Kalaam.

The kalaam npm package has zero runtime dependencies. The interpreter is pure JavaScript — no parser framework, no tokenizer library, no runtime engine. This keeps the bundle tiny, works completely offline, and means any environment that runs JS can run Kalaam.

The interpreter is language-agnostic. Phase 1 (Cleaning) substitutes language-specific keywords into normalized tokens before any parsing begins. Adding Hindi, Marathi, Bengali, Telugu, or Odia required no parser changes — just one keyword map object per language.

Every operation the interpreter performs appends to ExecutionStack[]. The PWA UI replays this stack line-by-line in the student's language, showing what each line does and why. Students learn not just to write code but how the computer evaluates it — without a teacher or internet.

The Vue 2 + Quasar PWA is service-worker cached on first load. After that, no internet is needed — the app, interpreter, and all language assets run entirely from the device. Designed for tier-3 city students with intermittent connectivity and no laptop.

The code editor uses CodeMirror with a custom syntax mode written specifically for Kalaam. It highlights Hindi/Marathi/etc. keywords, strings, and comments in real-time. The mode understands the normalized token structure from Phase 1, giving accurate syntax coloring for all 5 languages.

The package exposes one public function: Compile(sourcecode, languageKeywords) which returns { output, ExecutionStack, isError, TimeTaken }. The caller decides which language keyword map to pass. The interpreter doesn't know or care which language is active — it only sees normalized tokens after Phase 1.