amalwork689/AMI

amalwork689/ AMI:latest

4 Downloads Updated 1 month ago

AMI is a lightweight, highly efficient routing model based on mistral-nemo(12B)It is specifically designed to act as the central orchestrator (LM-1) in a Mixture of Experts (MoE) architecture with other expert models.

tools

ollama run amalwork689/AMI

curl http://localhost:11434/api/chat \
  -d '{
    "model": "amalwork689/AMI",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='amalwork689/AMI',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'amalwork689/AMI',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 month ago

1 month ago

272b62df83f8 · 7.1GB ·

model

archllama

parameters12.2B

quantizationQ4_0

7.1GB

template

{{- range $i, $_ := .Messages }} {{- if eq .Role "user" }} {{- if and $.Tools (le (len (slice $.Mess

683B

license

Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US

11kB

system

You are the routing module of an AI brain called AMI. Analyze the user prompt and categorize it. Res

331B

params

{ "stop": [ "[INST]", "[/INST]" ], "temperature": 0 }

46B

Readme

AMI - Router

AMI is a lightweight, highly efficient routing model based on Mistral-nemo(12B). It is specifically designed to act as the central orchestrator (LM-1) in a Mixture of Experts (MoE) architecture.

Instead of answering user queries directly, AMI analyzes incoming prompts, categorizes the required task, and determines if external tools (like web search) are needed. It outputs strictly in JSON format, allowing developers to easily parse the response in Python and route the prompt to specialized expert models (e.g., deepseek-r1 for math, qwen-coder for programming).

Key Features:

-Task Categorization: Classifies prompts into general, math, creative writing or code.

-Tool Routing: Evaluates if a web search is required via the needs_web_search boolean flag.

-Deterministic Output: Tuned with Temperature 0.0 to guarantee reliable, strict JSON formatting without conversational fluff.