ollama run qwen2.5
This code by Erick Engelke
November 26, 2025
Writing a simple AI program using PHP is pretty easy. All the parts you need are free, and they are available for Mac, Windows or Linux, your choice.
This will work on almost any computer, but if you have a GPU (all M1-M4 Macs do), it will do the AI much faster than if you do not.
If you are lucky, someone in your organization may allow you to use their pre-configured server running Ollama and you may have PHP already installed. If so, you can skip to the Trival Program section and just replace localhost with the dns name of the server.
Visit the web site ollama.com and follow the download instructions to install it.
Using a shell window, run the command to download and run a specific model, eg. qwen2.5.
ollama run qwen2.5Then enter a prompt:
>>> write a letter to eric the red.to which it will write a letter. Okay, once that works, we can install a language to call it programatically.
Use apt install php php-libcurl
Install homebrew. Then use the commands brew install to install php and libcurl.
You may need to edit php.ini.
Use an LLM to find instructions.
I made a single function library which does all the communcations between your application and Ollama/LLM.
It’s in the file ai_simple.php.
It exposes a function called ask_ai() which takes two parameters, $text and $model.
$text will contain the text of your query, and $model will contain the name of the model. It must be a loaded model, you cannot specify a new model not yet loaded in ollama.
Here is a trivial real program which writes a letter to a specified user, namely Eric the Red. You write it in a text editor, name it ai_test.php and from the shell run: php ai_test.php
<?php
// include the ask_ai function
include "ai_simple.php";
// the name of the model we will use
$model = "qwen2.5";
$aihost = "localhost"; // your computer, or a computer with a GPU
// define a name:
$name = "Eric the Red";
// define a question or request
$text = "Write a letter congraduatling $name on his journey.";
$results = ask_ai( $text, $model );
print $results;This one loads a file with data and writes the data back to disk.
<?php
// include the ask_ai function
include "ai_simple.php";
// the name of the model we will use
$model = "qwen2.5";
$aihost = "localhost"; // your computer, or a computer with a GPU
$input = "ai_fred.php"; // some input file
$output = "myresults.txt"; // some output file
// get some data
$context = file_get_contents( $input );
// ask a quesiton about that data or give commands
$question = "What does the following do: $context";
$results = ask_ai( $question, $model );
print "Writing results to $output\n";
file_put_contents( $output , $results );