REDDIT · REDDIT// 10h agoOPENSOURCE RELEASE

SHANA brings local voice AI to AMD Linux

SHANA is an open-source, fully local AI assistant for AMD GPU systems on Linux. It combines Ollama for local LLM inference, Whisper for speech-to-text, Piper for text-to-speech, and a Flask/Socket.IO web UI with persistent session memory. The project is positioned specifically for AMD hardware, with the README calling out Linux Mint and gfx1151/Strix Halo testing, and it defaults to `qwen2.5:7b-instruct` for the model backend.

// ANALYSIS

Strong niche utility: this is not a general AI demo, it is a clear “AMD-first, offline-first” assistant stack for people who want local voice chat without CUDA or cloud services.

–The core value prop is straightforward: everything runs on-device, including voice input, voice output, memory, and model inference.
–The project stacks known local components rather than inventing a new model, which makes it practical and reproducible.
–The Linux-only and AMD-specific focus narrows the audience, but that focus is also the product’s main differentiator.
–The repo looks very early-stage from a packaging maturity standpoint, so this reads more like a useful open-source prototype than a polished end-user app.

// TAGS

local-aiamdlinuxollamawhisperpiper-ttsvoice-assistantopen-sourceofflineprivacy

DISCOVERED

10h ago

2026-04-17

PUBLISHED

10h ago

2026-04-17

RELEVANCE

8/ 10

AUTHOR

Practical-Cupcake259