Eidolon Chat for Mac brings the complete Eidolon Home AI ecosystem to Apple Silicon Macs — featuring the same local-first architecture, personalities, memory system, and multimodal capabilities available on Windows and Linux. Built specifically for Apple Silicon, Eidolon runs natively on M1, M2, M3, and M4 chips with full Metal GPU acceleration. No cloud. No subscription. No data leaving your Mac.

Summary

• A complete local AI system, native on Apple Silicon
• Compatible with M1, M2, M3, and M4 Macs (all variants)
• Native Metal GPU acceleration
• No drivers, Homebrew, or external dependencies required
• Requires macOS 13 Ventura or later

---

Local Architecture and Privacy

Eidolon operates under a local-first architecture. By default:

• No data is transmitted externally
• Conversations remain on your Mac
• AI inference is handled locally
• External connections are optional and fully user-controlled

For advanced tasks, users can optionally connect Eidolon to cloud-based AI providers via API, but this behavior is explicitly enabled, fully configurable, and never automatic.

---

Built for Apple Silicon

Eidolon Chat for Mac is compiled natively for arm64. Not a port. Not Rosetta. Not a browser wrapper. The installer automatically sets up the complete environment:

• Standalone Python 3.12 (arm64, no Homebrew required)
• llama-server with Metal GPU acceleration
• Piper TTS and Whisper STT for voice interaction
• Native ffmpeg binaries
• All dependencies inside an isolated virtual environment

Apple Silicon’s unified memory architecture is particularly well suited for local AI inference. Unlike traditional systems with separate RAM and VRAM pools, Apple Silicon allows the GPU and CPU to access the same unified memory space. On systems with 32 GB unified memory, the Mike 12B model can run fully inside GPU-accessible memory with no swap usage, delivering performance comparable to dedicated desktop GPUs.

Supported Hardware

• MacBook Air (M1, M2, M3, M4)
• MacBook Pro (all Apple Silicon variants)
• iMac with Apple Silicon
• Mac mini
• Mac Studio
• Mac Pro with Apple Silicon

Intel Macs are not supported. All included binaries are arm64 only.

Supported Operating Systems

• macOS 13 Ventura
• macOS 14 Sonoma
• macOS 15 Sequoia

Virtual machines are not supported. The licensing system requires identifiable physical hardware.

---

AI Engine and Scalability

Eidolon Chat adapts to available hardware.

Currently Supported Models

Mike 4B Lightweight, responsive, and optimized for 16 GB Macs. Mike 12B Higher-quality responses, improved reasoning depth, and stronger contextual understanding. Recommended for systems with 32 GB RAM or more. Additional models and specialized AI modules will continue to be integrated over time.

---

Core Features

🧠 Full AI Chat

A complete conversational AI system with context-aware responses, multi-turn interaction, and natural continuous dialogue.

---

🎭 Four Built-in Personalities

Thomas — The Sarcastic Analyst

Razor-sharp wit and dry humor. Ideal for technical analysis and code review.

Natasha — The Visionary

Creative, unconventional, and idea-driven. Designed for brainstorming and exploratory thinking.

Jason — The Balanced Mind

Measured reasoning and balanced judgment for complex decisions and structured conversations.

Sue — The Empath

Calm, supportive, and emotionally aware. Designed for reflective conversations and personal guidance.

---

🧩 Persistent Memory

Local long-term memory across sessions. Eidolon can retain:

• relevant interactions
• preferences
• ongoing projects
• recurring discussion topics
• behavioral continuity over time

---

👁️ Computer Vision

Multimodal image understanding integrated directly into the chat. Upload:

• screenshots
• photos
• diagrams
• documents
• visual references

and ask contextual questions about their contents. Image analysis happens locally on your hardware.

---

📄 Document Analysis

Upload documents for:

• summarization
• information extraction
• contextual interpretation
• conversational analysis

---

🧠 PIF (Prompt Instruction Framework)

Advanced behavioral guidance system for:

• role definition
• response shaping
• objective management
• interaction customization

---

🌐 Internet Search (Optional)

Integrated internet search with AI-assisted filtering and summarization. Web access is manually activated and integrated directly into the conversational workflow.

---

🎤 Voice Interaction

Full voice I/O powered by:

• Whisper (speech-to-text)
• Piper TTS (text-to-speech)

Speak naturally with Eidolon and receive spoken responses entirely offline. Available on systems with 16 GB RAM or more.

---

☁️ Optional Cloud Integration

Optional API integration with:

• OpenAI
• Anthropic Claude
• Google Gemini
• other compatible providers

Cloud routing is fully manual and never automatic. Users remain in control of when external models are used.

---

System Philosophy

Eidolon Chat is built around a simple principle: The user remains in control of the AI system. This philosophy shapes every aspect of the platform:

• local execution
• modular architecture
• transparency
• persistent memory
• configurable behavior
• strong privacy guarantees

Eidolon is designed as a practical alternative to centralized cloud AI platforms.

---

System Requirements

Component	Minimum	Recommended
Chip	Apple M1	M2 Pro / M3 / M4
RAM	16 GB unified memory	32 GB or more
Storage	15 GB free	25 GB free
OS	macOS 13 Ventura	macOS 14 Sonoma or later

Virtual machines are not supported.

---

Kickstarter Backers

If your Kickstarter tier included Eidolon Chat, the macOS version is already available in the My Eidolon section of your account at eidolonhub.com.

---

£69 — One Purchase. Fully Local. Yours to Run.

Support: [email protected]