08 Agent Architectures
Phase 2 - Agent Core
Stage 8 of 14. Previous: Stage 7 - RAG and Memory. Next: Stage 9 - Building Agents.
Goal
Learn common agent architecture patterns and when each pattern is appropriate.
Learn
- ReAct agents
- RAG agents
- Planner-executor workflows
- Routing and prompt chaining
- Evaluator-optimizer loops
- DAG agents and tree-of-thought patterns
- When a simpler workflow is better than an agent
Build
Compare at least two architectures for the same task and document the tradeoffs in cost, quality, latency, and debuggability.
Exit Criteria
- You can avoid overbuilding when a single LLM call or fixed workflow is enough.
- You can choose an architecture based on task uncertainty and failure modes.
- You can explain the tradeoff between autonomy and control.
Checkpoint
Use the Stage 8 checkpoint before moving on.