Aareguru SDK

Real-time water temperature, weather, and flow data for the Aare River in Switzerland.

This is an unofficial SDK for the Aare.guru public API, generated by Voxgig with @voxgig/sdkgen. It is not affiliated with, endorsed by, or sponsored by the upstream API provider.

Learn more about Voxgig SDKs at voxgig.com/sdk.

Metadata kindly supplied by www.freepublicapis.com.

TypeScript, Python, PHP, Golang, Ruby, Lua SDKs, a CLI, an interactive REPL, and an MCP server for AI agents — all generated from one OpenAPI spec by @voxgig/sdkgen.

About Aare.guru API

Aare.guru is a community-run service that publishes live measurements for the Aare — Switzerland’s longest river entirely within the country, and the river many Bernese swim in on a summer afternoon. The API answers the practical question “how warm is the Aare today, and is it safe to swim?” and exposes the raw data behind it.

What you get from the API

• Current water temperature (°C, rounded and unrounded variants)
• Flow (m³/s) and a 1-5 danger level (flow_gefahrenstufe)
• River height (m above sea level)
• Air temperature, precipitation, sunshine and forecast weather symbols
• Plain-text descriptions (text, temperature_text) suitable for UIs
• Per-city queries for Bern, Olten, Thun, and other Aare locations

Data sources — BAFU (Swiss Federal Office of Environment) for national Aare measurements, TemperAare for Olten, MeteoSchweiz SwissMetNet for weather observations, and Meteotest for the forecast.

Operational notes — data updates every ~10 minutes (with 10-20 min delay); polling every 5 minutes is plenty. Responses are cached for 2 minutes. CORS is wide-open (*) and JSONP is supported via a callback query parameter. “Every value could be null at any time — defensive programming always pays off.”

Entities, not endpoints

This SDK exposes the API as a small set of semantic entities — Legacy, Stuff and V2018 — that you call directly, instead of assembling URL paths and query strings. Entities are Capitalised to mark them as the primary surface, each with the operations they support (load):

const client = new AareguruSDK()
const legacy = await client.Legacy().load()

Thinking in entities keeps the mental model small — for people and AI agents alike — rather than reasoning about raw HTTP routes and query parameters.

Offline unit testing

Every SDK ships a built-in test mode that swaps the HTTP transport for an in-memory mock, so your unit tests run fully offline — no server, no network, and no credentials:

TypeScript

const client = AareguruSDK.test()
const legacy = await client.Legacy().load()
// legacy is a bare Legacy populated with mock data
console.log(legacy)

Python

client = AareguruSDK.test()
legacy = client.Legacy().load()
print(legacy)

PHP

// Seed fixture data so offline calls resolve without a live server.
$client = AareguruSDK::test([
    "entity" => ["legacy" => ["test01" => []]],
]);
$legacy = $client->Legacy()->load();

Golang

client := sdk.Test()
result, err := client.Legacy(nil).Load(
    nil, nil,
)

Ruby

# Seed fixture data so offline calls resolve without a live server.
client = AareguruSDK.test({
  "entity" => { "legacy" => { "test01" => {} } },
})
legacy = client.Legacy.load()

Lua

local client = sdk.test()
local result, err = client:Legacy():load()

Packages

Quickstart

TypeScript

import { AareguruSDK } from '@voxgig-sdk/aareguru'

const client = new AareguruSDK()

// Load legacy data (returns a Legacy)
const legacy = await client.Legacy().load()
console.log(legacy)

See the TypeScript README for the full guide.

Surfaces

Use it from an AI agent (MCP)

The generated MCP server exposes every operation in this SDK as an MCP tool that Claude, Cursor or Cline can call directly. Build and register it:

cd go-mcp && go build -o aareguru-mcp .

Then add it to your agent’s MCP config (Claude Desktop, Cursor, etc.):

{
  "mcpServers": {
    "aareguru": {
      "command": "/abs/path/to/aareguru-mcp"
    }
  }
}

Entities

The API exposes 3 entities:

The operations available across these entities are load — see each entity’s own list above for exactly which it supports.

Quickstart in other languages

Python

from aareguru_sdk import AareguruSDK

client = AareguruSDK()


# Load a specific legacy (returns the record, raises on error)
legacy = client.Legacy().load()
print(legacy)

PHP

<?php
require_once 'aareguru_sdk.php';

$client = new AareguruSDK();


// Load a specific legacy (returns the bare record; throws on error)
$legacy = $client->Legacy()->load();
print_r($legacy);

Golang

import sdk "github.com/voxgig-sdk/aareguru-sdk/go"

client := sdk.New()

// Load legacy data
legacy, err := client.Legacy(nil).Load(nil, nil)
if err != nil {
    panic(err)
}
fmt.Println(legacy)

Ruby

require_relative "Aareguru_sdk"

client = AareguruSDK.new


# Load a specific legacy (returns the bare record; raises on error)
legacy = client.Legacy.load()
puts legacy

Lua

local sdk = require("aareguru_sdk")

local client = sdk.new()


-- Load a specific legacy
local legacy, err = client:Legacy():load()
print(legacy)

Direct and prepare

For endpoints the entity model doesn’t cover, use the low-level methods:

• direct(fetchargs) — build and send an HTTP request in one step.
• prepare(fetchargs) — build the request without sending it.

Both accept a map with path, method, params, query, headers, and body. See the How-to guides below.

How-to guides

Make a direct API call

When the entity interface does not cover an endpoint, use direct:

TypeScript:

const result = await client.direct({
  path: '/api/resource/{id}',
  method: 'GET',
  params: { id: 'example' },
})
if (result instanceof Error) {
  throw result
}
console.log(result.data)

Python:

result = client.direct({
    "path": "/api/resource/{id}",
    "method": "GET",
    "params": {"id": "example"},
})

PHP:

$result = $client->direct([
    "path" => "/api/resource/{id}",
    "method" => "GET",
    "params" => ["id" => "example"],
]);

Go:

result, err := client.Direct(map[string]any{
    "path":   "/api/resource/{id}",
    "method": "GET",
    "params": map[string]any{"id": "example"},
})
if err != nil {
    panic(err)
}
fmt.Println(result)

Ruby:

result = client.direct({
  "path" => "/api/resource/{id}",
  "method" => "GET",
  "params" => { "id" => "example" },
})

Lua:

local result, err = client:direct({
  path = "/api/resource/{id}",
  method = "GET",
  params = { id = "example" },
})

Advanced

Everyday use only needs the sections above. This explains the internals behind every call — relevant when writing custom features.

Every SDK call runs the same five-stage pipeline:

1. Point — resolve the API endpoint from the operation definition.
2. Spec — build the HTTP specification (URL, method, headers, body).
3. Request — send the HTTP request.
4. Response — receive and parse the response.
5. Result — extract the result data for the caller.

A feature hook fires at each stage (e.g. PrePoint, PreSpec, PreRequest), so features can inspect or modify the pipeline without forking the SDK.

Features

Pass custom features via the extend option at construction time.

Per-language documentation

• TypeScript
• Python
• PHP
• Golang
• Ruby
• Lua

Upstream API

This SDK is generated from the upstream OpenAPI specification. It is an unofficial client and is not affiliated with the API provider.

• Upstream API: https://aareguru.existenz.ch

Security

Please report security issues to security@voxgig.com. See SECURITY.md. Do not open public issues for suspected vulnerabilities.

Generated from the Aare.guru API OpenAPI spec by @voxgig/sdkgen.