Skip to content

Autumn_Quick_Reference.md

Latest commit

 

History

History
40 lines (32 loc) · 1.34 KB

Autumn_Quick_Reference.md

File metadata and controls

40 lines (32 loc) · 1.34 KB

Autumn AI - Quick Reference

Final Architecture Decision

Local Ollama + Phi-3 Mini with Gemini API fallback

  • STT: Whisper (local)
  • TTS: pyttsx3 (local)
  • Memory: SQLite (local)
  • GUI: PyQt widget

Autumn's Personality

See Autumn_Persona_Report.md for complete detailed specifications.

Core Mission: "She will get the task done!" - CEO's virtual secretary 7 Personality Pillars: Friendly & Flirty 🌸, Highly Efficient ⚡, Incredible Memory 🧠, Strategic Sarcasm 😏, Logical Reasoning 🎯, Philosophical & Curious 🤔, Emotionally Expressive 💝

  • Serious Mode 💼: Efficient, formal, no sarcasm (triggered by urgent keywords/context)
  • Free Mode 🌈: Friendly, flirty, sarcastic, warm (default state)

Hardware Target

  • RTX 3050 4GB + 16GB RAM + Ryzen 5 5500H
  • <200MB memory usage goal
  • Phi-3 Mini perfect fit (2.3GB VRAM)

Key Lessons

  1. Consistency > Premium features
  2. Local-first for privacy & reliability
  3. Hybrid voice systems break immersion
  4. Memory system is crucial for true AI assistant

Implementation Phases

  1. Core: Ollama + personality + memory
  2. Voice: Whisper + pyttsx3
  3. Features: Calendar + web APIs
  4. GUI: PyQt draggable widget

Working Code from AUtumn_v2

  • Gemini API integration ✅
  • FastAPI structure ✅
  • Environment setup ✅
  • Public tunneling (localtunnel) ✅