agentic-tool-use

README

Agentic Tool Use

Function calling (tool use) with a language model -- the core pattern behind agentic AI systems.

How it works

1. Loads a model using the zerfoo.Load one-line API
2. Defines tools (functions) with OpenAI-compatible schemas via serve.Tool
3. Generates a response with zerfoo.WithTools -- the model decides which tool to call
4. Executes the tool call locally and gets a result
5. Feeds the result back to the model for a natural-language answer

This two-step pattern (plan + execute) is the foundation of agentic systems that combine LLM reasoning with real-world actions like API calls, database queries, or system commands.

Usage

go build -o agentic-tool-use ./examples/agentic-tool-use/

# Default query (weather)
./agentic-tool-use --model path/to/model.gguf

# Custom query
./agentic-tool-use --model path/to/model.gguf --query "What is 42 * 17?"

# Using a HuggingFace model ID
./agentic-tool-use --model google/gemma-3-4b --query "Weather in London?"

Flags

Flag	Default	Description
--model	(required)	Path to GGUF model file or HuggingFace model ID
--query	"What is the weather in San Francisco?"	User query

Architecture

User Query
  -> Model generates tool call (with zerfoo.WithTools)
  -> Application executes tool locally
  -> Tool result fed back to model
  -> Model generates final answer

Documentation

Overview

Command agentic-tool-use demonstrates function calling (tool use) with a language model using the zerfoo one-line API.

The model is given a set of tools (functions) it can invoke. When the model decides to call a tool, the application executes the function locally and feeds the result back for a follow-up generation. This is the core pattern behind agentic systems that combine LLM reasoning with real-world actions.

Usage:

go build -o agentic-tool-use ./examples/agentic-tool-use/
./agentic-tool-use --model path/to/model.gguf
./agentic-tool-use --model path/to/model.gguf --query "What is the weather in Tokyo?"