| {"pattern": "Model Routing", "category": "Orchestration", "description": "Routes each request to the model best suited for the task based on capability, latency, cost, modality, or policy.", "components": ["router", "model registry", "evaluation signals", "fallbacks"], "use_case": "Multi-model AI systems", "complexity": "medium"} | |
| {"pattern": "Planner-Executor", "category": "Agents", "description": "Separates planning from execution so one component creates a task plan while another performs the actions.", "components": ["planner", "executor", "state", "tools"], "use_case": "Long-horizon agent workflows", "complexity": "medium"} | |
| {"pattern": "Supervisor-Worker", "category": "Multi-Agent", "description": "Uses a supervising agent to delegate specialized subtasks to worker agents and combine their results.", "components": ["supervisor", "worker agents", "task queue", "result merger"], "use_case": "Parallel research and enterprise automation", "complexity": "high"} | |
| {"pattern": "Retrieval-Augmented Generation", "category": "Knowledge", "description": "Retrieves external knowledge at inference time and injects relevant context into model generation.", "components": ["retriever", "index", "reranker", "generator"], "use_case": "Knowledge-grounded assistants", "complexity": "medium"} | |
| {"pattern": "Tool-Using Agent", "category": "Agents", "description": "Lets a model select and call external tools such as APIs, browsers, databases, or code execution.", "components": ["agent model", "tool registry", "schemas", "execution layer"], "use_case": "Autonomous task execution", "complexity": "medium"} | |
| {"pattern": "Memory-Augmented Agent", "category": "Memory", "description": "Adds persistent memory so an agent can reuse relevant information across steps or sessions.", "components": ["agent", "memory store", "retrieval", "memory policy"], "use_case": "Persistent assistants and long-horizon agents", "complexity": "medium"} | |
| {"pattern": "Verifier Loop", "category": "Validation", "description": "Uses a separate verifier, critic, test, or model to check intermediate or final outputs before acceptance.", "components": ["generator", "verifier", "retry policy", "acceptance rule"], "use_case": "High-reliability reasoning and coding", "complexity": "medium"} | |
| {"pattern": "Human-in-the-Loop", "category": "Control", "description": "Introduces explicit human review or approval at selected stages of an AI workflow.", "components": ["agent", "risk policy", "approval interface", "audit trail"], "use_case": "High-impact or regulated workflows", "complexity": "medium"} | |
| {"pattern": "World-Model Planning", "category": "World Models", "description": "Uses a predictive model of the environment to simulate possible outcomes before selecting an action.", "components": ["world model", "planner", "state representation", "policy"], "use_case": "Robotics, simulation, and Physical AI", "complexity": "high"} | |
| {"pattern": "Fallback Routing", "category": "Reliability", "description": "Switches to an alternate model, provider, tool, or execution path when the preferred path fails.", "components": ["primary route", "fallback route", "failure detector", "policy"], "use_case": "Resilient production AI systems", "complexity": "medium"} | |
| {"pattern": "Multi-Agent Debate", "category": "Multi-Agent", "description": "Multiple agents propose or critique answers before a final synthesis or decision is produced.", "components": ["proposer agents", "critic agents", "judge", "shared context"], "use_case": "Reasoning and review workflows", "complexity": "high"} | |
| {"pattern": "Agentic RAG", "category": "Knowledge", "description": "Combines retrieval with agent planning so the system can iteratively search, inspect, and refine evidence.", "components": ["agent", "retriever", "search tools", "memory", "verifier"], "use_case": "Complex research and evidence synthesis", "complexity": "high"} | |
| {"pattern": "Reflection Loop", "category": "Reasoning", "description": "Lets a model review its own output, identify weaknesses, and produce an improved revision.", "components": ["generator", "reflection prompt", "revision loop"], "use_case": "Reasoning, writing, and coding improvement", "complexity": "low"} | |
