spineapi 0.0.1

A simple machine learning inference library.

Spine API

Access your ML models through HTTP requests.

It is simple, clean, and easy to use. It works with jupyter notebooks too.

POST http://your_server_ip/api/your_project/your_function => your_result

### Installation
```bash
$ pip instal spine-api

Usage

1. Import the library

from spine import Connection

2. Define your inference function

def hello_function (input):
  # ...
  # do something with input
  # ...
  return output

3. Specify name and description to initialize your project

spine_connection = Connection(
  project_name="hello_project",
  description="An inference API for my ML model",
  author="", # Optional
  link="" # Optional
)

4. Register your function(s)

spine_connection.register_function(
  pathname='hello_function',
  function=hello_function,
  # ============ Optional ==================
  # Set True if you want to protect this API
  requiresAuth=False,
  authToken="xxx",
  # ========================================
)

The function will be accessible through /api/hello_project/hello_function

5. Run

spine_connection.run()

That's it! You can now communicate with your ML model through HTTP post requests.

6. Send requests

Note You have to first run JSON.stringify(input) for the request data.

const data = {
  input: JSON.stringify(YOUR_INPUT)
  // ============ Optional ==================
  // Required if the API is protected
  authToken: "xxx",
  // ========================================
}

Post request body to the endpoint.

const url = 'https://spineapi.com/api/hello_function';

fetch(
  url,
  {
    method: 'POST',
    body: JSON.stringify(data),
    headers: new Headers({ 'Content-Type': 'application/json' })
  }
)
.then(res => res.json())
.catch(error => console.error('Error:', error))
.then(response => console.log('Success:', response));

Self hosted server

Prerequisites

1. Make sure your server is accessible through port 3000
2. Have Node.js installed on your server

Installation

$ git clone https://github.com/spineapi/spine-api
$ cd spine-api/server
$ npm install

Run server

$ node server.js

Now you can access your server through http://YOUR_SERVER_IP:3000. Copy the passcode.

Specify url and passcode For self hosted server, you'll have to specify url and passcode.

from spine import Connection
spine_connection = Connection(
  name="hello_project",
  description="a demo app for my ML model",
  url="http://YOUR_SERVER_IP:3000",
  passcode="YOUR_PASSCODE"
)

For other steps, it's all the same.

License

MIT