ReAct (Reasoning + Acting) execution engine with step-by-step observation injection and scratchpad history tracking
ReAct (Reasoning + Acting) execution engine with step-by-step observation injection and scratchpad history tracking
A TypeScript template for ReAct agents that use local or OpenAI-compatible models, call local and MCP tools, preserve conversation state, and stream responses over HTTP.
ReAct (Reasoning and Acting) agent implementation using Spring Boot 4.0.1 and Spring AI 2.0.0-M1. This project demonstrates how to build an AI agent that combines reasoning with tool execution, integrating OpenAI for response generation and Tavily for web search. It includes a node-based system (AgentNode and ToolNode)
A minimal, dependency-free ReAct (Reasoning + Acting) agent loop framework for TypeScript — bring your own LLM, register tools, and run a Think-Act-Observe loop with pluggable guardrail hooks.
A minimal, dependency-free implementation of the ReAct (Reasoning + Acting) agent loop in ~110 lines of Python — no LangChain, no magic, just a transparent loop with Pydantic-validated tool calls, supporting OpenAI, Anthropic, and local Ollama models.
A multi-action ReAct runtime with dependency-aware tool execution, parallel actions, and iterative reasoning.
To associate your repository with the reasoning-and-acting topic, visit your repo's landing page and select "manage topics."