Recently watsonx Orchestrate has added support for agentic capabilities. Orchestration agents can route to assistants, skills and agents. This post focusses on the routing between assistants and actions.
With watsonx Orchestrate you can easily create personalized generative AI-powered assistants and agents to automate and accelerate your work, see the documentation What’s new in IBM watsonx Orchestrate: AI agent available on IBM Cloud.
This post is part of a mini-series about agents in Orchestrate:
- watsonx Orchestrate Agent Routing to Assistants (this post)
- watsonx Orchestrate Agent Routing to Skills
- watsonx Orchestrate Agent Routing to Agents
Example
Let’s look at an example. The three questions are answered by three assistants.
Assistants can have three different types of actions:
- Custom-built actions
- AI-guided actions
- Skill-based actions
Custom-built Actions
The following simple example of a custom-built action always returns hardcoded text.
Obviously, these actions are typically more sophisticated. The post Accessing LLMs from watsonx Assistant explains how to leverage Generative AI from watsonx.ai.
AI-guided Actions
AI-guided actions use Large Language Models and some context, for instance documents, to respond to user input.
Skill-based Actions
Skill-based actions allow accessing lots of predefined skills in watsonx Orchestrate without having to code anything.
Additionally, pro-code developers can build their own conversational skills.
AI Agent Configuration
The orchestration agent can be defined in the user interface. First the Large Language Model is selected.
Next the system prompt can be changed.
Via a drop-down box existing assistants can be chosen.
The description of the assistant is key, since the orchestration agent uses this description to determine where to route to.
You can also integrate assistants defined in other watsonx Orchestrate instances.
Next Steps
Check out watsonx Orchestrate, the AI for business productivity, to learn more.