from CaviraOSS that turns uploaded study materials into interactive learning resources. Its README describes it as slightly inspired by NotebookLM and aimed at students, educators and researchers who want contextual chat, generated notes and audio without relying on a single hosted service. The feature set is broad for a community project. Uploaded PDF, DOCX, Markdown and TXT files can be used for contextual chat, Cornell-style SmartNotes, non-overlapping flashcards for spaced repetition, quizzes with hints, explanations and scoring, and podcast generation from notes or topics. The README also lists voice transcription for lecture recordings, a homework planner, an exam simulator called ExamLab, a debate practice mode and a study companion. That breadth suggests a tooling experiment covering many study workflows rather than a narrowly focused utility, so builders should expect uneven depth across features and test the ones they actually plan to use. Under the hood, the stack is Node.js and TypeScript on the backend with LangChain and LangGraph, and Vite, React and TailwindCSS on the frontend. WebSocket streaming is used for real-time chat, notes and podcast generation. Storage defaults to JSON with file-based persistence for generated content, with optional vector database support for embeddings and retrieval. Document parsing uses pdf-lib, mammoth and pdf-parse. Docker and Docker Compose are provided for both development and production, with the frontend on port 5173 in development and 8080 in production, and the backend on 5000. Provider choice is a central design point. Supported LLM backends include Google Gemini, OpenAI GPT, Anthropic Claude, xAI Grok, MiniMax, Ollama for local use, OpenRouter,