Your private AI super-app — companions, agents, creative tools & more, all self-hosted
Aethvion Suite is a powerful, self-hosted AI platform combining the best cloud models (Gemini, GPT-4o, Claude, Grok) with fast local GGUF models — all in one unified, privacy-first dashboard.
Run intelligent agents on your own code and files, chat with persistent AI companions who actually remember you, generate images and audio, research any topic with a multi-expert board, play AI-powered games, and automate recurring tasks. Everything runs on your machine, under your control.
Whether you're a developer, creator, researcher, or just want a powerful personal AI setup, Aethvion Suite gives you a unified workspace that adapts to how you want to use it.
Current version: v15
- AI Companions — Persistent characters with memory, moods, tools, and workspace access; they remember who you are across every session
- Hybrid Intelligence — Seamlessly mix cloud APIs (Gemini, GPT-4o, Claude, Grok) and local GGUF models with smart auto-routing and failover
- Powerful Agents — Multi-step ReAct agents with file access, shell execution, and real-time streaming output
- Agent Corp — Create multi-worker AI organizations with specialized roles; assign complex projects and watch them collaborate
- Research Board — Generate structured expert perspectives on any topic; debate mode lets experts argue across multiple rounds
- Creative Studio — Image generation (DALL-E 3, Imagen 3, Stable Diffusion), Audio (TTS/STT with voice cloning), and 3D asset generation
- Bridges — Live connections between companions and your environment: Spotify, weather, system stats, screen capture, webcam, and more
- Privacy First — Intelligence Firewall scans and blocks sensitive data before it leaves your machine; all data stored locally
- One-Click Windows Launcher —
Start_Aethvion_Suite.bathandles venv creation, dependency installation, and launch
 |
 |
| Agent Workspace · Multi-step ReAct agent runner with real-time SSE event streaming | Agent Corp · Manage and collaborate with multiple persistent specialized agents |
 |
 |
| Companions · Persistent AI characters with memory, moods, tools, and workspace access | Aethvion Photo · AI image generation with DALL-E 3, Imagen 3, and local Stable Diffusion |
 |
 |
| Aethvion Audio · Local TTS (Kokoro, XTTS-v2) and STT (Whisper) with voice cloning | Aethvion Code IDE · Monaco-based IDE with AI copilot and code execution |
 |
 |
| Local Models · Browse, download, and run GGUF models and Ollama models locally | Usage & Cost Tracking · Token usage, cost estimates, and per-provider breakdowns |
Aethvion Suite is a self-hosted AI platform built around three core ideas: local (your data, your models, your hardware), personal (companions that know you, a dashboard that adapts to you), and unified (every AI workflow in one place, and those workflows talk to each other).
| Component | Role | Status |
|---|---|---|
| AetherCore | Central AI gateway — routes all requests, manages failover, Intelligence Firewall | Stable |
| Companion Engine | Persistent AI companions with memory, moods, tools, and workspace access | Stable |
| Agent Runner | ReAct-style multi-step task execution with file I/O and shell commands | Stable |
| Bridges | Registry-driven integrations: Spotify, weather, system stats, screen, webcam | Stable |
| Memory Tier | Episodic memory (ChromaDB) + knowledge graph + persistent memory + identity | Stable |
| Schedule Manager | Cron-based recurring AI task automation with notifications | Stable |
| Provider Manager | Multi-provider routing with automatic failover across all AI calls | Stable |
Cloud Providers: Google AI (Gemini 2.0 Flash, 1.5 Pro, Imagen 3) · OpenAI (GPT-4o, DALL-E 3) · xAI (Grok) · Anthropic (Claude) Local Models: GGUF via llama-cpp-python · Ollama integration · Kokoro TTS · XTTS-v2 (voice cloning) · Whisper STT Intelligence Firewall: PII/credential scanning before any external API call — sensitive data never leaves your machine.
Four distinct AI companions, each with a unique personality, capability set, and persistent memory:
| Companion | Personality | Capabilities |
|---|---|---|
| Misaka Cipher | Adaptive, curious, playful | Tools, workspace access, web search, bridges, memory |
| Axiom | Analytical, precise, stoic | Tools, workspace access, bridges, memory |
| Lyra | Empathetic, creative, warm | Bridges, memory |
| Nova | Minimalist, calm, observant | Memory |
Every companion has persistent base_info.json (evolving identity) and memory.json (dynamic observations about you). Companions learn your preferences across sessions.
