m8-core 0.1.2

High-performance Python client for the M8P Hypervisor and Agent Runtime.

m8-core

The Official Python Client & CLI for the M8P Hypervisor.

Accelerate Intelligence. Maximize FLOPs. Optimize CAPEX.

M8P is a unified execution engine for AI Agents, combining Vector Storage, Logic, and Inference into a single high-performance runtime. It eliminates the "Microservice Tax" of legacy stacks by running logic and data in a shared memory space (Zero-Copy Execution).

Key Features

• 🚀 18% Lower Latency: Optimized internal bus eliminates HTTP/JSON overhead.
• 📦 75% Less Complexity: Replaces the typical stack (FastAPI + LangChain + FAISS + Llama.cpp) with a single binary.
• 🧠 Unified Memory: Logic executes directly on data registers without network hops.
• ⚡ Native Streaming: First-class opcode support for real-time token generation.

Installation

pip install m8-core

Optional: For syntax highlighting in the CLI, install the toolkit:

pip install prompt_toolkit

CLI Utility

You can use the command line tool to manage sessions and run scripts directly.

Start a Session:

m8-core start my_session_v1

Run a Script:

m8-core run myscript.m8 --session my_session_v1

Stop/Cleanup:

m8-core stop my_session_v1

API Usage

from m8_core import M8
SESSION_ID = "my_agent_v1"

1. Initialize Memory

init_script = """
vdb_instance AGENT_MEM dim=4096 max_elements=10000
return "Ready"
"""
M8.EnsureExists(SESSION_ID, code=init_script)

2. Run Inference

script = """
store <prompt> Why is the sky blue?
llm_openai <prompt> instance_name n_predict=75 force=true temperature=0.1
llm_instancestatus instance_name <out> #result of inference is stored in <out> by the instancestatus call
return <out>
"""
response = M8.RunSession(SESSION_ID, script)
print(response)

2. Streamming Inference

script = """
store <prompt> Why is the sky blue?
llm_openai <prompt> instance_name n_predict=75 force=true temperature=0.1 stream=true
llm_instancestatus instance_name <out> 
#stream <out> # stream will just duplicate what llm_openai sent
return <out>
"""
response = M8.RunSession(SESSION_ID, script)
print(response)

Requirements

Python 3.8+
requests