Aether Dev

macOS Digital Life System — Long Memory, Real Emotion, Agent-Native

Aether Dev is a macOS developer-side foundation for a long-memory digital life system — not a passive chatbot, but a persistent digital entity with structured memory architecture, emotion modeling, and autonomous reflection cycles.

Architecture

┌────────────────────────────────────────────┐
│  Flutter macOS App (Dart)                  │
│  ChatScreen · DeveloperDashboard · Memory │
├────────────────────────────────────────────┤
│  FastAPI Backend (Python 3.13)            │
│  ┌──────────┬──────────┬────────────────┐ │
│  │ Routers  │ Services │ Plugins        │ │
│  │ health   │ chat     │ example_memory │ │
│  │ convos   │ memory   │ (extensible)   │ │
│  │ chat     │ emotion  │                │ │
│  │ memory   │ audit    │                │ │
│  │ emotion  │ event    │                │ │
│  │ dev      │ embed    │                │ │
│  │ plugins  │ vector   │                │ │
│  │ WebSocket│ model    │                │ │
│  └──────────┴──────────┴────────────────┘ │
├────────────────────────────────────────────┤
│  SQLite + Vector Store + Plugin Runtime   │
│  macOS Application Support sandbox        │
└────────────────────────────────────────────┘

Key Features

L1 / L2 / L3 Memory System

• L1 — Conversation context window with configurable message limit
• L2 — Auto-extracted long-term memories surfaced during conversation
• L3 — Persistent vector-store entries with embedding support (mock, OpenAI, Chroma, LanceDB)

PAD Emotion Engine

• Pleasure-Arousal-Dominance emotional state model
• Energy level tracking independent of emotion
• State update API with event-bus notification

Autonomous REM Cycle

• Manual trigger endpoint (POST /developer/rem/run)
• Memory consolidation via service pipeline
• Runtime event logging per REM pass

Plugin System

• Manifest-based plugin discovery from plugins/ directory
• Enable / disable at runtime via API
• Example memory plugin shipped

Developer Dashboard

• Runtime status viewer (pid, uptime, memory count, emotion state)
• Audit log browser
• Structured log viewer
• Flutter macOS native window

Quick Start

# Backend
cd backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload --host 127.0.0.1 --port 8000

# Health check
curl http://127.0.0.1:8000/health

# Send a message
curl -X POST http://127.0.0.1:8000/conversations \
  -H "Content-Type: application/json" \
  -d '{"title":"Hello"}'

curl -X POST http://127.0.0.1:8000/chat/message \
  -H "Content-Type: application/json" \
  -d '{"content":"What can you remember about me?"}'

# Flutter frontend
cd frontend/flutter_app
flutter run -d macos

API Endpoints (21 routes)

Category	Method	Path
Health	GET	/health
Conversations	POST /GET	/conversations
Conversations	GET	/conversations/{id}/messages
Chat	POST	/chat/message
Chat	WS	/ws/chat/{id}
Memory	POST	/memory/write
Memory	GET	/memory/query
Memory	GET	/memory/recent
Memory	DELETE	/memory/{id}
Memory	POST	/memory/{id}/level/{level}
Emotion	GET	/emotion/status
Emotion	POST	/emotion/update
Developer	GET	/developer/runtime
Developer	GET	/developer/logs
Developer	GET	/developer/audit
Developer	POST	/developer/rem/run
Plugins	GET	/plugins
Plugins	POST	/plugins/{name}/enable
Plugins	POST	/plugins/{name}/disable

Test Results

backend:  3/3 passed  (FastAPI endpoint tests)
flutter:  1/1 passed  (widget tests)
analyze:  No issues found

Project Structure

Aether_Dev/
├── backend/           FastAPI backend (Python)
│   ├── app/
│   │   ├── routers/    21 API route handlers
│   │   ├── services/   8 service modules
│   │   └── tests/      MVP test suite
│   └── scripts/        Build & start scripts
├── frontend/          Flutter macOS app (Dart)
│   └── flutter_app/
│       ├── lib/        Screens, models, services
│       └── macos/      Native macOS runner
├── plugins/           Plugin manifests & implementations
├── docs/              Architecture, API, roadmap docs
└── scripts/           CI scripts (test, build, package)

Tech Stack

Layer	Technology
Backend	FastAPI 0.115, Uvicorn, Python 3.13
Database	SQLite (via aiosqlite)
Vector Store	Local JSON + Chroma / LanceDB adapters
Frontend	Flutter 3.x, Dart, macOS native
Packaging	PyInstaller + macOS .dmg
Testing	Pytest, Flutter test

Roadmap

More details in docs/evolution_roadmap.md. The MVP is the stable base — real model gateway, WebSocket streaming UI, Chroma/LanceDB backends, and release packaging are the next increment.

License

MIT — see LICENSE.