At the time of writing this doc, prompt tools are only compatible with some models. Only OpenAI models and Anthropic models support it.
Prompt Tools
Overview
Function calling is the capability to connect LLMs to external tools and APIs, enabling effective tool usage and interaction with external systems. This functionality allows LLMs to execute specific tasks and fetch data by invoking predefined functions based on natural language prompts. On our platform, you can explore function calling through the use of Prompt Tools. This allows you to mimic and test an agentic flow.
Create a prompt tool
Maxim allows you to create three different types of prompt tools to enhance your workflows:
-
Function-based Prompt tools
- Create prompt tools by writing functions directly within the editor.
- These functions are fully customizable and can be tailored to your specific use cases.
- Learn more about creating function-based tools here.
-
API-integrated Prompt tools
- Leverage functions exposed through APIs as prompt tools.
- This allows seamless integration of external services directly into your prompts.
- Alternatively, you can integrate functions exposed through APIs for use as prompt tools in your prompts. Read more about this here
-
Schema-based Prompt tools
- Use the Schema tool type to define structured prompts with direct schema input.
- Ideal for evaluating tool call accuracy, schema-based tools ensure outputs align with predefined structures.
Mechanism of Prompt tool
Here is a dummy prompt tool for your reference. "Pincode" is a prompt tool that returns the name of the place corresponding to a given Pincode. The "Pincode" tool is sent to the model along with the user prompt.
Use a Prompt tool
You can add a prompt tool to a prompt through its configuration. Additionally, you can attach multiple prompt tools to an individual prompt to validate the model's ability to recognize user input and appropriately select the correct prompt tool for function execution. For more information, please visit this page.