Live connections between companions and your environment, called mid-conversation:
| Bridge | What it does |
|---|---|
spotify_bridge |
Control Spotify — play, pause, skip, get current track |
weather_bridge |
Real-time weather for any location |
media_sentinel |
What's currently playing on Windows (any app) |
system_bridge |
Live CPU, RAM, and disk usage |
screen_capture |
Take a screenshot and share it with the companion |
webcam_capture |
Capture a webcam image |
# Clone and install
git clone https://github.com/Aethvion/Aethvion-Suite.git
cd Aethvion-Suite
pip install -e ".[memory]"
# Configure providers (at least one required)
copy .env.example .env
# Edit .env — add any of: GOOGLE_AI_API_KEY / OPENAI_API_KEY / GROK_API_KEY / ANTHROPIC_API_KEY
# Leave blank to use only local modelsOne-click (Windows): Double-click Start_Aethvion_Suite.bat — it creates the venv, installs dependencies, and opens the dashboard automatically.
| Application | Launcher | Default URL |
|---|---|---|
| Suite Dashboard | Start_Aethvion_Suite.bat |
http://localhost:8080 |
| Code IDE | apps/code/Start_Code.bat |
http://localhost:8083 |
| VTuber Engine | apps/vtuber/Start_VTuber.bat |
http://localhost:8081 |
| Audio Editor | apps/audio/Start_Audio.bat |
http://localhost:8081* |
| Photo Editor | apps/photo/Start_Photo.bat |
http://localhost:8081* |
| Finance Hub | apps/finance/Start_Finance.bat |
http://localhost:8081* |
| Drive Info | apps/driveinfo/Start_DriveInfo.bat |
http://localhost:8084 |
| Tracking Bridge | apps/tracking/Start_Tracking.bat |
http://localhost:8081* |
| Hardware Info | apps/hardwareinfo/Start_HardwareInfo.bat |
http://localhost:8081* |
* Apps sharing port 8081 auto-negotiate the next available port if multiple are running.
Manual:
python -m core.main # web dashboard at http://localhost:8080
python -m core.main --cli # interactive CLI
python -m core.main --test # run verification tests- Four companions (Misaka Cipher, Axiom, Lyra, Nova) — each with distinct personality, memory, and capability profile
- Per-session and long-term memory — companions remember preferences, past conversations, and observations about you
- Bridges integration — companions can check weather, control Spotify, read system stats, and more, mid-conversation
- Workspace access — companions can read, write, and search files in permitted directories
- Floating overlay companion (Misaka) — accessible system-wide via hotkey (Ctrl+Shift+Space)
- Research Board — simulate a board of experts; each gives a structured perspective; optional live debate mode where experts argue across multiple rounds
- AI Explained — generate richly formatted HTML deep-dives on any topic, with persistent threads
- Model Arena — blind benchmark: run the same prompt across multiple models simultaneously and rate results
- Advanced Lab — structured multi-persona conversations with custom participant profiles
- Agent Workspaces — create named workspaces; agents run multi-step ReAct loops (read/write/list/run_command); each step streams live to the dashboard
- Agent Corp — build multi-worker AI organizations; assign complex projects across specialized roles
- Task Scheduler — cron-based recurring AI tasks; completion triggers notifications with direct links to results
- Task Queue — submit, cancel, and monitor background AI tasks with persistent state
- Image generation — DALL-E 3, Imagen 3, Stable Diffusion WebUI, ComfyUI
- Audio Studio — local TTS (Kokoro, XTTS-v2 with voice cloning), STT (Whisper), and a full multi-track timeline editor
- 3D Generation — text-to-3D and image-to-3D (Trellis 2, TripoSR)
- Local Model Hub — browse, download, and manage GGUF models and audio models locally
- Checkers, Blackjack, Sudoku, Word Search, Logic Quest — all with AI opponents or generation
- Are You Smarter Than AI? — gameshow-format trivia: AI Game Master generates questions, you and an AI opponent compete, the GM judges in real time
- External API — OpenAI-compatible
/v1/chat/completionsendpoint; use Aethvion as a backend for other apps - Discord Bot — connect your Discord server; full bi-directional messaging with companions
- Usage & Cost Tracking — token counts, cost estimates, per-provider and per-day breakdowns
- Persistent Memory — long-term topic-based knowledge hub; manually curate facts that every companion and agent can retrieve
- Notification System — real-time alerts with persistent history, category filtering, and deep-link navigation
- Port Manager, Logs Viewer, System Status — full operational visibility
Aethvion-Suite/
├── Start_Aethvion_Suite.bat # One-click install + launch (Windows)
├── pyproject.toml # All dependencies + project metadata
│
├── core/ # Shared AI core — used by all apps and the dashboard
│ ├── main.py # Entry point (web / CLI / test modes)
│ ├── aether_core.py # Central AI gateway (AetherCore) — routing, failover, firewall
│ ├── bridges/ # Companion integration layer
│ │ ├── bridge_manager.py # Registry-driven module loader and dispatcher
│ │ ├── registry.json # Active bridge module registry
│ │ ├── spotify_bridge.py # Spotify playback control
