Agentflow
Complex LLM Workflows from Simple JSON.
No commits in 38 months — may not be actively maintained. See maintained alternatives →
open-sourceagent-frameworks
321
Stars
+0
Stars/month
0
Commits (90d)
0
Releases (6m)
Star Growth
Overview
Agentflow是一个工作流程执行工具,通过人类可读的JSON文件驱动LLM执行复杂的多步骤任务。该工具填补了聊天界面(如ChatGPT)和完全自主系统(如Auto-GPT)之间的空白,提供了一个平衡的解决方案:既有足够的结构来确保可靠执行,又保持了灵活性来处理动态内容。用户可以用简单的英文在JSON中定义工作流程,使用变量来创建基于输入变化的动态输出,并构建自定义函数来扩展文本生成之外的功能。Agentflow通过命令行界面运行,让LLM按步骤执行预定义的任务序列,同时支持实时监控和调试。目前该项目使用OpenAI API作为底层LLM服务,专注于提供可重复、可控的AI工作流程执行能力。
Deep Analysis
Key Differentiator
vs AutoGPT / LangChain agents: deterministic step-by-step workflow execution from JSON definitions — balanced between chat flexibility and autonomous agent unpredictability, with custom function support
⚡ Capabilities
- • Run LLM-powered workflows defined in plain English JSON files
- • Variable substitution for dynamic outputs based on user input
- • Custom function building and execution beyond text generation
- • Step-by-step workflow execution with configurable task settings
- • Verbose real-time task completion output
- • Temperature and function call settings per task
🔗 Integrations
OpenAI API
✓ Best For
- ✓ Developers wanting structured, repeatable LLM workflows vs. freeform chat
- ✓ Multi-step content generation pipelines (e.g., market research → analysis → report)
- ✓ Teams needing predictable LLM execution with human-readable workflow definitions
✗ Not Ideal For
- ✗ Interactive conversational AI applications
- ✗ Autonomous agents needing self-directed decision making
- ✗ Non-developers without Python experience
Languages
Python
Deployment
pip installlocal CLI execution
⚠ Known Limitations
- ⚠ Only supports OpenAI as LLM backend
- ⚠ Currently in development stage
- ⚠ No web UI — command-line only
- ⚠ Workflow definition limited to JSON format
- ⚠ No built-in memory or conversation persistence
Pros
- + 人类可读的JSON格式使非技术用户也能轻松创建和修改AI工作流程
- + 在聊天式交互和完全自主系统之间提供了良好的平衡,确保工作流程的可靠性和可控性
- + 支持自定义函数和变量系统,允许用户扩展功能并创建动态内容生成流程
Cons
- - 目前仍在开发阶段,可能缺乏生产环境所需的稳定性和完整功能
- - 依赖OpenAI API,需要外部服务和API密钥,可能产生使用成本
- - 需要Python环境和手动配置,对非技术用户存在一定的技术门槛
Use Cases
- • 自动化内容生成管道,如批量创建营销文案、产品描述或技术文档
- • 构建需要多个步骤的数据处理工作流程,如信息提取、分析和报告生成
- • 创建可重复的AI辅助业务流程,如客户服务响应模板或内容审核工作流
Getting Started
1. 获取OpenAI API密钥并克隆项目仓库到本地;2. 从example.env创建.env文件并添加API密钥,运行pip install -r requirements.txt安装依赖;3. 使用python -m run --flow=example命令运行示例工作流程,然后修改或创建自己的JSON工作流程文件
Alternatives
A
AutoGPT
AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.
A
AgentPilot
A versatile workflow automation platform to create, organize, and execute AI workflows, from a single LLM to complex AI-driven workflows.
G
Griptape
Modular Python framework for AI agents and workflows with chain-of-thought reasoning, tools, and memory.
T
TaskWeaver
The first "code-first" agent framework for seamlessly planning and executing data analytics tasks.
Compare Agentflow
Maintain Agentflow?
Show your live rank in your README, or put Agentflow in front of every visitor to AgentoolRank.