支持MCP的AI对话引擎,支持MCTS分析、FastAPI后端和异步操作,构建高级LLM应用
概览
- 一个生产准备的FastAPI后端和MCP服务器,利用蒙特卡洛树搜索(MCTS)智能探索和评估对话路径,提升响应质量。 - 自动配置 Redis、PostgreSQL/PGVector 和 Prometheus: - : http://localhost:8000(健康检查,条件指标)- : http://localhost:8001/mcp/v1(对话分析)- : localhost:6379 - : localhost:5432 - : http://localhost:9090 - : http://localhost:3000 (管理员/管理员) 通过对对话路径应用高级搜索算法,增强了LLM的响应优化。CAE不再生成单一回应,而是: 1. 在任何对话情境下 2.预测结果 3.(情商、说服力、乐于助人)4.
概述(AI 翻译)
使用搜索算法优化LLM响应——一个生产准备的FastAPI后端和MCP服务器,利用蒙特卡洛树搜索(MCTS)智能探索和评估对话路径,提升响应质量。
目录
- 快速启动(Docker Compose)
- 概述
- 建筑学
- 服务模式与使用情况
- Claude桌面集成
- 手动设置
- 发展
- 测试
- 贡献
- 路线图
- 许可证
快速入门(Docker Compose)
推荐安装 - 自动配置 Redis、PostgreSQL/PGVector 和 Prometheus:
先决条件
- Docker ≥ 24.0
- Docker Compose v2
一命令设置
你得到了什么
- API 后端:
http://localhost:8000(健康检查,条件指标) - MCP 服务器:
http://localhost:8001/mcp/v1(对话分析) - Redis 缓存:
localhost:6379 - PostgreSQL/PGVector:
localhost:5432
可选监控(添加 --profile monitoring)
- 普罗米修斯指标:
http://localhost:9090 - Grafana 仪表盘:
http://localhost:3000(管理员/管理员)
关机
概述
本页由 AI 翻译,技术细节请以原文为准。
README
以下为原文 README 节选,完整内容请查看原文。
Conversational Analysis Engine (CAE)
Optimize LLM responses using search algorithms - A production-ready FastAPI backend and MCP server that leverages Monte Carlo Tree Search (MCTS) to intelligently explore and evaluate conversation paths for enhanced response quality.
Table of Contents
- Quick Start (Docker Compose)
- Overview
- Architecture
- Service Modes & Usage
- Claude Desktop Integration
- Manual Setup
- Development
- Testing
- Contributing
- Roadmap
- License
Quick Start (Docker Compose)
Recommended Installation - Automatically provisions Redis, PostgreSQL/PGVector, and Prometheus:
Prerequisites
- Docker ≥ 24.0
- Docker Compose v2
One-Command Setup
git clone https://github.com/yourusername/ConversationalAnalysisEngine
cd ConversationalAnalysisEngine
docker compose up --build
What You Get
- API Backend:
http://localhost:8000(health checks, conditional metrics) - MCP Server:
http://localhost:8001/mcp/v1(conversation analysis) - Redis Cache:
localhost:6379 - PostgreSQL/PGVector:
localhost:5432
Optional Monitoring (Add --profile monitoring)
docker compose --profile monitoring up --build
- Prometheus Metrics:
http://localhost:9090 - Grafana Dashboard:
http://localhost:3000(admin/admin)
Shutdown
docker compose down -v
Overview
The Conversational Analysis Engine (CAE) enhances LLM response optimization by applying advanced search algorithms to conversation paths. Instead of generating single responses, CAE uses Monte Carlo Tree Search (MCTS) to:
- Generate multiple response branches for any conversation context
- Simulate conversation continuations to predict outcomes
- Score paths based on goal-specific metrics (emotional intelligence, persuasiveness, helpfulness)
- Select the optimal response through intelligent exploration
Dual Architecture:
- FastAPI Backend: Production-ready API with health checks, metrics, and monitoring
- MCP Server: Monte Carlo Tree Search optimization via Model Context Protocol
Architecture
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手?先看 mcpskill.org 教程 →配置
{
"mcpServers": {
"conversational-analysis-engine": {
"command": "docker",
"args": [
"compose", "-f", "/path/to/ConversationalAnalysisEngine/docker-compose.yml",
"up", "mcp", "--build"
],
"env": {
"LLM_API_KEY": "your_openai_api_key"
}
}
}
}