Autonomous Agent
AgentLoopsAgent
Self-directed agents plan, act, and observe to reach objectives.
Evaluator-Optimizer
WorkflowLoopsWorkflow
Generator and evaluator models refine outputs through feedback loops.
Orchestrator-Workers
WorkflowWorkflow
A central orchestrator delegates subtasks to specialized workers.
Parallelization
WorkflowWorkflow
Run multiple LLM calls in parallel to reduce total latency.
Prompt Chaining
WorkflowWorkflow
LLM calls are sequenced with conditional branching to refine outputs.
Retrieval-Augmented Generation
RAGKnowledgeKnowledge
Ground LLM responses in external knowledge bases for accurate, up-to-date answers.
Routing
WorkflowLow latencyWorkflow
Route incoming requests to specialized services based on intent.
Self-Critique
ReflectionEvaluationReflection
Enable agents to evaluate and improve their own outputs through structured self-reflection.
Tool Calling
Tool UseFunction CallingCapability
Enable LLMs to invoke external tools, APIs, and functions dynamically.