01 Architecture
Agent Architecture & Design Patterns
Compare ReAct, Plan-and-Execute, Reflexion, Tree-of-Thought, event-driven agents, goal-driven agents, stateful agents, and orchestration frameworks.
DesignFrameworksState
Agentic AI Tutorial Series
Learn how to design, build, and operate production AI agents
A structured DataKnobs tutorial library for agent architecture, memory systems, tool orchestration, planning, multi-agent systems, observability, reliability, and safety.
6 Core Themes
27+ Tutorials
Production Patterns
Tutorial focus
From prototypes to governed agent systems
Move beyond simple agent demos with practical lessons on architecture trade-offs, memory design, reliable tool use, cost-aware planning, multi-agent coordination, and safety controls.
Design patterns
ReAct, Plan-and-Execute, Reflexion, Tree-of-Thought, hierarchical agents.
Production controls
Observability, guardrails, sandboxing, permissioning, and fail-safe execution.
Learning Path
AI agent curriculum organized by production readiness
Start with architecture, then layer in memory, tools, planning, multi-agent coordination, and safety controls.
02 Memory
Memory Systems in Agentic AI
Design short-term, long-term, episodic, compressed, retrieved, hybrid, and self-evolving memory systems for agent workflows.
Vector DBKnowledge GraphEpisodic
03 Tools
Tool Use & Function Calling at Scale
Build reliable tool schemas, dynamic routing, latency-aware orchestration, multi-step tool chaining, failure recovery, and hallucination mitigation.
Function CallingSchemasRecovery
04 Planning
Planning, Reasoning & Decision Making
Understand planning algorithms, reactive execution, cost-aware planning, symbolic reasoning integration, and uncertainty handling.
ReasoningCostUncertainty
05 MAS
Multi-Agent Systems
Explore coordination protocols, negotiation, voting, delegation, conflict resolution, emergent behavior, simulations, and orchestrator choices.
CoordinationSimulationEmergence
06 Safety
Reliability, Safety & Guardrails
Design fail-safe agents with sandboxing, permissioning, prompt-injection defenses, observability, and human review loops.
GuardrailsObservabilityHITL
Featured Tutorials
Start with the tutorials most teams need first
These lessons create a strong foundation before you go deeper into memory, tools, planning, and safety.
Overview
AI Agent Slides
A broad introduction to agentic AI concepts, architecture decisions, and production considerations.
AI Agent SlidesArchitecture
Beyond ReAct
Compare ReAct with planning, reflection, and tree-search style agent reasoning patterns.
Beyond React TutorialOperations
Agent Observability
Trace decisions, tool calls, memory retrieval, errors, and cost across agent execution paths.
Agent Observability TutorialPlanning
Cost-Aware Planning
Balance quality, latency, token usage, tool calls, and budget in agent decision loops.
Cost Aware PlanningAll Tutorials
Browse the complete AI agent slide library
Each card links to a subfolder tutorial page.
01
Agent Architecture & Design Patterns
Beyond ReAct
ReAct vs Plan-and-Execute vs Reflexion vs Tree-of-Thought.
Hierarchical Agents
Manager-worker and planner-executor multi-agent structures.
Event-Driven vs Goal-Driven Agents
When to trigger agents by events versus explicit objectives.
Stateless vs Stateful Agents
Production trade-offs for context, memory, recovery, and governance.
Agent Frameworks
LangGraph, AutoGen, CrewAI, and orchestration architecture.
02
Memory Systems in Agentic AI
Memory Strategies
Long-term vs short-term memory implementation choices.
Memory Architectures
Vector DB, knowledge graph, and hybrid memory designs.
Episodic Memory
Let agents learn from prior tasks, outcomes, and decisions.
Memory Compression
Summarization, retrieval optimization, and context management.
Self-Evolving Memory
Design memory systems that improve over repeated workflows.
03
Tool Use & Function Calling at Scale
04
Planning, Reasoning & Decision Making
Planning Algorithms
DFS/BFS-style reasoning, search, and MCTS analogies.
Planning vs Reactive Execution
Choose planning depth based on task complexity and risk.
Cost-Aware Planning
Manage tokens, latency, model calls, and tool execution cost.
Symbolic + LLM Reasoning
Combine structured logic with language-model reasoning.
Uncertainty Handling
Confidence, ambiguity, fallback, and escalation strategies.
05
Multi-Agent Systems
Multi-Agent Systems
Coordination protocols, negotiation, voting, and delegation.
Emergent Behavior
Understand and test unexpected behavior in agent ecosystems.
Conflict Resolution
Resolve disagreements between agents, plans, or outputs.
MAS Simulation
Build simulation environments for testing multi-agent workflows.
06
Reliability, Safety & Guardrails
Fail-Safe Agents
Graceful degradation, fallback paths, and safe stopping conditions.
Sandboxing Agents
Permissioning and containment for autonomous agent actions.
Prompt Injection Defense
Protect agent workflows from malicious or conflicting instructions.
Agent Observability
Trace reasoning, decisions, tool calls, memory, errors, and approvals.
Reference
Folder Indexes
Build production agents
Use DataKnobs to move from agent tutorials to governed workflows.
Translate agent architecture lessons into production-ready assistants, workflow automation, decision systems, and data products with the right observability, permissions, evaluation, and governance controls.
Explore with DataKnobs
Design agent systems with reliable data, tools, memory, and control points.