chapter two
2 Turn a task into an agent architecture
This chapter covers
- Identifying the capabilities an agent needs with the seven-module capability stack
- Connecting capabilities through the Perception-Reasoning-Action loop and runtime architecture
- Comparing execution structures with the two-axis pattern map
- Deciding when one agent needs delegation and where autonomy needs safeguards
- Applying the decisions to Argus and run its first PRA loop against a real code diff
"Each pattern describes a problem which occurs over and over again in our environment, and then describes the core of the solution to that problem, in such a way that you can use this solution a million times over."
— Christopher Alexander, A Pattern Language (1977)
An agent architecture begins with a business task and its constraints: what the system must perceive, remember, decide, and change; which failures are tolerable; and where people must remain in control. This chapter gives you a sequence for turning those constraints into an architecture. You will identify the capabilities the task needs, connect them through a runtime, compare execution structures, and sketch a first design before selecting any pattern.