PHPackages jarir-ahmed/php-llm - PHPackages - PHPackages [Skip to content](#main-content)[PHPackages](/)[Directory](/)[Categories](/categories)[Trending](/trending)[Leaderboard](/leaderboard)[Changelog](/changelog)[Analyze](/analyze)[Collections](/collections)[Log in](/login)[Sign up](/register) 1. [Directory](/) 2. / 3. [Utility & Helpers](/categories/utility) 4. / 5. jarir-ahmed/php-llm ActiveLibrary[Utility & Helpers](/categories/utility) jarir-ahmed/php-llm =================== Framework-agnostic unified AI/LLM toolkit for PHP — LLMs, embeddings, vector databases, RAG, agents, memory, cost tracking and structured output. Merges and extends manik/neuro + manik/cortex with zero framework coupling. v1.0.0(yesterday)00MITPHPPHP ^8.3 Since Jun 8Pushed yesterdayCompare [ Source](https://github.com/jarir2020/jarir-ahmed-php-llm)[ Packagist](https://packagist.org/packages/jarir-ahmed/php-llm)[ RSS](/packages/jarir-ahmed-php-llm/feed)WikiDiscussions main Synced yesterday READMEChangelogDependencies (3)Versions (2)Used By (0) jarir-ahmed/php-llm =================== [](#jarir-ahmedphp-llm) A **framework-agnostic** unified AI/LLM toolkit for PHP — one fluent API over many LLM, embedding and vector-database providers, plus RAG, agents, memory, **cost/token tracking** and **structured output**. > This package merges and extends [`manik/neuro`](https://github.com/dev-manik-mia/neuro) and [`manik/cortex`](https://github.com/dev-manik-mia/cortex) (which are the same Laravel package under two names) into a single library with **zero framework coupling** — no `laravel/framework` dependency. Works in Laravel, Symfony, WordPress, Slim, or plain PHP. [![PHP](https://camo.githubusercontent.com/c8d8dad6beb757a2b8acba331d16140813699543b88a37af0a81f20bd35f61de/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f5048502d382e332532422d626c7565)](https://php.net) [![License](https://camo.githubusercontent.com/f8df3091bbe1149f398a5369b2c39e896766f9f6efba3477c63e9b4aa940ef14/68747470733a2f2f696d672e736869656c64732e696f2f62616467652f6c6963656e73652d4d49542d677265656e)](LICENSE) --- What changed vs. neuro/cortex ----------------------------- [](#what-changed-vs-neurocortex) Areaneuro / cortexphp-llmFrameworkRequires `laravel/framework`**None** — pure PHP + GuzzleBootstrapServiceProvider + Facade`Client::create()` factoryHTTP`Illuminate\Support\Facades\Http`Guzzle-backed `Support\Http` (same fluent API) + retry + true SSE streamingConfigLaravel `config()``Support\Config` (dot-access, env fallback)DB (memory/pgvector)`Illuminate\...\DB`raw `PDO` with **self-healing** tablesSession memoryLaravel session`SessionStore` ($\_SESSION / in-memory / pluggable)Cost trackingflags only**`Pricing` + `Usage`**: real USD estimates per callStructured output—**JSON-schema `->structured()`** (native or instruct+parse)Providers7 LLMs**+ DeepSeek, Groq, Azure OpenAI, OpenRouter**Fluent clientshared, stateful builder**fresh `Generation` per call** + `ask()`, `conversation()`Install ------- [](#install) ``` composer require jarir-ahmed/php-llm ``` Requires PHP 8.3+. `ext-pdo` is only needed for the persistent-memory and pgvector drivers. Quick start ----------- [](#quick-start) ``` use JarirAhmed\PhpLlm\Client; $ai = Client::create([ 'defaults' => ['llm' => 'openai'], 'llm' => ['openai' => ['api_key' => 'sk-...']], ]); // one-shot echo $ai->ask('Explain Laravel in one line.'); // fluent — a fresh, isolated request each time $res = $ai->chat() ->provider('openai')->model('gpt-4o') ->system('You are terse.') ->message('Capital of France?') ->chat(); echo $res['content']; // "Paris." echo $res['usage']['total_tokens']; // 27 echo '$' . $res['cost']; // estimated USD cost ``` Any value you don't override falls back to the