Compare Architectures
Side-by-side: use case, requirements, latency, cost, strengths, weaknesses, failure modes, example — and when to choose each.
Workflow vs AgentSingle agent vs Multi-agentRAG vs Fine-tuningVector search vs BM25Function calling vs MCPDirect API integration vs MCPPlan-then-execute vs ReActShort-term memory vs Long-term memoryLLM-as-judge vs Deterministic evaluatorsHuman-in-the-loop vs Human-on-the-loopSupervisor vs PipelinePrompt caching vs Semantic cachingRouter vs SupervisorFramework vs No framework
| Router | Supervisor | |
|---|---|---|
| Use case | Classify a request once and dispatch to one handler | Coordinate several workers across multiple turns toward one goal |
| Requirements | Classifier (small model or rules) plus handlers | Coordinator loop, workers, shared state, termination |
| Latency | One cheap classification call | Multiple coordination rounds |
| Cost | Minimal; can use a small model | Coordinator tokens on every round |
| Strengths | Simple, fast, easy to evaluate as a classifier | Can decompose, retry, and combine partial results |
| Weaknesses | One shot: cannot recover if it routes wrong | Complexity and bottleneck; harder to trace |
| Failure modes | Misclassified intent lands in the wrong handler | Endless delegation; workers overwrite shared state |
| Example | Billing vs technical vs sales inquiry | Produce a due-diligence report from many sources |
| Choose this when | Choose a router when a single classification decides everything; it is cheaper and more testable. | Choose a supervisor when the task needs iterative decomposition and synthesis across workers. |