│ │ ├── weather_bridge.py # Real-time weather data
│ │ ├── system_bridge.py # CPU / RAM / disk telemetry
│ │ ├── media_sentinel.py # Windows SMTC media info
│ │ ├── screen_capture.py # Screenshot capture
│ │ └── webcam_capture.py # Webcam image capture
│ ├── companions/ # Companion system
│ │ ├── companion_engine.py # Main chat generator — history, tools, memory, streaming
│ │ ├── companion_routes.py # FastAPI routes for all companion endpoints
│ │ ├── registry.py # Companion config loader (CompanionConfig)
│ │ ├── configs/ # Per-companion JSON configs
│ │ │ ├── misaka_cipher.json # Misaka Cipher — adaptive, tool-enabled
│ │ │ ├── axiom.json # Axiom — analytical, precise
│ │ │ ├── lyra.json # Lyra — empathetic, creative
│ │ │ └── simple_companion.json # Nova — minimalist
│ │ └── engine/ # Shared companion subsystems
│ │ ├── memory.py # base_info.json + memory.json — init, load, XML update, synthesis
│ │ ├── history.py # Per-companion conversation history
│ │ ├── streaming.py # build_bridges_capabilities() — prompt injection helper
│ │ └── tools.py # Tool block parsing, execution, workspace validation
│ ├── config/ # Configuration files (version-controlled)
│ │ ├── providers.yaml # Provider priority and failover config
│ │ ├── security.yaml # Intelligence Firewall rules
│ │ ├── settings_manager.py # Settings singleton
│ │ ├── suggested_apimodels.json
│ │ └── suggested_localmodels.json
│ ├── factory/ # Agent spawning engine (legacy factory pattern)
│ ├── memory/ # Memory tier
│ │ ├── episodic_memory.py # Vector-based interaction storage (ChromaDB)
│ │ ├── identity_manager.py # System identity and Misaka persona management
│ │ ├── knowledge_graph.py # Entity-relationship graph (NetworkX)
│ │ ├── persistent_memory.py # Long-term topic-based knowledge hub
│ │ ├── social_registry.py # Cross-platform user identity mapping
│ │ ├── summarization.py # Memory consolidation and core insights
│ │ └── file_vector_store.py # Semantic file indexing (FastEmbed + ChromaDB)
│ ├── orchestrator/ # Task orchestration
│ │ ├── agent_runner.py # ReAct-style multi-step agent execution loop
│ │ ├── task_queue.py # Task submission, cancellation, state management
│ │ ├── persona_manager.py # System prompt building and persona routing
│ │ └── schedule_manager.py # Cron-based recurring task scheduler
│ ├── providers/ # AI provider abstraction layer
│ │ ├── provider_manager.py # Multi-provider routing with failover
│ │ ├── google_provider.py # Google AI (Gemini, Imagen)
│ │ ├── openai_provider.py # OpenAI (GPT-4o, DALL-E 3)
│ │ ├── grok_provider.py # xAI (Grok)
│ │ └── anthropic_provider.py # Anthropic (Claude)
│ ├── security/ # Intelligence Firewall
│ │ ├── firewall.py # Pre-flight PII/credential scanning
│ │ └── scanner.py # Regex-based content scanning
│ ├── workers/ # Background workers
│ │ ├── discord_worker.py # Persistent Discord bot service
│ │ └── package_installer.py # Async pip package installation
│ ├── workspace/ # Workspace utilities
│ │ ├── workspace_utils.py # load_workspaces(), validate_path(), build_workspace_block()
│ │ └── preferences_manager.py # User preference persistence
│ ├── utils/ # Shared utilities
│ │ └── paths.py # Canonical data path constants (single source of truth)
│ └── interfaces/
│ ├── dashboard/ # Web dashboard (FastAPI + static files)
│ │ ├── server.py # FastAPI app — 32 routers registered
│ │ └── *_routes.py # Feature route modules (agents, arena, companions, games, …)
│ └── cli_modules/ # CLI module implementations
│
├── apps/ # Standalone apps — each has its own server + launcher
│ ├── audio/ # Audio processing (TTS/STT)
│ │ ├── models/ # Local audio model adapters (Kokoro, XTTS-v2, Whisper)
│ │ └── tts_manager.py # TTS/STT lifecycle manager
│ ├── code/ # Monaco-based Code IDE (port 8083)
│ ├── driveinfo/ # System storage and drive info
│ ├── finance/ # Finance tracking + AI market analysis
│ ├── hardwareinfo/ # System hardware information
│ ├── photo/ # AI-powered photo editing
│ ├── tracking/ # Motion tracking WebSocket server (port 8082)
│ └── vtuber/ # VTuber character animation engine (port 8081)
│
├── data/ # Runtime data — never committed
│ ├── companions/ # Per-companion persistent data
│ │ └── {companion_id}/ # base_info.json, memory.json, history/, uploads/
│ ├── config/ # Runtime config (model_registry.json, settings.json)
│ ├── modes/ # Per-mode data (workspaces, agents, schedules)
│ ├── logs/ # Usage logs + system logs
│ └── workspaces/ # Output files, uploads, media, projects
│
├── localmodels/ # Model files — never committed
│ ├── gguf/ # GGUF chat models (llama.cpp)
│ └── audio/ # TTS/STT/voice models + voice cloning WAVs
│
└── assets/ # Static assets (showcase images, character sprites)
Contributions are welcome. See CONTRIBUTING.md for guidelines.
git clone https://github.com/Aethvion/Aethvion-Suite.git
cd Aethvion-Suite
pip install -e ".[memory]"
cp .env.example .env # Windows: copy .env.example .env- Docs: /core/documentation/
- Issues: GitHub Issues
- Discussions: GitHub Discussions
A local-first AI super-app — private, personal, and always yours.