| {"pattern": "Self-Consistency", "category": "Reasoning", "description": "Generates multiple candidate solutions and selects the most consistent result across them.", "components": ["sampler", "candidate generator", "aggregator"], "use_case": "Reasoning and uncertainty reduction", "complexity": "low"} | |
| {"pattern": "Mixture of Agents", "category": "Multi-Agent", "description": "Combines outputs from multiple specialized agents using a final aggregation or synthesis stage.", "components": ["specialized agents", "router", "aggregator"], "use_case": "Broad multi-domain tasks", "complexity": "high"} | |
| {"pattern": "Event-Driven Agent", "category": "Agents", "description": "Triggers agent actions from events such as messages, file changes, webhooks, or system state changes.", "components": ["event bus", "agent", "handlers", "state"], "use_case": "Automation and monitoring", "complexity": "medium"} | |
| {"pattern": "State Machine Agent", "category": "Agents", "description": "Constrains agent behavior to explicit states and transitions for predictable workflow execution.", "components": ["state machine", "agent", "transition rules", "tools"], "use_case": "Reliable structured workflows", "complexity": "medium"} | |
| {"pattern": "Graph-Based Workflow", "category": "Orchestration", "description": "Represents an AI workflow as nodes and edges so branching, looping, and dependencies are explicit.", "components": ["workflow graph", "nodes", "edges", "state"], "use_case": "Complex agent orchestration", "complexity": "medium"} | |
| {"pattern": "Checkpointed Agent", "category": "Memory", "description": "Persists task state at checkpoints so long-running work can recover after interruption or failure.", "components": ["agent", "checkpoint store", "state serializer", "recovery logic"], "use_case": "Long-horizon workflows", "complexity": "medium"} | |
| {"pattern": "Hierarchical Memory", "category": "Memory", "description": "Separates short-term, task-level, and long-term memory to improve context management and retrieval quality.", "components": ["working memory", "task memory", "long-term memory", "retrieval policy"], "use_case": "Persistent agent systems", "complexity": "high"} | |
| {"pattern": "Context Compression", "category": "Memory", "description": "Compresses older interaction history into summaries or structured state to preserve useful context efficiently.", "components": ["summarizer", "context store", "retrieval policy"], "use_case": "Long-context agent workflows", "complexity": "medium"} | |
| {"pattern": "Permission-Gated Tool Use", "category": "Control", "description": "Requires policy checks or human approval before an agent can execute selected tools or actions.", "components": ["tool registry", "permission engine", "identity", "approval layer"], "use_case": "Enterprise agents and high-impact automation", "complexity": "high"} | |
| {"pattern": "Sandboxed Execution", "category": "Control", "description": "Runs model-generated code or actions inside an isolated environment with restricted permissions and resources.", "components": ["sandbox", "resource limits", "execution engine", "audit"], "use_case": "Coding agents and untrusted execution", "complexity": "medium"} | |
| {"pattern": "Observability-First Agent", "category": "Observability", "description": "Captures traces, tool calls, state transitions, costs, and failures as first-class runtime data.", "components": ["tracing", "logs", "metrics", "audit store"], "use_case": "Production agent operations", "complexity": "medium"} | |
| {"pattern": "Cost-Aware Routing", "category": "Orchestration", "description": "Chooses models or tools using both capability requirements and explicit cost constraints.", "components": ["router", "cost model", "capability registry", "budget policy"], "use_case": "Cost-efficient AI platforms", "complexity": "medium"} | |
| {"pattern": "Latency-Aware Routing", "category": "Orchestration", "description": "Selects execution paths based on latency budgets while preserving minimum capability requirements.", "components": ["router", "latency telemetry", "SLA policy", "fallbacks"], "use_case": "Real-time AI applications", "complexity": "medium"} | |
| {"pattern": "Model Cascade", "category": "Orchestration", "description": "Starts with a cheaper or smaller model and escalates to stronger models only when needed.", "components": ["small model", "confidence gate", "large model", "router"], "use_case": "Efficient inference systems", "complexity": "medium"} | |
