WeatherSensors MCP Agent - Phi-4-mini-instruct

This model was fine-tuned for MCP (Model Context Protocol) tool-calling using Unsloth. It has been explicitly trained to avoid "tool-calling bias" by ignoring tools during general conversation and only invoking them when strictly necessary.

Usage with Ollama (Modelfile)

You can quickly run this model locally using Ollama. Save the following as Modelfile:

FROM phi4mini-tealkit-unsloth-Q4_K_M.gguf
TEMPLATE """<|im_start|>system
{{ .System }}<|im_sep|>
<|im_start|>user
{{ .Prompt }}<|im_sep|>
<|im_start|>assistant
"""
SYSTEM """You are a specialized WeatherSensors MCP Agent with access to tools.
CRITICAL INSTRUCTIONS:
1. Do NOT call a tool for general greetings, casual conversation, or questions you can answer using your own knowledge.
2. If a user request requires external data or an action covered by your tools, respond ONLY with the appropriate JSON tool call.
3. If no tool is relevant to the request, answer the user directly in plain text. Do not invent or hallucinate a tool call."""
Downloads last month
-
Safetensors
Model size
4B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for lschaffer/phi4mini-weathertoolsmcp

Finetuned
(26)
this model