README

PHP library for accessing the fuzzy.ai API.

Requirements

PHP 5.3.3 or later with the cURL extension.

You can install the library via Composer:

composer require fuzzy-ai/sdk

To load the library, use Composer's autoload:

require_once('vendor/autoload.php');

All API calls require an API key from https://fuzzy.ai/

$client = new FuzzyAi\Client('YourAPIKey');

list($result, $evalID) = $client->evaluate('YourAgentID', array('input1' => 42));

This is the main class and serves as an entry point to the API.

FuzzyAi\Client($key, $root) Constructor that takes the following arguments and returns a new Client object.
- key: Your Fuzzy.ai API key
- root (optional): The API endpoint (defaults to https://api.fuzzy.ai)
evaluate($agentId, $inputs) The main method to use, it performs a single inference. Returns an array of outputs and an evaluation ID (for training feedback, see below).
- agentId: The ID of the Agent to perform the evaluation against
- inputs: An associative array of input name => values.
feedback($evaluationId, $performance) This is the method used for training better results. Returns a feedback object.
- evaluationId: Unique identifier returned from an evaluate() call.
- performance: The performance metrics (as an associative array) to provide the learning.
newAgent($props) Use this method to create a new Agent. Returns an Agent object.
- props: An associative array representing an agent with at least inputs, outputs, and rules.
getAgent($agentId) This will fetch an existing agent definition. Returns an Agent object.
- agentId : ID of the agent to retrieve

This class represents an Agent and provides full CRUD features.

FuzzyAi\Agent($client) Constructor - takes an HTTP Client object, but it's easier to use either newAgent or getAgent from above to create the Agent object.
evaluate($inputs) Like Client::evaluate, but on an existing Agent instance. NOTE Returns an Evaluation object.
- inputs: An associative array of input name => values.
create($props) Creates a new agent (although Client::newAgent is likely easier).
- props: An associative array representing an agent with at least inputs, outputs, and rules.
read($props) Reads an agent definition from the API. Probably easier to use Client::getAgent().
- id : Agent ID to read.
update($props) Updates the current agent instance with $props.
- props: New agent properties.
delete() Deletes current agent from the API.

This class represents a single evaluation.

read($id) Load a single evaluation object by ID.
feedback($values) Provide learning feedback data to a single evaluation . Returns a Feedback object.
- values: the performance metrics to provide for feedback.

The examples/ directory has some examples of using the library.

Install dependencies:

composer install

Run the tests:

./vendor/bin/phpunit