๐Ÿช„ Built-in Fuzzy Search ยท Always-cited sources ยท Trained on your support docs

Maya is the AI support assistant your school team always wanted.

Ask Maya anything about QuickSchools billing, fee tracking, discipline, student records, or reporting. She looks up your actual GitHub Knowledge Base and live ClickUp tickets โ€” and cites every answer so you never have to guess.

Download Maya Desktop v0.2

Maya is a local AI desktop app โ€” AI models run entirely on your machine, your QuickSchools data never leaves your computer. Answers are always grounded in your GitHub Knowledge Base and ClickUp tickets, with sources cited. Pick the installer for your OS below.

๐ŸŽ
macOS ยท Apple Silicon
M1, M2, M3, M4 series (arm64)
v0.2 ยท Universal binary (.dmg)
Format.dmg
Size101 MB
macOS12 Monterey+
โฌ‡ Download for Mac (M-series)
โŠž
Windows ยท x64
Windows 10 / 11 installer
v0.2 ยท Electron Setup (.exe)
Format.exe installer
Size81.5 MB
Windows10 22H2 / 11
โฌ‡ Download Windows .exe
๐Ÿง
Linux ยท Debian / Ubuntu
x86_64 .deb package
v0.2 ยท Debian package (.deb)
Format.deb (apt)
Size76.6 MB
LinuxUbuntu 22.04+, Debian 12+
โฌ‡ Download Linux .deb
Alternative for other distros: AppImage (113 MB) โ†—

Hardware Requirements

Recommended specs based on the AI model you choose inside the app. Larger models give better answers but need more RAM and disk space.

๐Ÿ’ป

Minimum

  • โœ“
    8 GB RAM โ€” runs 1.7B parameter models comfortably
  • โœ“
    5 GB free disk โ€” app + 1 compact model
  • โœ“
    Apple M1 / Intel i5 / AMD Ryzen 5 or newer
  • โœ“
    macOS 12 ยท Windows 10 ยท Ubuntu 22.04+
๐Ÿ’ช

Recommended

  • โœ“
    16 GB RAM โ€” best for 3Bโ€“4B models
  • โœ“
    15 GB free SSD โ€” multiple models + KB cache
  • โœ“
    Apple M2 Pro / Intel i7 / Ryzen 7 or better
  • โœ“
    GPU acceleration (Metal / CUDA) for 2โ€“5ร— speed
๐Ÿš€

Performance

  • โœ“
    1.7B model โ€” ~30+ tokens/sec on M2 Air
  • โœ“
    3B model โ€” ~15โ€“25 tokens/sec on 16 GB laptop
  • โœ“
    4B model โ€” ~10โ€“20 tokens/sec on 32 GB / M2 Pro
  • โœ“
    KB lookup โ€” sub-second semantic search

Available AI Models

All models run locally inside Maya. Choose inside the app based on your available RAM โ€” you can switch or download more at any time.

Gemma 3 4B
๐Ÿ“ฆ ~3.1 GB ๐Ÿง  16 GB+ RAM

Google's open model, excellent at following instructions precisely with strong English comprehension.

Qwen 2.5 3B
๐Ÿ“ฆ ~1.9 GB ๐Ÿง  12 GB+ RAM

Fast mid-size model, solid general-purpose answers without the 4B-size footprint.

Llama 3.2 3B
๐Ÿ“ฆ ~1.9 GB ๐Ÿง  12 GB+ RAM

Meta's latest small Llama, tuned for chat and structured output with strong safety alignment.

SmolLM2 1.7B
๐Ÿ“ฆ ~1.2 GB ๐Ÿง  8 GB+ RAM

Ultra lightweight, runs smoothly on 8 GB laptops. Great first model.