This PHP framework enables developers to build customizable, production-ready LLM applications. It targets PHP developers looking to integrate LLM capabilities into their projects, offering a structured approach to creating agents, managing memory, and implementing complex patterns like RAG and tool usage.

How It Works

Neuron AI provides an extensible Agent class that abstracts LLM interactions, memory management, and tool execution. Developers define an AIProviderInterface to switch between LLM backends (Anthropic, Ollama, OpenAI, Gemini) and use a SystemPrompt class to construct detailed instructions for the LLM. Tools can be defined with callable functions and properties, or integrated via an MCP server connector, allowing agents to perform external tasks. For RAG, a separate RAG class is available, requiring configuration of embeddings and vector stores.

Quick Start & Requirements

• Install via Composer: composer require inspector-apm/neuron-ai
• Requires PHP ^8.1.
• Official documentation: https://inspector-apm.github.io/neuron-ai/

Highlighted Details

• Supports multiple LLM providers (Anthropic, Ollama, OpenAI, Gemini) with easy switching.
• Enables custom tool integration and external tool discovery via MCP server.
• Facilitates RAG implementation with support for various vector stores and embeddings providers.
• Offers structured output capabilities for data extraction and integration.

Maintenance & Community

• Community support via the Inspector Forum.
• Contribution guidelines are available.

Licensing & Compatibility

• MIT License.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The framework is presented as a development kit, and specific implementations for retrieving YouTube transcriptions or connecting to MCP servers are shown as examples requiring developer implementation.