DeepLM

DeepLM is a self-hosted Next.js language-learning app for grammar correction, translation styles, and English/German tense practice with Persian glosses. All generation is handled server-side through OpenAI.

Architecture

flowchart LR
  browser[Next.js PWA] --> routes[Next.js API routes]
  routes --> openai[OpenAI Chat API]

Loading

The repository is a single Next.js App Router application. The browser calls same-origin /api/* route handlers; OPENAI_API_KEY is never exposed to the client.

Requirements

• Node.js 20+
• An OpenAI API key

Quick start

copy .env.example .env.local
npm install
npm run dev

Set OPENAI_API_KEY in .env.local, then visit http://localhost:3000.

Configuration

Variable	Purpose
OPENAI_API_KEY	Server-side OpenAI key, required for generation.
OPENAI_MODEL	Chat model to use; defaults to gpt-4o-mini.

API

Method	Path	Purpose
GET	/api/health	Application and OpenAI configuration status.
GET	/api/languages	Language, locale, and tense metadata.
GET	/api/changelog	Release list for the Versions tab.
POST	/api/grammar	Grammar correction and styled translation.
POST	/api/tenses	English or German tense chart.
POST	/api/tenses/explain	Tense explanation and examples.

Deployment

Deploy the repository root to Vercel or any Node-compatible host. Configure OPENAI_API_KEY and optionally OPENAI_MODEL as server environment variables. Do not prefix either variable with NEXT_PUBLIC_.

Versioning

Canonical semver is VERSION. Log changes in CHANGELOG.md.

License

MIT. Copyright Master2iT.