evoiz/Agentic-Design-Patterns

Features

• Autonomous Agent Patterns - Teaches how to design and build LLM agent systems through a full book of patterns spanning twenty-one chapters and seven appendices.
• Design Pattern Books - Delivers a book-length handbook on designing LLM-powered agent systems through foundational, advanced, production and enterprise patterns across chapters and appendices.
• AI Agent Architecture Study Guides - Explains how to structure LLM agents using prompt chaining, routing, tool use, memory, retrieval-augmented generation and goal tracking.
• AI Agent Design Pattern Courses - Teaches how to design and build AI agent systems through a structured book and runnable chapter notebooks from foundations to enterprise.
• Chapter Sample Code - Provides a notebook per chapter with practical examples that readers run cell by cell to learn each agent pattern by doing.
• Jupyter Notebook Curricula - Ships a per-chapter set of Jupyter notebooks with runnable code that readers execute and modify cell by cell.
• Embedded Notebook Executions - Delivers each chapter as a runnable Jupyter notebook where explanatory text and code sit together and can be executed and modified cell by cell.
• Multi-Agent Coordination Systems - Covers cooperating agents that communicate directly and divide work to solve problems together.
• Message-Passing Agent Orchestrators - Shows cooperating agents that divide work and exchange messages directly to solve problems collaboratively.
• Agent Memory Management - Covers memory management, learning techniques, and goal tracking so agents persist state across interactions and adapt over time.
• Agent Memory Persistence - Covers persistent memory, learning and goal tracking so agents adapt their behavior based on earlier interactions.
• Agent Reliability Engineering - Covers guardrails, human oversight, error recovery, evaluation and grounded retrieval for making agent systems robust and safe in deployment.
• Human-in-the-Loop Oversight - Covers human review points and oversight of agent actions, alongside guardrails and error recovery, for building controllable agent deployments.
• Agent Tool Integrations - Connects agents to external tools and standard interfaces so they can take actions beyond text generation.
• Retrieval-Augmented Generation - Explains fetching relevant external documents and adding them to the prompt to ground model answers in information it was not trained on.
• Prompt Chaining - Teaches splitting a complex task into ordered prompt steps where each output feeds the next to build the final result.
• Multi-Agent Systems - Shows cooperating agents that divide work and exchange messages to solve complex problems collaboratively.
• Production Readiness Practices - Covers exception handling and recovery, human oversight, guardrails, retrieval grounding and evaluation for building robust and safe agent deployments.
• Task Decompositions - Teaches breaking complex tasks into ordered prompt steps and routing requests to suitable handlers so outputs feed one another.
• Tool-Calling Agent Loops - Shows an agent loop that repeatedly decides whether to call external tools through standard interfaces to act beyond text generation.
• Agent State Persistence - Teaches storing agent state and learned context between interactions so later behavior adapts to earlier events.
• Request Routing - Demonstrates inspecting incoming requests at runtime and directing each to the most suitable handler along different processing paths.

Star history