ReadBuddy Project Context

Loads architectural, design, and schema context for the ReadBuddy reading comprehension project.

Updated Jul 26, 2026
One-click install
npx skills add https://github.com/ConcernedDeyb/readbuddy --skill readbuddy-project-context-concerneddeyb
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ReadBuddy Project Context
Source: https://github.com/ConcernedDeyb/readbuddy/tree/main/.agents/skills/readbuddy-context
Command: npx skills add https://github.com/ConcernedDeyb/readbuddy --skill readbuddy-project-context-concerneddeyb

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Working on the ReadBuddy capstone project requires constant awareness of its tech stack, binding rules, database schema, and manuscript requirements; this Skill injects that full project context automatically so code changes and documentation stay consistent with the source-of-truth documents. ## Core Features & Use Cases - Tech Stack & VRAM Constraints: Enforces Next.js/FastAPI architecture, mms-1b-all ASR with language adapters, Gemma 3 4B via Ollama, and the 8GB GPU phase-separated model loading budget. - Binding Rules R-1 through R-21: Applies constraints on AI usage, data privacy for minors, authentication (httpOnly JWT cookies, bcrypt), language framing, and UI design standards. - Schema & Manuscript Guidance: Provides the 12-table PostgreSQL schema, Phil-IRI rubric, user flow maps, and the SMCC capstone manuscript structural checklist. - Use Case: When asked to add a new student-facing feature, the Skill ensures the implementation respects rules like no student self-registration (R-16), encouraging feedback tone, and the existing reading_sessions schema. ## Quick Start Ask the AI to implement or review any ReadBuddy feature, such as adding a new endpoint for teacher reading-test assignments, and it will apply the project's stack, rules, and schema automatically.

Frequently Asked Questions about ReadBuddy Project Context

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
What tech stack does the ReadBuddy project use?

ReadBuddy uses a Next.js frontend with WebSocket audio streaming and a FastAPI async Python backend. AI inference runs locally with the facebook/mms-1b-all CTC model for speech recognition, Gemma 3 4B via Ollama for comprehension questions, EasyOCR for photo text extraction, and PostgreSQL for storage.

How does ReadBuddy handle pronunciation scoring and reading assessment?

Pronunciation scoring uses only the CTC-based mms-1b-all ASR model with per-language adapters, never an LLM. Results are mapped to the Phil-IRI rubric (Independent, Instructional, Frustration) based on word recognition and comprehension percentages, taking the lower of the two tiers.

What are the VRAM constraints for running ReadBuddy models?

The project targets an 8GB GPU with phase-separated loading: the ASR model uses roughly 1.2-1.9GB during reading, Gemma 3 4B uses about 2.5GB during comprehension tests, and models are released between phases with gc.collect() and torch.cuda.empty_cache().

Can students register their own accounts in ReadBuddy?

No, students cannot self-register; only teachers create student accounts per rule R-16. Teachers themselves must register, verify their email, and receive admin approval before accessing their dashboard.

What are the known limitations of the ReadBuddy system?

The system supports only one guided reading session at a time, runs on a single-device deployment during the pilot, and lacks production auth hardening such as rate-limiting and CSRF protection. Both production models are also unverified on real target-population student recordings.