Stream原生AI代理，由Timeplus提供支持

PulseBot 是一个轻量级、可扩展的 AI 代理框架，利用 Timeplus 流式数据库作为实时消息路由、可观测性和内存存储的骨干。

✨ 特色

• 流原生架构 - 所有通信都通过Timeplus流流
• 多提供商LLM支持 - Anthropic Claude、OpenAI、OpenRouter、Ollama 和 NVIDIA
• 多代理协作 - 通过基于看板的消息队列为复杂任务生成协调代理团队
• 向量存储器 - 使用存储在 Timeplus 中的嵌入进行语义搜索
• SQL-原生调度 - Timeplus 任务取代传统的 cron 作业
• 交互式工作区 - 构建和发布动态工件和可运行的网页应用
• 可扩展技能 - 基于插件的工具系统，兼容 OpenClaw，支持 ClawHub 注册表
• 工具调用钩子 - 在执行前后拦截、批准、拒绝或修改工具调用
• 多渠道 - Telegram、网络聊天，且可便捷扩展至 Slack/WhatsApp
• 实时可观测性 - 所有LLM调用和工具执行记录在流中

PulseBot

Stream-native AI Agent powered by Timeplus

PulseBot is a lightweight, extensible AI agent framework that uses Timeplus streaming database as its backbone for real-time message routing, observability, and memory storage.

✨ Features

• Stream-Native Architecture - All communication flows through Timeplus streams
• Multi-Provider LLM Support - Anthropic Claude, OpenAI, OpenRouter, Ollama, and NVIDIA
• Multi-Agent Collaboration - Spawn coordinated agent teams for complex tasks via kanban-based message queues
• Vector Memory - Semantic search using embeddings stored in Timeplus
• SQL-Native Scheduling - Timeplus Tasks replace traditional cron jobs
• Interactive Workspaces - Build and publish dynamic artifacts and runnable web apps
• Extensible Skills - Plugin-based tool system with OpenClaw compatibility and ClawHub registry
• Tool Call Hooks - Intercept, approve, deny, or modify tool calls before/after execution
• Multi-Channel - Telegram, webchat, with easy extension to Slack/WhatsApp
• Real-Time Observability - All LLM calls and tool executions logged to streams
• Production Ready - Docker deployment, async architecture, structured logging

🏗️ Architecture

🚀 Quick Start

Prerequisites

• Python 3.11+
• Timeplus or Proton

Installation

# Clone repository
git clone https://github.com/timeplus-io/PulseBot.git
cd PulseBot

# Install package
pip install -e .

# Generate default config
pulsebot init

Configuration

Edit config.yaml with your API keys:

agent:
  name: "PulseBot"
  model: "claude-sonnet-4-20250514"  # or use ollama model

providers:
  anthropic:
    api_key: ${ANTHROPIC_API_KEY}
  
  # For local testing with Ollama
  ollama:
    enabled: true
    host: "http://localhost:11434"
    default_model: "llama3"

workspace:
  base_dir: "./workspaces"
  workspace_port: 8001
  internal_api_key: "${WORKSPACE_INTERNAL_KEY}"

Using Ollama (Local Testing)

# Install Ollama (macOS)
brew install ollama

# Pull a model
ollama pull llama3

# Start Ollama server
ollama serve

# Update config.yaml to use Ollama
# Set agent.provider: "ollama" and agent.model: "llama3"

Setup & Run