Problem
Currently all executions have hooks. We need to pare down the requirements of the execution engine to allow invoking an agent definition without any hooks (known as a Task Agent)
Proposed solution
Split the task invocation into two "types" of tasks:
- Script tasks
- LLM tasks
- LLM tasks are simply a framework provided script task that calls an engine
Alternatives considered
No response
Area
Orchestrator
Additional context
No response
Problem
Currently all executions have hooks. We need to pare down the requirements of the execution engine to allow invoking an agent definition without any hooks (known as a Task Agent)
Proposed solution
Split the task invocation into two "types" of tasks:
Alternatives considered
No response
Area
Orchestrator
Additional context
No response