Rodrigo Antunes

This plugin provides Ollama integration for the PHP AI Client SDK using OpenAI-compatible endpoints.
Features:
<ul>
<li>Text generation via <code>/v1/chat/completions</code></li>
<li>Dynamic model listing via <code>/v1/models</code></li>
<li>Automatic provider registration</li>
<li>Keyless/local setup support</li>
</ul>
Requirements:
<ul>
<li>PHP 7.4 or higher</li>
<li>Ollama server available</li>
<li>WordPress AI Client plugin/runtime support</li>
</ul>

Roddyka Local AI Connector for Ollama

Roddyka Local AI Connector for Ollama for the PHP AI Client SDK.

Per Søderlind

This plugin registers exo as an AI provider in WordPress 7’s AI Client SDK and Connectors page.
<a href="https://github.com/exo-explore/exo" rel="nofollow ugc">exo</a> connects all your devices into an AI cluster, enabling you to run frontier models locally. It exposes an OpenAI-compatible API that this plugin connects to.
Features:
<ul>
<li>Registers exo as a WordPress AI provider</li>
<li>OpenAI-compatible text generation via exo’s chat completions API</li>
<li>Auto-detect active models from your running exo cluster</li>
<li>Capability detection — displays model capabilities (Text, Code, Thinking, Vision) as badges</li>
<li>“Connect & Detect” / “Save & Re-detect” connector flow</li>
<li>Optional API key authentication with secure storage</li>
<li>Configurable endpoint URL (default: <code>http://localhost:52415</code>)</li>
<li>Settings integrated into WordPress 7’s Connectors page</li>
</ul>
Choosing a Model:
exo exposes every model in its catalog, but only models actively loaded on your cluster will respond. Use “Connect & Detect” on the Connectors page to discover which models are running.
Recommended — Instruct models produce clean, usable output for WordPress AI features (title generation, content suggestions, etc.):
<ul>
<li><code>Llama-3.2-3B-Instruct-8bit</code> — ~3 GB, fast, great for short tasks</li>
<li><code>Meta-Llama-3.1-8B-Instruct-4bit</code> — ~4 GB, good balance of speed and quality</li>
<li><code>Llama-3.3-70B-Instruct-4bit</code> — ~35 GB, best quality, needs a larger cluster</li>
</ul>
Avoid — Reasoning/thinking models (Qwen3.5, DeepSeek, GLM, Nemotron-Nano) spend most tokens on internal chain-of-thought, producing slow responses with minimal visible output.
To load a model: <code>exo run mlx-community/Llama-3.2-3B-Instruct-8bit</code>
Requirements:
<ul>
<li>WordPress 7.0 or later</li>
<li>PHP 8.3 or later</li>
<li>A running exo cluster (see <a href="https://github.com/exo-explore/exo" rel="nofollow ugc">exo documentation</a>)</li>
</ul>
<h3>Configuration</h3>
The plugin can be configured via the Connectors page or environment variables:
<ul>
<li><code>AIPRFOEX_ENDPOINT</code> — exo API endpoint (default: <code>http://localhost:52415</code>)</li>
<li><code>AIPRFOEX_API_KEY</code> — Optional API key for authentication</li>
<li><code>AIPRFOEX_MODEL</code> — Model name to use (auto-detected if empty)</li>
</ul>
You can also define these as constants in <code>wp-config.php</code>.

AI Provider for exo

Connect WordPress to exo — run frontier AI models locally on your device cluster.

Roddyka Local AI Connector for Ollama vs AI Provider for exo