Skip to content

Latest commit

 

History

History
93 lines (73 loc) · 2.35 KB

File metadata and controls

93 lines (73 loc) · 2.35 KB

Steve AI Agent Architecture

Overview

Steve AI uses a LangChain-inspired agent architecture with the ReAct (Reasoning + Acting) framework for autonomous decision-making in Minecraft.

Core Components

1. Agent Chain (AgentChain.java)

  • Orchestrates the reasoning and action loop
  • Manages chain state across multiple steps
  • Builds context from Steve's memory and environment

2. Agent Executor (AgentExecutor.java)

  • Executes agent decisions
  • Manages tool selection and invocation
  • Handles max iteration limits and error recovery

3. ReAct Agent (ReActAgent.java)

  • Implements the ReAct reasoning framework
  • Generates thoughts based on observations
  • Selects actions using thought history
  • Synthesizes final answers from action results

4. Tools (ToolWrapper.java)

Available tools for agent interaction:

  • Build: Structure construction
  • Mine: Resource gathering
  • Attack: Combat actions
  • Pathfind: Navigation

5. Memory Systems

Conversational Memory (ConversationalMemory.java)

  • Stores chat history with timestamps
  • Maintains user/assistant message pairs
  • Token-limited buffer for context management

Vector Store (VectorStore.java)

  • Semantic search over past experiences
  • Cosine similarity for relevance ranking
  • 384-dimensional embeddings
  • Metadata-enriched storage

6. Prompt Management (PromptTemplate.java)

  • Dynamic prompt formatting
  • Variable substitution
  • Template validation

Architecture Flow

User Input
    ↓
ReActAgent.reason()
    ↓
Generate Thought ← Relevant Memories (Vector Store)
    ↓
Select Action
    ↓
AgentExecutor.execute()
    ↓
ToolWrapper.invoke()
    ↓
Action Result
    ↓
Update Memory
    ↓
Synthesize Answer

Configuration

Agent behavior is controlled via langchain_config.json:

  • Model selection and parameters
  • Memory limits and vector store settings
  • Tool definitions
  • Prompt templates

Integration with Minecraft

The agent system integrates with Minecraft through:

  • SteveEntity: The physical entity in the game
  • ActionExecutor: Bridges agent decisions to game actions
  • SteveMemory: Persistent memory across game sessions

Future Enhancements

  • Multi-agent coordination using shared memory
  • Long-term episodic memory with summarization
  • Tool learning and dynamic tool creation
  • Hierarchical planning for complex tasks