README

Cohere PHP Client

A PHP client to interact with the Cohere® API, designed to be framework-agnostic, simple to use, and fully compatible with PSR-18.

This package provides an easy and structured way to use Cohere's powerful language models — for embeddings, chat, classification, tokenization, and more — in any PHP project.

Features 🛠️

✅ Support for Cohere API v1/v2 endpoints (chat, embed, classify, tokenize, etc.)
✅ Compatible with any PSR-18 HTTP Client
✅ File upload and multipart support
✅ Developer-friendly DTOs and response builders
✅ Easily extendable with your own endpoints

Requirements ⚙️

PHP 8.2 or higher
Composer
A PSR-18 compatible HTTP client ( e.g. Guzzle, Symfony HttpClient, HTTPlug clients)

Installation 📝

Install the library via Composer:

composer require talleu/cohere-php-client

Then install your preferred HTTP client

Using Symfony HttpClient:

composer require symfony/http-client

Using Guzzle:

composer require guzzlehttp/guzzle

Or another client implementing PSR18

Basic Usage 🎯

Minimal example

use Talleu\CohereClient\Cohere;

$client = Cohere::client('your-api-key'); 

// Call the embed endpoint
$embeds = $client->embed()->create([
      'Cohere is amazing!',
      'Let’s try embedding some text.'
  ]);

var_dump($embeds);

Or to simple chats with LLM

    $chat = $client->chat()->create([
        [
            'role' => 'user',
            'content' => 'how are you ?'
        ]
    ]);

Authentication 🔐

You can pass the API key directly in the http client:

Cohere::client('your-api-key');

Or use an environment variable (recommended):

#.env
COHERE_API_KEY=your-api-key

Available Endpoints 📚

The following endpoints are supported:

Endpoint	Class	Description
`v2/embed`	`Embed`	Generate embeddings from input text
`v2/chat`	`Chat`	Perform conversational chat with a LLM
`v1/classify`	`Classify`	Text classification based on custom labels
`v1/tokenize`	`Tokenize`	Token-level breakdown of input text
`v1/detokenize`	`Detokenize`	De-tokenify tokens to text
`v1/connectors`	`Connector`	Cohere connectors
`v1/embed-jobs`	`EmbedJob`	Async embed jobs
`v2/rerank`	`Rerank`	Retrieve Cohere available models
`v1/models`	`Model`	Produces an ordered array with text
`v1/datasets`	`Dataset`	Create a dataset by uploading a file, retrieve datasets
`v1/finetuning`	`FineTuning`	Trains, deploy, list datasets

You can access them via:

$client->embed();
$client->chat();
$client->tokenize();
$client->detokenize();
$client->classify();
$client->rerank();
$client->dataset();
$client->fineTuning();
$client->model();
$client->embedJob();
$client->connector();
// etc.

Then you can use it simple :

$connector = $client->connector()->create($name, $url, ['model' => 'command-a-03-2025']);
$connector = $client->connector()->get($id);
$connector = $client->connector()->list();

Each endpoint returns a strongly typed DTO with the result of the API call.

To send a request without using provided resources, you can use directly the CohereClient sendRequesst() method

$chat = Cohere::client()->sendRequest('GET', '/v2/chat', $body);

📚 Documentation by endpoints

Documentation 📚

🧠 Cohere official docs: https://docs.cohere.com
📘 API Reference: https://docs.cohere.com/reference
📦 This PHP client wraps the API endpoints in a friendly OO API with typed DTOs.

Contributing 🤝

PRs are welcome! If you’d like to add support for more endpoints, improve tests or add features, feel free to open an issue or submit a PR.

talleu / cohere-php-client

Maintainers

Details