How it works

There are three classes you need to know:

Class	Side	What it does
Server	Server	Listens for TCP connections. Creates a Channel for each client that connects.
Channel	Server	Represents one connected client. You define Network_{action} methods here to handle messages from that client.
ConnectionListener	Client	Connects to a server. You define Network_{action} methods here to handle messages from the server. Call pump() every frame to process them.

Architecture

┌────────────────────┐            ┌───────────────────────┐
│      Client        │            │      Server           │
│                    │            │                       │
│ ConnectionListener │ ◄────────► │  Server               │
│  - pump()          │     TCP    │  - launch()           │
│  - send()          │            │  - on_connect()       │
│  - Network_*()     │            │  - on_disconnect()    │
│                    │            │                       │
│                    │            │  Channel (per client) │
│                    │            │  - send()             │
│                    │            │  - Network_*()        │
└────────────────────┘            └───────────────────────┘

Message dispatch

The design philosophy is simple: messages are Python dicts. You send a dict with an "action" key, and repod automatically routes it to a method named Network_{action} on the other side.

Client sends:  {"action": "move", "x": 10, "y": 20}
                        │
                        ▼
Server calls:  channel.Network_move({"action": "move", "x": 10, "y": 20})

repod handles the TCP connection, serialization (msgpack), framing (length-prefix), and threading for you. You just define handler methods and send dicts.