Simple PHP LLM with SQLite and Token Probability Matching

A lightweight PHP demo of a rule-based language model using SQLite and probabilistic token scoring. Learn how to simulate AI-style responses with native PHP.

PHP Code Editor

<?php
// PHP LLM with SQLite + probabilistic scoring based on token frequency
$db = new PDO("sqlite::memory:");
$db->setAttribute(PDO::ATTR_ERRMODE, PDO::ERRMODE_EXCEPTION);
$db->exec("CREATE TABLE responses (id INTEGER PRIMARY KEY, text TEXT)");
$db->exec("CREATE TABLE tokens (
response_id INTEGER,
token TEXT,
count INTEGER DEFAULT 1,
UNIQUE(response_id, token)
)");
function tokenize(string $text): array {
return array_filter(preg_split('/[^a-z0-9]+/i', strtolower($text)));
}
// Learning phase: ingest responses and index tokens
function learn(PDO $db, string $text) {
$stmt = $db->prepare("INSERT INTO responses (text) VALUES (?)");
$stmt->execute([$text]);
$id = $db->lastInsertId();
 
הההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההההה
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
PHP Version

Execution Result

Ready to execute

Click the "Run Script" button to see the output here

Description

PHP LLM with SQLite Token Scoring

This example demonstrates how to build a lightweight rule-based language model in pure PHP, using:

  • SQLite as a local memory store
  • Token frequency matching to simulate probabilistic scoring
  • Natural language prompts to return the most relevant response

It’s a fully self-contained script, ideal for demos, education, or sandbox experiments.

What It Does

  • Tokenizes training sentences (responses)
  • Stores token frequency per response in a SQLite table
  • On user input, calculates a match score based on overlapping tokens
  • Returns the most relevant response with a confidence score

Comments

Ctrl + Enter to send

No comments yet

Be the first to share your thoughts!