[Feature]: Tool Calling Support

[dolphyisnoomi]

Background & Description

With some new LLMs, there are new features called tool calling, such as with Qwen 4B 2507 Instruct with Qwen Agent. I would like LlamaSharp to harness this power and become more agentic, as shown in this Python example:

from qwen_agent.agents import Assistant

llm_cfg = {
    'model': 'Qwen3-4B-Instruct-2507',
    'model_server': 'http://localhost:8000/v1',  # api_base
    'api_key': 'EMPTY',
}

tools = [
    {
        'mcpServers': {
            'time': {
                'command': 'uvx',
                'args': ['mcp-server-time', '--local-timezone=Asia/Shanghai']
            },
            'fetch': {
                'command': 'uvx',
                'args': ['mcp-server-fetch']
            }
        }
    },
    'code_interpreter',
]

bot = Assistant(llm=llm_cfg, function_list=tools)

messages = [{
    'role': 'user',
    'content': 'https://qwenlm.github.io/blog/ Introduce the latest developments of Qwen'
}]

for responses in bot.run(messages=messages):
    pass

print(responses)

This way we can get more out of LlamaSharp inference and help the future of agentic AI.

API & Usage

Maybe you could try integrating Qwen Agent or a similar solution into LlamaSharp.

How to implement

No response

[Webslug]

It can be done in LlamaSharp. I did it today with Tinyagent-1.1b.

You have to load the LLamaSharp.KernelMemory for RAG operations. It's flimsy but works.

using LLamaSharp.KernelMemory;

[rockofme1]

It can be done in LlamaSharp. I did it today with Tinyagent-1.1b.

You have to load the LLamaSharp.KernelMemory for RAG operations. It's flimsy but works.

using LLamaSharp.KernelMemory;

Are there any examples of such an implementation?

[SignalRT]

@rockofme1

You should use SemanticKernel.

• You can find examples of function calling in (SemanticKernel) Here:

https://github.com/microsoft/semantic-kernel/tree/main/dotnet/samples/Concepts

• And you can find examples of Model Context Protocol, here:

https://github.com/microsoft/semantic-kernel/tree/main/dotnet/samples/Demos

It depends if you want to use functions / tools in your code or implemented in a MCP server.