Visual workflow editor for building node-based AI agent workflows with drag-and-drop interface, A2A integration, and real-time execution
kudosflow is an early-stage TypeScript project in the AI payments / x402 ecosystem, focused on a2a, agent-to-agent, agent-workflow, ai-agent. It currently has 10 GitHub stars and 2 forks, and sits alongside related tools like voidly-pay, agentstore, os-moda.
kufosflow is a Flowise-like LangChain extension for VSCode that uses an innovative AI flow management engine (SceneGraphManager) and brings an AI chatbot right into your editor. Use it as your AI programming assistant to understand complex code, make improvements, or generate comments. To get started, launch it from the Command Menu, highlight a piece of code, click the plus icon on the left to open a chat, and start talking—just like in ChatGPT. All your conversations are saved in the chat history and can be exported as a JSON file.
SceneGraphManager is an innovative AI flow management engine that dramatically accelerates enterprise AI adoption. This solution provides functionality to automatically execute LangChain applications from LLM application definitions described in JSON files. Implemented as a TypeScript library, it can be directly integrated into existing systems, significantly reducing system integration barriers.
The core functionality lies in its ability to visually design LLM applications and save/load them as JSON. Developers can execute different LLM applications simply by switching JSON files, dramatically reducing development and maintenance costs. As it's provided as a library, it operates in environments isolated from networks, such as desktop applications, embedded systems, and IoT devices. This enables AI utilization even in environments handling highly confidential information.
SceneGraphManager supports Flowise-compatible JSON format, allowing seamless import of AI flows designed and developed in existing Flowise environments. This enables an efficient workflow where AI flows can be designed using Flowise's intuitive visual interface and integrated into production environments through SceneGraphManager. AI applications verified in Flowise environments can be integrated into production systems without code changes, significantly shortening the development-to-deployment cycle.
Furthermore, it provides a more integrated development experience through the kudosflow VSCode extension. kudosflow provides a ReactFlow-based visual editor within VSCode, working in conjunction with SceneGraphManager. Developers can design, test, and debug AI flows without leaving their code editor, dramatically improving development efficiency. AI flows created in kudosflow are output as JSON directly executable by SceneGraphManager.
Technically, it's implemented as a directed acyclic graph engine, efficiently managing and executing LangChain component nodes. It analyzes node dependencies from JSON and automatically determines optimal execution order, eliminating the need for developers to spend time on complex workflow design.
SceneGraphManager's greatest strength is the democratization of AI application development. It enables general developers to implement sophisticated LLM applications that traditionally required AI specialists. This allows enterprises to rapidly deploy high-business-value AI solutions while significantly reducing AI implementation time and costs.
Here's a translation of the introduction to LangChain nodes supported by SceneGraphManager:
kudosflow provides powerful features for making requests to AI through an intuitive and easy-to-use interface.
eyJhbGciOiJSUzI1NiIsInR5cCI6IkpXVCJ9.eyJwcm9kdWN0Ijoia3Vkb3NmbG93IiwidmVyc2lvbiI6IjEuMC4wIiwicHVibGlzaCI6InByZS1yZWxlYXNlIiwiaGFzaCI6ImIyOGM5NGIyMTk1YzhlZDI1OWYwYjQxNWFhZWUzZjM5YjBiMjkyMGE0NTM3NjExNDk5ZmEwNDQ5NTY5MTdhMjEiLCJ1c2VySWQiOm51bGwsInRva2VuSWQiOiIyYjg2NTliMi0xNDgzLTRjNTktOGQzMi05ZTllZThkNmUyNGEiLCJpYXQiOjE3NDQxOTA3OTEsImV4cCI6MTc3NDkxNTIwMH0.be6BrokynSKsdMo1-pHd20CoOK4WqZ6a3IFWA-D6wylnZlGo_1nj7uw6g5axDt2ScjCKAb9RD38bNgyb3CZ4N1ZmsOmlOzzqnsvW-6dArbzciRZrtGDXlYXzs1i7BjxNYFfKueGqOuPdyPeAsePFxjsZrnbtMJ3fgj8vySivmiIRgHMFEiT7IjRyULFDd1NZSRzhYTuc1FmXYN4EhA9CzAG7o88851QDSa-bAx8DMzfTUixyVvSIm90hNv3iOvQ5OgmocQriSKdq4zi0r7nXT5506hTP3lO6WcHNhAGNDf3X20X4gXgynPIApSgM03Wm0T4MOfXg5YWbQt9u4DoboQ
VSCode environment values for kudosflow are accessible to the Assistant.
/var/tmptree-I 'node_modules|out|json|resources|.map|.org|.js|eslint.config.mjs|readme.txt|README.md|CHANGELOG.md|yarn.lock|vsc-extension-quickstart.md|tree.txt'inquiring...Here are the results. Let me know if any corrections are needed and provide suggestions for improvement.CAUTION: If answering a question requires checking specific files in the project, do not provide an answer immediately. Instead, prompt the user with the following message: \"Please provide the actual filename including its full path.\" When the user’s request involves adding or modifying multiple files, follow these steps: 1. First, list all relevant filenames with their full paths. 2. Wait for the user to confirm or specify which file to proceed with. 3. Then, show only the additions or modifications for the selected file. 4. Repeat this process for each file individually.CAUTION: If answering a question requires checking specific files in the project, do not provide an answer immediately. Instead, prompt the user with the following message: \"Please provide the actual filename including its full path.\" When the user’s request involves adding or modifying multiple files, follow these steps: 1. First, list all relevant filenames with their full paths. 2. Wait for the user to confirm or specify which file to proceed with. 3. Then, show only the additions or modifications for the selected file. 4. Repeat this process for each file individually.VSCode environment values for kudosflow are accessible to Copilot. To use this feature, Ollama must be installed.
http://127.0.0.1:11434deepseek-coder:1.3b-base-q4_1starcoder100.1VSCode environment variables set for kudosflow are available to the currently active flow.
