🪐 Uniflow

A high-performance, extremely flexible, and easily extensible universal workflow engine.

📝 Overview

Uniflow is designed to manage a wide range of tasks, from short-term jobs to long-term processes. It supports declarative workflow definitions and allows for dynamic changes to data flows. With built-in extensions, you can implement complex workflows and add or remove nodes to expand its functionality as needed.

This system empowers you to deliver customized experiences through your service and continuously enhance its capabilities.

🎯 Core Values

Performance: Maximize throughput and minimize latency across various environments.
Flexibility: Modify and adjust workflows in real-time.
Extensibility: Easily extend functionality with new components.

🚀 Quick Start

🛠️ Build and Install

To begin, install Go 1.23 or later. Then, follow these steps to build the source code:

git clone https://github.com/siyul-park/uniflow

cd uniflow

make init
make build

The executable will be located in the dist directory after building.

⚡ Running an Example

To run a basic HTTP request handler example using ping.yaml:

- kind: listener
  name: listener
  protocol: http
  port: 8000
  ports:
    out:
      - name: router
        port: in

- kind: router
  name: router
  routes:
    - method: GET
      path: /ping
      port: out[0]
  ports:
    out[0]:
      - name: pong
        port: in

- kind: snippet
  name: pong
  language: text
  code: pong

To start the workflow, run:

uniflow start --from-nodes example/ping.yaml

Verify it's running by calling the HTTP endpoint:

curl localhost:8000/ping
pong#

⚙️ Configuration

Settings can be adjusted via the .uniflow.toml file or environment variables.

TOML Key	Environment Variable Key	Example
`database.url`	`DATABASE.URL`	`mem://` or `mongodb://`
`database.name`	`DATABASE.NAME`	-
`collection.nodes`	`COLLECTION.NODES`	`nodes`
`collection.secrets`	`COLLECTION.SECRETS`	`secrets`

📊 Benchmark

The benchmark below was conducted on a Contabo VPS S SSD (4 cores, 8GB) using the Apache HTTP benchmarking tool. It measures the performance of the ping.yaml workflow, including listener, router, and snippet nodes.

ab -n 102400 -c 1024 http://127.0.0.1:8000/ping

This is ApacheBench, Version 2.3 <$Revision: 1879490 $>
Benchmarking 127.0.0.1 (be patient)
Server Hostname:        127.0.0.1
Server Port:            8000
Document Path:          /ping
Document Length:        4 bytes
Concurrency Level:      1024
Time taken for tests:   122.866 seconds
Complete requests:      1024000
Failed requests:        0
Total transferred:      122880000 bytes
HTML transferred:       4096000 bytes
Requests per second:    8334.29 [#/sec] (mean)
Time per request:       122.866 [ms] (mean)
Time per request:       0.120 [ms] (mean, across all concurrent requests)
Transfer rate:          976.67 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    2   3.8      0      56
Processing:     0  121  53.4    121     593
Waiting:        0  120  53.4    121     592
Total:          0  123  53.3    123     594

Percentage of the requests served within a certain time (ms)
  50%    123
  66%    143
  75%    155
  80%    163
  90%    185
  95%    207
  98%    240
  99%    266
 100%    594 (longest request)

📚 Learn More

Getting Started: Learn how to install the CLI and manage workflows.
Key Concepts: Understand nodes, connections, ports, and packets.
Architecture: Explore how workflows are executed and node specifications are loaded.
Debugging: Learn how to debug workflows, set breakpoints, and start sessions.
User Extensions: Learn to add features and integrate with external services.

🌐 Community and Support

Discussion Forum: Share questions and feedback.
Issue Tracker: Submit bug reports and request new features.

📜 License

This project is released under the MIT License. You are free to use, modify, and distribute it as per the terms of the license.

siyul-park / uniflow

readme