🕸️ DisSysLab — Build distributed apps by connecting functions

DisSysLab (aka dsl) is a lightweight teaching framework where you build distributed programs as graphs of plain Python callables. Each node is just a function — often a call into familiar libraries (NumPy/SciPy, requests, OpenAI). Edges carry messages. Agents run concurrently.

DisSysLab is designed for first-year undergrads. This is an early release; it will evolve, and feedback is welcome.

TL;DR – try it

git clone https://github.com/kmchandy/DisSysLab.git
cd DisSysLab
python -m venv .venv && source .venv/bin/activate   # Windows: .\.venv\Scripts\Activate.ps1
python -m pip install -e .
python -m modules.ch01_networks.simple_network

---

Core idea

• A program is specified as directed graph where nodes are agents and edges are message channels.

• An agent is a callable. It can be:

  • a pure Python function (def f(x): ...),
  • a class instance with __call__,
  • a thin wrapper around a standard library function (NumPy, SciPy, requests),
  • or a service call (e.g., OpenAI) behind a simple adapter.

• You don’t use concurrency primitives such as send, receive, and threads. Instead you connect functions.

from dsl import network

def src():
   for i in range(10):              # zero-arg source → yields messages
    yield i

def double(x): return 2 * x     # transform
def show(x): print(x)           # sink

g = network([
  (src, double),
  (double, show),
])
g.run_network()

👉 Start Here

Module 1. An introduction to dsl.

Modules

1. Module 1 — Intro → modules/ch01_networks/README_1.md
2. Module 2 — Sources → modules/ch02_sources/README_1.md
3. Module 3 — OpenAI agents → modules/ch03_GPT/README_1.md
4. Module 4 — Numerics → modules/ch04_numeric/README.md