Tool Steps

Tool steps are the building blocks that define the logic and functionality of custom tools within the HALO platform. By combining different types of tool steps, you can create sophisticated workflows that automate processes, integrate with external systems, and deliver personalized responses. This knowledge article explores the various tool steps available and how they can be effectively utilized to enhance your agent's capabilities.

Available Tool Steps

1. Tools

Description: This step allows you to call an existing tool within your new tool.
Usage: Ideal for leveraging pre-built functionality without recreating it, promoting code reuse and maintaining consistency across your toolset.
Example: A customer service tool might call a separate "account validation" tool before proceeding with specific account actions.

2. API (HTTP) Call

Description: Enables interaction with external systems or databases through API endpoints.
Usage: Requires proper API documentation, endpoint URLs, and authentication credentials to retrieve, delete, add, or modify data in external systems.
Considerations:

• Ensure you have all necessary connection details before implementation

• Verify API permissions align with your intended operations

• Handle response codes appropriately to manage errors

3. LLM Interaction

Description: Leverages large language model capabilities to process inputs and generate customized outputs.
Usage: Takes parameters from your tool or data retrieved from API calls to create tailored responses, summaries, or transformations.
Example: Transforming technical data retrieved from a system API into a customer-friendly explanation.

4. Code Execution

Description: Allows execution of custom Python code within your tool.
Usage: Provides flexibility for complex data manipulation, calculations, or logic that may be difficult to achieve with other tool steps.
Considerations: Maintain security best practices when creating executable code.

5. User Interaction

Description: Creates standardized prompts for gathering information from users.
Usage: Defines specific questions with optional recommended answers to guide users through providing necessary information.
Benefits:

• Ensures consistent data collection

• Improves user experience with clear guidance

• Can reduce errors by suggesting valid response options

6. Approval Step

Description: Integrates human oversight through structured approval workflows.
Usage: Links to predefined approval flows (configured in HALO Governance) that route decisions to authorized employees.
Process: When this step is reached, the tool pauses execution until the appropriate reviewer provides approval, after which the tool continues its execution.

7. Knowledge

Description: Retrieves specific information from your knowledge base using targeted queries.
Usage: Searches knowledge repositories based on parameters passed to the tool.
Advanced Options: By specifying context, you can narrow search results to include or exclude certain categories of knowledge, which is particularly useful when your knowledge base contains information about multiple entities.

8. Logical Step

Description: Implements conditional logic to determine the execution path of subsequent steps.
Usage: Functions similarly to an "if statement" in programming, where different conditions trigger different paths through the remainder of the tool.
Example: If a customer's account status is "premium," the tool might follow one path with enhanced options, while standard accounts follow another.

9. Parallel Execution

Description: Enables simultaneous execution of multiple tool steps.
Usage: Improves efficiency by running independent processes concurrently rather than sequentially.
Outcome: Guarantees multiple results from your tool, with each parallel path potentially producing different outputs.

Combining Tool Steps for Advanced Workflows

The true power of HALO's tooling capabilities emerges when combining different tool steps into cohesive workflows. For example:

1. Begin with User Interaction to collect necessary information

2. Use API Call to retrieve related data from your systems

3. Apply Knowledge steps to gather relevant contextual information

4. Implement Logical Steps to determine the appropriate course of action

5. For complex cases, insert an Approval Step for human oversight

6. Utilize LLM Interaction to format the final response in natural language

Conclusion

Tool steps provide a flexible framework for creating custom capabilities within your HALO implementation. By understanding the unique strengths of each tool step type and how they can work together, you can design efficient, powerful tools that enhance your agent's ability to serve users and integrate with your existing systems.

When designing tools, consider the user journey and business requirements to determine which combination of tool steps will deliver the most effective solution. With practice, you'll be able to create increasingly sophisticated tools that streamline operations and improve user experiences.