Skip to content

14 - Agents & Reasoning

This module covers core LLM agent technologies including tool use, reasoning strategies, memory systems, and multi-agent collaboration.

Module Structure

14-agents-reasoning/
├── 01-tool-use/                # Tool Calling
├── 02-reasoning/               # Reasoning Strategies
├── 03-memory-systems/          # Memory Systems
├── 04-planning/                # Task Planning
└── 05-multi-agent/             # Multi-Agent

Core Content

01 - Tool Use

TechnologyDescription
Function CallingOpenAI-style function definitions
Tool RegistryDecorator registration, tag management
Tool ExecutorTimeout control, retry mechanism
Structured OutputJSON Schema constraints

02 - Reasoning Strategies

TechnologyDescriptionUse Case
Chain-of-ThoughtStep-by-step reasoningComplex reasoning
ReActReasoning + Action loopTool usage
Tree-of-ThoughtsTree search explorationExploratory problems
Self-ConsistencyMulti-path votingImprove accuracy
ReflectionSelf-correctionError fixing

03 - Memory Systems

TypeDescription
Short-termConversation context window
Long-termVector database storage
EpisodicHistorical interaction records
SemanticKnowledge graphs

04 - Task Planning

TechnologyDescription
Task DecompositionBreak down complex tasks
Plan GenerationStep planning
Plan ExecutionMonitoring and adjustment
Re-planningFailure recovery

05 - Multi-Agent

PatternDescription
DebateMultiple agents discuss to reach consensus
CollaborativeDivision of labor
HierarchicalMain agent coordinates sub-agents

Learning Path

Function Calling → CoT/ReAct → ToT → Memory → Planning → Multi-Agent

Released under the MIT License.