README

PHP Client Library to interact with Ollama API.

The intention is to work with models on your Ollama setup and not the create a delete models. Therefore, this library will not implement any APIs to create, move or delete models. Here is a list of the APIs we intend to implement and the status of the implementation:

Completion (without streaming support)
Chat Completion (without streaming support)
List local models
Show Model Information
List Running Models
Version

The checkout the Ollama API Docs for more information and the APIs we might miss.

This package contains some low-level API libraries as well as a convenient API wrapper for all APIs.

Installation

composer install tredmann/php-ollama

Convenience Wrapper

The easiest way to ask the LLM things is to use the convenience wrapper:

use Ollama\Ollama;

$ollama = new Ollama(model: 'gemma2:latest');

echo $ollama->completion(prompt: 'What is the capitol of Germany?');
// The capital of Germany is **Berlin**.

It does have a ton of limitations, but for quick results it is easy to use. I would highly encourage to look into the low-level library.

General way of working with the low-level library

Creating the Client

use Ollama\Client\OllamaClient;

$client = new OllamaClient(
    baseUrl: 'http://localhost:11434' // default
    );

Inject the client into the respected API

use Ollama\Api\Completion;

$completionApi = new Completion(client: $client);

Use the API by creating an API request

use Ollama\Requests\CompletionRequest;

$request = new CompletionRequest(
    model: 'phi3.5:latest',
    prompt: 'What is the capitol of Germany?' 
);

Use a request to query the API

$response = $completionApi->getCompletion(request: $request);

Use the response

echo $response->response;
// 'The capitol of Germany is Berlin.'

tredmann / php-ollama

Maintainers

Details