packaged defaults, which read from environment variables (e.g. `OPENAI_API_KEY`). So in production you can often just do `Client::create()`. LLM providers ------------- [](#llm-providers) ProviderKeyNotesOpenAI`openai`Anthropic`anthropic`Google Gemini`gemini`Ollama`ollama`local, no keyxAI Grok`grok`Mistral`mistral`Cohere`cohere`**DeepSeek**`deepseek`new**Groq**`groq`new**Azure OpenAI**`azure`new — set `base_url` (endpoint) + `deployment`**OpenRouter**`openrouter`new — optional `referer`/`title` headers``` $ai = Client::create([ 'defaults' => ['llm' => 'groq'], 'llm' => ['groq' => ['api_key' => getenv('GROQ_API_KEY')]], ]); echo $ai->ask('Why is the LPU fast?'); ``` Cost & token tracking ------------------------- [](#cost--token-tracking) Every `chat()` result carries normalized `usage` and an estimated `cost` (USD). Prices live in `Pricing` and are overridable: ``` use JarirAhmed\PhpLlm\Pricing\Pricing; use JarirAhmed\PhpLlm\Support\Usage; Pricing::set('my-finetune', 5.00, 15.00); // $/1M input, $/1M output $usd = Pricing::estimate('gpt-4o', new Usage(1_000_000, 1_000_000)); // 12.5 $breakdown = Pricing::breakdown('gpt-4o', new Usage(1000, 500)); ``` Hook every call for logging/metering: ``` use JarirAhmed\PhpLlm\Support\EventDispatcher; use JarirAhmed\PhpLlm\Events\MessageReceived; EventDispatcher::listen(MessageReceived::class, function (MessageReceived $e) { error_log("{$e->provider} {$e->model} took {$e->latency}s, cost \${$e->response['cost']}"); }); ``` Structured output (JSON) ------------------------ [](#structured-output-json) ``` $person = $ai->chat() ->message('Extract: "Ada Lovelace, born 1815, mathematician."') ->structured([ 'type' => 'object', 'properties' => [ 'name' => ['type' => 'string'], 'born' => ['type' => 'integer'], 'role' => ['type' => 'string'], ], 'required' => ['name', 'born'], ]); // ['name' => 'Ada Lovelace', 'born' => 1815, 'role' => 'mathematician'] ``` Uses native `response_format` JSON-schema on OpenAI-compatible providers, and an instruction + tolerant parser everywhere else. Streaming --------- [](#streaming) ``` foreach ($ai->chat()->message('Write a haiku')->stream() as $chunk) { echo $chunk['content']; flush(); } ``` OpenAI-family drivers stream true Server-Sent Events token-by-token via the Guzzle stream body. Conversations with memory ------------------------- [](#conversations-with-memory) ``` $chat = $ai->conversation('user-42', 'openai', 'conversation'); echo $chat->say('My name is Sam.'); echo $chat->say('What is my name?'); // remembers "Sam" echo $chat->totalCost(); ``` Memory drivers: `session` (web), `conversation` (in-process), `persistent` (PDO, self-creating table). ``` use JarirAhmed\PhpLlm\Client; Client::useDatabase('default', new PDO('sqlite:' . __DIR__ . '/ai.sqlite')); $ai->memory()->driver('persistent')->add('user-42', ['role' => 'user', 'content' => 'hi']); ``` Embeddings, vectors & RAG ----------------------------- [](#embeddings-vectors--rag) ``` $vec = $ai->embed('text to embed'); // ['embedding'=>[...], 'dimensions'=>1536] $ai->vector('qdrant')->createCollection('docs', 1536); $ai->vector('qdrant')->upsert('docs', [ ['id' => '1', 'vector' => $vec['embedding'], 'payload' => ['text' => '...']], ]); // RAG $ai->rag()->ingestion()->ingestFromPath('handbook.md', 'docs'); $answer = $ai->rag()->collection('docs')->question('What is the refund policy?')