| {"pattern": "Ensemble Verification", "category": "Validation", "description": "Uses several independent models or methods to cross-check a result before accepting it.", "components": ["candidate output", "verifiers", "aggregator", "acceptance policy"], "use_case": "High-confidence AI decisions", "complexity": "high"} | |
| {"pattern": "Synthetic Data Flywheel", "category": "Data", "description": "Generates synthetic examples, evaluates them, filters them, and feeds high-quality samples back into training.", "components": ["generator", "quality filter", "deduplication", "training pipeline"], "use_case": "Model improvement and domain adaptation", "complexity": "high"} | |
| {"pattern": "Active Learning Loop", "category": "Data", "description": "Selects uncertain or high-value examples for human or model labeling and uses them to improve the system.", "components": ["model", "uncertainty scorer", "labeler", "training loop"], "use_case": "Efficient supervised data collection", "complexity": "high"} | |
| {"pattern": "Evaluation Harness", "category": "Validation", "description": "Standardizes datasets, prompts, metrics, and execution settings for repeatable model or agent evaluation.", "components": ["task suite", "runner", "metrics", "result store"], "use_case": "Model and agent benchmarking", "complexity": "medium"} | |
| {"pattern": "Shadow Deployment", "category": "Validation", "description": "Runs a new AI system alongside production traffic without affecting users, enabling safe comparison.", "components": ["production system", "candidate system", "traffic mirror", "evaluation"], "use_case": "Safe model upgrades", "complexity": "high"} | |
| {"pattern": "Canary Model Rollout", "category": "Operations", "description": "Deploys a new model to a small portion of traffic before wider release while monitoring quality and failures.", "components": ["traffic splitter", "candidate model", "monitoring", "rollback"], "use_case": "Production model deployment", "complexity": "medium"} | |
| {"pattern": "Human Escalation", "category": "Control", "description": "Lets the AI system transfer a task to a human when uncertainty, risk, or policy thresholds are exceeded.", "components": ["risk detector", "escalation policy", "human queue", "handoff context"], "use_case": "Customer support and regulated automation", "complexity": "medium"} | |
| {"pattern": "Structured Output Contract", "category": "Interoperability", "description": "Requires models to produce outputs that conform to a predefined schema for reliable downstream processing.", "components": ["schema", "validator", "model", "parser"], "use_case": "Tool use and software integration", "complexity": "low"} | |
| {"pattern": "Capability Discovery", "category": "Interoperability", "description": "Allows agents or tools to advertise available capabilities and interface requirements dynamically.", "components": ["capability registry", "schemas", "discovery protocol", "router"], "use_case": "Open agent ecosystems", "complexity": "high"} | |
| {"pattern": "Agent-to-Agent Delegation", "category": "Interoperability", "description": "Lets one agent delegate a task to another specialized agent while preserving context and expected outputs.", "components": ["source agent", "target agent", "message schema", "task contract"], "use_case": "Distributed multi-agent systems", "complexity": "high"} | |
| {"pattern": "World-State Synchronization", "category": "World Models", "description": "Keeps the agent's internal representation aligned with changing external environment state.", "components": ["observation layer", "state model", "update mechanism", "planner"], "use_case": "Robotics and dynamic environments", "complexity": "high"} | |
| {"pattern": "Sensor Fusion Pipeline", "category": "Physical AI", "description": "Combines multiple sensor modalities into a unified representation for perception and downstream decision-making.", "components": ["sensors", "fusion model", "time synchronization", "state estimator"], "use_case": "Robotics, vehicles, and industrial AI", "complexity": "high"} | |
| {"pattern": "Edge-Cloud Split", "category": "Physical AI", "description": "Splits AI processing between low-latency edge devices and more capable cloud infrastructure.", "components": ["edge model", "cloud model", "router", "connectivity layer"], "use_case": "Robotics, wearables, and mobile AI", "complexity": "high"} | |