Peras#
A SaaS platform for producing AI-authored, learner-adaptive textbooks as EPUBs. A user describes what they want to learn, what they already know, and how they prefer to think; Peras turns that into a structured book project, generates sections through an LLM pipeline, and builds a downloadable EPUB.
What it does#
Peras is built around project-based textbook generation:
- Learner-aware setup — project creation captures audience, background knowledge, goals, tone, and style preferences
- Structured generation pipeline — planning and authoring steps produce a book outline, section drafts, and a coherent final structure
- EPUB build system — markdown sections are assembled into EPUB output with support for math, diagrams, and generated assets
- Job orchestration — generation and build jobs run through queued workflows with status tracking, logs, cancellation, and resume behavior
- Billing and credits — Stripe-backed credit packs meter generation work at the section level
- Admin and observability tooling — internal dashboards, generation diagnostics, GlitchTip errors, Prometheus metrics, and Grafana reporting
The implementation focus is less “single prompt to book” and more production system design: model selection, prompt contracts, evaluation loops, cost controls, job reliability, and product flows that make long-form generation inspectable.
Stack#
Python, FastAPI, SQLAlchemy, Alembic, React, Vite, TailwindCSS, Anthropic Claude, OpenRouter, Stripe, GlitchTip, Prometheus, Grafana.