AXON

AXON: AI-Powered Desktop Assistant

A voice- and text-controlled desktop assistant with dual-LLM routing.

Category
AI Systems · Agents
  • Python
  • Gemini
  • LLaMA
  • Speech recognition
  • Speech synthesis

Abstract preview - not a screenshot

01 / Problem

A voice- and text-controlled assistant for system commands, app launching, web search and conversational Q&A.

02 / Approach

  1. 01Dual-LLM routing: Gemini for command interpretation and screen-image analysis, LLaMA for open-ended conversation.
  2. 02Fuzzy command matching, robust to speech-recognition errors.
  3. 03Speech synthesis and interaction logging.

03 / Architecture

AXON system architectureVoice or text, then Fuzzy command matching, then LLM router, then Dual LLM (Gemini, LLaMA), then Action or answer, then Speech synthesis.01Voice or textMIC → SPEECH RECOGNITION · KEYBOARD02Fuzzy command matchingTOLERATES RECOGNITION ERRORS03LLM routerCOMMAND OR CONVERSATION?04GeminiCOMMANDS · SCREEN VISIONLLaMAOPEN CONVERSATION05Action or answerAPPS · SYSTEM · WEB SEARCH · Q&A06Speech synthesisSPOKEN REPLY · INTERACTION LOG
Fig. 01 - System architectureSimplified

04 / Key details

  • System commands, app launching, web search and conversational Q&A.
  • Gemini handles command interpretation and screen-image analysis; LLaMA handles open-ended conversation.
  • Fuzzy command matching robust to speech-recognition errors.
  • Speech synthesis and interaction logging.

05 / Results & signals

2 LLMs
Dual routing
Voice+Text
Input modes
Fuzzy
Command matching
Vision
Screen analysis

Next entry →

AirVision

Eleven regressors, one benchmark, a live AQI dashboard.