->answer(); echo $answer['answer']; ``` Vector drivers: `qdrant`, `pinecone`, `pgvector` (PDO), `weaviate`, `milvus`, `chroma`. Agents & tool calling ------------------------- [](#agents--tool-calling) ``` $agent = $ai->agent('openai') ->tool('get_time', fn () => date('c'), 'Current time') ->maxSteps(5); $result = $agent->run('What time is it?'); echo $result['response']; ``` Using inside a framework ------------------------ [](#using-inside-a-framework) Nothing framework-specific is required. In Laravel/Symfony just build the client once in a service/container binding: ``` $this->app->singleton(\JarirAhmed\PhpLlm\AIClient::class, fn () => \JarirAhmed\PhpLlm\Client::create([ 'llm' => ['openai' => ['api_key' => config('services.openai.key')]], ])); ``` You can also plug your framework's session and DB in: `SessionStore::use($yourStore)` and `Client::useDatabase('default', $pdo)`. Testing without network ----------------------- [](#testing-without-network) ``` $ai = Client::create(['llm' => ['openai' => ['api_key' => 'test']]]); $ai->fake(); // all providers return deterministic fakes $this->assertSame('fake response', $ai->ask('anything')); ``` For driver-level tests, inject a Guzzle mock: `PendingRequest::useClient($guzzleWithMockHandler)`. Architecture ------------ [](#architecture) ``` Client::create() ── AIClient (hub) ├─ chat()/generate() ─ Generation (fluent, per-request: usage+cost+structured) ├─ conversation() ─ Conversation (memory-backed multi-turn) └─ managers ─ LLM / Embedding / Vector / Image / Speech / RAG / Memory / Agent └─ Drivers (Support\Http → Guzzle) Support: Config · Http/PendingRequest/Response · Database(PDO) · SessionStore · EventDispatcher · Env · Usage Pricing: USD cost tables + estimate/breakdown ``` Credits ------- [](#credits) Built on the excellent work of **Manik Mia** (`manik/neuro`, `manik/cortex`). This package re-architects that code to be framework-agnostic and adds cost tracking, structured output, extra providers, and a polished fluent client. License ------- [](#license) MIT. ### Health Score 40 — FairBetter than 86% of packages Maintenance100 Actively maintained with recent releases Popularity0 Limited adoption so far Community2 Small or concentrated contributor base Maturity48 Maturing project, gaining track record How is this calculated?**Maintenance (25%)** — Last commit recency, latest release date, and issue-to-star ratio. Uses a 2-year decay window. **Popularity (30%)** — Total and monthly downloads, GitHub stars, and forks. Logarithmic scaling prevents top-heavy scores. **Community (15%)** — Contributors, dependents, forks, watchers, and maintainers. Measures real ecosystem engagement. **Maturity (30%)** — Project age, version count, PHP version support, and release stability. ### Release Activity Cadence Unknown Total 1 Last Release 1d ago ### Community Maintainers ![](https://www.gravatar.com/avatar/20700f34ff813055154e843bbdbe04d33320c1e6a81bbb3301cad67cb8350fd3?d=identicon)[jarircse16](/maintainers/jarircse16) --- Tags aivectoropenaiagentsGeminillmanthropicdeepseekembeddingsraggroq ### Code Quality TestsPHPUnit ### Embed Badge ![Health badge](/badges/jarir-ahmed-php-llm/health.svg) ``` [![Health](https://phpackages.com/badges/jarir-ahmed-php-llm/health.svg)](https://phpackages.com/packages/jarir-ahmed-php-llm) ``` ### Alternatives [aws/aws-sdk-php AWS SDK for PHP - Use Amazon Web Services in your PHP project 6.3k532.1M2.5k](/packages/aws-aws-sdk-php)[cognesy/instructor-php The complete AI toolkit for PHP: unified LLM API, structured outputs, agents, and coding agent control 317117.1k1](/packages/cognesy-instructor-php)[neuron-core/neuron-ai The PHP Agentic Framework. 1.9k496.1k32](/packages/neuron-core-neuron-ai)[vizra/vizra-adk Vizra Agent Development Kit - A comprehensive Laravel package for building intelligent AI agents. 29431.7k](/packages/vizra-vizra-adk)[soukicz/llm LLM client with support for cache, tools and async requests 4411.4k](/packages/soukicz-llm) PHPackages © 2026 [Directory](/)[Categories](/categories)[Trending](/trending)[Changelog](/changelog)[Analyze](/analyze)