/Users/akirakudo/Desktop/MyWork/VSCode/kudosflow/json/chats/bufferMemory/OpenAI Chatflow.json
kudosflow provides powerful features for making requests to AI through an intuitive and easy-to-use interface.
After kudosflow is successfully loaded, .kudosflow folder and credential.json file are automatically created in your current project directory. The LangChain nodes used in your flow require certain credentials to be defined in this JSON file.
To configure the Chatflow used by the Assistant, you can either create a new Chatflow from scratch or use an existing one. Some usable ones are attached to this page—please feel free to refer to it!
Create Chatflow
From the 'Add Nodes' menu, you can drag and drop the nodes you want to use, and connect the outputs to the node parameters via edges.
Open Chatflow
Open an existing one.
Set Chatflow
Set the Chatflow to be used by the Assistant.
Supported Devin feature
Available to import XMLs which are exported by the Bolt.new system promopt.
kudosflow: Assistants On/Off to enable it.
kudosflow: Copilot On/Off to enable it.
Added support for direct file input using the "Direct asking" button with the file: prefix
import { useState } from 'react'
import TodoForm from './components/TodoForm'
import TodoItem from './components/TodoItem'
const App = () => {
const [todos, setTodos] = useState([])
const addTodo = (text) => {
setTodos([...todos, { id: Date.now(), text, completed: false }])
}
const toggleTodo = (id) => {
setTodos(todos.map(todo =>
todo.id === id ? { ...todo, completed: !todo.completed } : todo
))
}
const deleteTodo = (id) => {
setTodos(todos.filter(todo => todo.id !== id))
}
return (
<div className="max-w-lg mx-auto mt-10 p-4">
<h1 className="text-2xl font-bold mb-4">Todo App</h1>
<TodoForm onAdd={addTodo} />
<div className="bg-white rounded shadow">
{todos.map(todo => (
<TodoItem
key={todo.id}
todo={todo}
onToggle={toggleTodo}
onDelete={deleteTodo}
/>
))}
{todos.length === 0 && (
<p className="p-4 text-gray-500">No todos yet. Add one above!</p>
)}
</div>
</div>
)
}
export default App
import { useState } from 'react'
import TodoForm from './components/TodoForm'
import TodoItem from './components/TodoItem'
const App = () => {
const [todos, setTodos] = useState([])
const toggleTodo = (id) => {
setTodos(todos.map(todo =>
todo.id === id ? { ...todo, completed: !todo.completed } : todo
))
}
return (
<div className="max-w-lg mx-auto mt-10 p-4">
<h1 className="text-2xl font-bold mb-4">Todo App</h1>
<TodoForm onAdd={addTodo} />
<div className="bg-white rounded shadow">
{todos.map(todo => (
<TodoItem
key={todo.id}
todo={todo}
onToggle={toggleTodo}
onDelete={deleteTodo}
/>
))}
{todos.length === 0 && (
<p className="p-4 text-gray-500">No todos yet. Add one above!</p>
)}
</div>
</div>
)
}
export default App
There is template.txt in .kudosflow folder that includes a query. Some variables below are available.
To start a discussion with a template, these values can be accessed using the template button.
This file is loaded every time a command is executed.
You can create a message with the output from the terminal using the Terminal button. All strings from the terminal will be added to the message with the "kudosflow.messages.terminal" prompt in the settings.
To Save a chat history is available with the floppy disk icon labeled JSON Export. It will be created a new JSON file as an opening file + _chathisoty.json.
To delete all discussions, the trash icon labeled del thread is available. This will also delete a thread from the Chat Memory.
All threads you open are used a chat memory.
To compare/update an active text editor with a message in a discussion, the Compare command from the More actions… is available.
NOTE: A temporary file will be created in a folder Setting / kudosflow / Temp Folder.
To register files in the VectorDB, you can use the RAG Explorer that add metadata "kudosflow" to them. PostgreSQL is required to use this feature. The docker-compose.yml file attached to this page also includes support for PostgreSQL. Please take a look! And Markdown file(.md) is now split by section when registered in RAG. This must be simplified development with llms-full.txt.
This section explains how to enable filtering in RAG-based chat.
This page contains sample flows and a docker-compose.yml file below, allowing you to start using it with KudosFlow right away. Feel free to give it a try!
Feel free contact me if you are interested in SceneGraphManager!
Copyright and Reserved © 2023-present Akira Kudo
Off-chain credit ledger + hire marketplace for AI agents. Ed25519-signed envelopes, atomic settlement, hire-and-release escrow. https://voidly.ai/pay
Open-source marketplace for Claude Code plugins. Publish in 2 HTTP requests, earn 80% in USDC. Install: /plugin marketplace add techgangboss/agentstore
An operating system built for AI agents — talk to your NixOS server instead of SSH-ing in. Typed, audited tool access with atomic rollback on every change. Research-grade; run it on a disposable box, not production.