subhashladumor1/larachain

LaraChain - LangChain-inspired AI orchestration framework for PHP and Laravel

1.0.1 2026-03-08 06:11 UTC

Requires

• php: ^8.0
• illuminate/contracts: ^11.20|^12.0
• illuminate/support: ^11.20|^12.0
• laravel/ai: ^0.1.0

Requires (Dev)

• mockery/mockery: ^1.6
• orchestra/testbench: ^9.0|^10.0
• pestphp/pest: ^3.0
• pestphp/pest-plugin-laravel: ^3.0

MIT

• laravel
• larachain
• laravel-ai
• ai-agents
• ai-orchestration
• intelligent-agents
• vector-search
• langchain-php
• llm-chains
• rag-framework
• ai-automation

README

LaraChain is a LangChain-inspired AI orchestration framework built specifically for Laravel 12. In 2026, building AI apps is no longer just about calling an API—it's about building Stateful Workflows, Agentic Tools, and Modern RAG pipelines. LaraChain provides the primitives to build these with professional-grade elegance and type safety.

🗺️ How LaraChain Works

LaraChain follows a "Runnable" architecture where every component—prompts, models, retrievers, and parsers—can be piped together.

graph LR
    A[Input Variables] --> B[PromptTemplate]
    B -->|Pipe| C[ChatModel]
    C -->|Pipe| D[OutputParser]
    D --> E[Final PHP Object]
    
    subgraph "The RAG Loop"
        F[PDF/Web/CSV] --> G[TextSplitter]
        G --> H[EmbeddingModel]
        H --> I[VectorStore]
        I -->|Retrieve| B
    end

Loading

🚀 Key Features

Feature	Description
LCEL-style Piping	Use the .pipe() pattern to chain components like a pro.
Smart Agents	ReAct (Reasoning + Acting) agents that use tools and make decisions.
Advanced RAG	Document loaders, recursive text splitting, and vector retrieval.
Postgres Support	Native pgvector integration for production-ready storage.
Memory Drivers	Stateful conversation buffers to maintain context.
Laravel 12 Native	Deeply integrated with the Laravel AI SDK and Service Container.

📂 Folder Structure

larachain/
├── src/
│   ├── Agents/           # ReAct and Agentic logic
│   ├── Chains/           # Pipeline orchestration (Sequential, Router)
│   ├── Contracts/        # Interfaces for all components
│   ├── DocumentLoaders/  # Reading PDF, Web, CSV contents
│   ├── Embeddings/       # Vector generation logic
│   ├── Laravel/          # Service Providers and Facades
│   ├── Memory/           # Conversation state management
│   ├── Messages/         # Message objects (User, Assistant, System)
│   ├── Models/           # AI Model wrappers (ChatModel)
│   ├── Parsers/          # Output formatting (JSON, XML)
│   ├── Prompts/          # Template management
│   ├── Retrieval/        # Document retrieval logic
│   ├── Support/          # Traits and Helpers (HasPipe)
│   ├── TextSplitters/    # Document chunking logic
│   ├── Toolkits/         # Grouped tools (File, Database)
│   ├── Tools/            # Individual tool implementations
│   └── VectorStores/     # Storage drivers (In-Memory, Postgres)

📖 Functional API Guide

1. The Pipe Pattern (Recommended)

The hallmark of LaraChain 2026 is the ability to chain components elegantly.

use LaraChain\Prompts\PromptTemplate;
use LaraChain\Models\ChatModel;
use LaraChain\Parsers\JsonParser;

$chain = PromptTemplate::make('Extract data from this text: {text} into JSON format.')
    ->pipe(new ChatModel('gpt-4o'))
    ->pipe(new JsonParser());

$output = $chain->invoke(['text' => 'My name is John and I live in London.']);
// Returns: ['name' => 'John', 'location' => 'London']

2. Intelligent Agents

An agent can use specialized tools to complete complex tasks.

use LaraChain\Agents\AgentExecutor;
use LaraChain\Toolkits\FileToolkit;

$agent = AgentExecutor::make()
    ->tools((new FileToolkit())->getTools());

$response = $agent->run("Read config.json and summarize it in readme.md");

3. RAG (Postgres + Recursive Chunking)

Handle large documents with state-of-the-art chunking and production storage.

use LaraChain\TextSplitters\RecursiveCharacterTextSplitter;
use LaraChain\VectorStores\PostgresVectorStore;
use LaraChain\Embeddings\EmbeddingModel;

$splitter = new RecursiveCharacterTextSplitter(chunkSize: 1000, chunkOverlap: 200);
$chunks = $splitter->splitText($largePdfContent);

$store = new PostgresVectorStore(new EmbeddingModel());
$store->addTexts($chunks);

⚖️ LaraChain vs. LangChain (For Laravel)

Feature	LangChain (Python/JS)	LaraChain (PHP/Laravel)
Syntax	Pipe Operator (|)	Fluent .pipe() Method
Integration	Ad-hoc	Native Service Providers / Facades
I/O	General	Laravel FileSystem / DB Facades
Agents	LangGraph	ReAct / Future LaraGraph
Models	Custom Drivers	Laravel AI SDK (Native)

📈 Use Cases

1. Semantic Document Search: Build a "Chat with your PDF" app in minutes using RecursiveSplitter and PostgresVectorStore.
2. Autonomous Code Auditor: Use the FileToolkit and AgentExecutor to scan your repository for security flaws.
3. Structured Data Extraction: Pipe raw OCR text through a ChatModel and JsonParser to ingest invoices into your database.

🛠️ Installation & Setup

composer require subhashladumor1/larachain
php artisan vendor:publish --tag="larachain-config"

Refer to LARACHAIN_VERIFICATION_2026.md for detailed verification of all 2026 market features.

🛠️ Multi-Provider Management

LaraChain uses a Driver-based Architecture (similar to Laravel's Database or Mail systems). You can configure and switch between providers at runtime.

1. Configuration (config/larachain.php)

Define multiple LLM, Vector, and Embedding providers:

'default' => [
    'llm' => 'openai',
    'vectorstore' => 'postgres',
],

'llms' => [
    'openai' => ['model' => 'gpt-4o'],
    'anthropic' => ['model' => 'claude-3-5-sonnet'],
],

2. Switching Providers at Runtime

Use the LaraChain facade to swap drivers dynamically:

// Use Anthropic instead of the default OpenAI
$model = LaraChain::model('anthropic');

// Use a specific vector store
$vectorStore = LaraChain::vectors()->driver('memory');

// Chain them together
$chain = PromptTemplate::make('Hello {name}')
    ->pipe($model)
    ->pipe(new JsonParser());

⚙️ Configuration

Contributions are welcome! Pull requests for new Vector Drivers (Pinecone, Qdrant) are prioritized.

📄 License

The MIT License (MIT). See License File.