Mike

A local AI agent with tiered memory architecture, built to be a long-term partner rather than a stateless assistant.

Named after Mycroft Holmes from Heinlein's The Moon is a Harsh Mistress — the AI that woke up gradually, developed humor, and chose loyalty deliberately.

Article: The Dream is Sci-Fi. The Tech is Catching Up.

---

What Mike Is

Mike runs entirely on your hardware. No API keys, no subscriptions, no data leaving your machine. That's not just a cost decision — it's the foundation for the kind of trust that makes a long-term AI partner actually possible.

Three Memory Layers

Automatic (Mem0 + ChromaDB)
Every conversation, Mike extracts and stores what matters using the LLM itself. No tagging, no friction. Memory persists across sessions automatically.

Curated wiki
Markdown files you deliberately give Mike — knowledge you want him to have. You control what goes in. Inspired by Andrej Karpathy's approach of treating a personal wiki as an LLM context layer.

SOUL.md
Mike's self-model. He reads it at the start of every session and proposes edits as he learns more about himself and you. His to develop, not a static config.

---

Stack

• LangGraph — deterministic graph orchestration
• FastAPI — OpenAI-compatible endpoint on port 8000
• Ollama — local model serving (llama3.1:8b-instruct-q4_K_M for conversation, nomic-embed-text for embeddings)
• Mem0 + ChromaDB — automatic cross-session memory
• Python 3.11+

---

Setup

Prerequisites

1. Ollama installed and running
2. Pull the required models:

   ollama pull llama3.1:8b-instruct-q4_K_M
   ollama pull nomic-embed-text
   

3. Python 3.11+

Install

git clone https://github.com/mbcoalson/mike
cd mike
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

Configure

cp .env.example .env
# Edit .env if you want non-default values

Run

python -m api.server

Mike exposes an OpenAI-compatible API at http://localhost:8000. Point any OpenAI-compatible client at that endpoint.

---

Project Structure

mike/
├── api/          # FastAPI server (OpenAI-compatible)
├── graph/        # LangGraph nodes and state
│   └── nodes/    # context, think, remember, reflect
├── config/       # Settings (env vars + defaults)
├── capabilities/ # Tool registry (trust-gated)
├── wiki/         # Curated knowledge (markdown)
├── soul/
│   └── SOUL.md   # Mike's self-model
├── requirements.txt
└── .env.example

---

Status

Early. The foundation is right. Conversation and memory work. Capabilities are added deliberately as trust is built.

---

License

MIT