python.qmd

---
title: "Python"
---

## Getting Started

Install [`rembus`](https://pypi.org/project/rembus/) using `pip` or `uv`:

```bash
pip install rembus
```

```bash
uv add rembus
```

For the basic concepts about node types and addressing see the
[Julia docs](julia.qmd#rembus-node-types)

Rembus provides both synchronous and asynchronous APIs:

```python
import rembus as rb

# Synchronous API
cli = rb.node("mynode")
```

```python
import rembus as rb

# Asynchronous API
cli = await rembus.component("mynode")
```

The returned `cli` object represents a Rembus component and exposes the core
communication primitives:

- `rpc` — request a remote service
- `expose` — expose a local function as an RPC service
- `publish` — publish messages to a topic
- `subscribe` — subscribe to topic updates

---

## RPC Call

```{mermaid}
flowchart LR
  C(("Client")) --> S((Server))
  style C fill:#009E73, color:white
  style S fill:#0072B2, color:white
```

This example shows a Python client invoking an RPC service implemented in Julia.

### Julia Server

```julia
using Statistics
using Rembus

function stats(df)
    return Dict(
        "min"  => minimum(df.value),
        "max"  => maximum(df.value),
        "mean" => mean(df.value),
        "std"  => std(df.value)
    )
end

bro = component()
expose(bro, stats)

println("up and running")
wait(bro)
```

### Python Client

```python
import rembus as rb
import pandas as pd
from random import random

nrows = 1_000_000

df = pd.DataFrame({
    "name": [f"kpi_{i}" for i in range(nrows)],
    "ts": range(nrows),
    "value": [random() for _ in range(nrows)],
})

cli = rembus.node("foo")
summary = cli.rpc("stats", df)
cli.close()
```

### Data Flow

1. The Python client sends a pandas DataFrame.
2. Rembus converts it into a Julia DataFrame.
3. The Julia service executes the computation.
4. The result is returned as a Python dictionary.
5. The WS connection is closed.

---

## Publish / Subscribe

This example shows a couple of subsribed nodes implemented in Python and a
Julia publisher.

```{mermaid}
flowchart LR
  C(("Publisher")) --> S1((Subscriber 1))
  C --> S2((Subscriber 2))
  style C fill:#009E73, color:white
  style S1 fill:#0072B2, color:white
  style S2 fill:#0072B2, color:white
```

### Subscribers (Python)

```python
import rembus as rb

def mytopic(data):
    print(f"[Sub-1] mytopic: {data}")

sub = rembus.node(name="sub-1")
sub.subscribe(mytopic)

print("up and running")
sub.wait()
```

```python
import rembus

def mytopic(data):
    print(f"[Sub-2] mytopic: {data}")

sub = rembus.node(name="sub-2")
sub.subscribe(mytopic)

print("up and running")
sub.wait()
```

### Publisher (Julia)

```julia
using Rembus

pub = component([
  "ws://:3001/client",
  "ws://:3002/client"
])

sensor1 = Dict("T" => 18.3, "H" => 45.2)
sensor2 = Dict("P" => 2.3)

publish(
    pub,
    "mytopic",
    Dict(
        "sensor#1" => sensor1,
        "sensor#2" => sensor2,
    )
)

close(pub)
```