FAIR

Federated and Independent Repositories

AI Provider for OpenAI Compatible Servers

Plugin Banner

AI Provider for OpenAI Compatible Servers

by George Stephanis

Download
Description

This plugin registers a provider for the WordPress AI Client that talks to any server implementing the OpenAI Chat Completions API — including Ollama, LM Studio, vLLM, llama.cpp, and LocalAI.

Run AI features in WordPress against models on your own hardware. No third-party AI service, no per-token billing, and no content leaving your network.

Features

  • Connect to any OpenAI-compatible base URL, local or remote (defaults to Ollama’s http://localhost:11434/v1).
  • Autodetect available models from the server’s /models endpoint, or specify models manually.
  • Setup card on the Connectors settings screen with a one-click connection test.
  • Best-effort detection of the running backend (Ollama, vLLM, or LM Studio), swapping in the brand’s name and icon on the connector card once detected.
  • Optional API key (bearer token) for servers that require authentication.
  • Custom HTTP headers for reverse proxies and gateways.
  • Advanced options for reasoning models: disable thinking blocks, and an R1-style message format that folds the system prompt into the first user message (needed for DeepSeek-R1).
  • Context length override and a vision/multimodal support toggle.

Requirements

  • The WordPress AI Client must be available (WordPress 7.0+).
  • A running OpenAI-compatible inference server reachable from your web server.

Developer Hooks

This plugin provides several filters for other plugins to dynamically extend its functionality:

  • connectors_ai_openai_compatible_servers_request_data: Filters the completions request body payload before it is sent to the endpoint. Passed arguments: $data (array), $modelId (string).
  • connectors_ai_openai_compatible_servers_request_headers: Filters HTTP headers for all request pathways. Passed arguments: $headers (array), $path (string), $method (string).
  • connectors_ai_openai_compatible_servers_detect_provider: Adds or overrides detection checks when probing server origins. Passed arguments: $detected_type (null|string), $base_url (string), $headers (array), $probe_args (array).
  • connectors_ai_openai_compatible_servers_model_options: Modifies the SupportedOption array for a resolved model. Passed arguments: $options (array), $modelId (string), $mode (string: ‘manual’ or ‘autodetect’).
  • connectors_ai_openai_compatible_servers_model_capabilities: Modifies the CapabilityEnum array for a resolved model. Passed arguments: $capabilities (array), $modelId (string), $mode (string: ‘manual’ or ‘autodetect’).

Credits

The Ollama, vLLM, and LM Studio brand icons shown when a server is detected are from
LobeHub’s lobe-icons, used under the MIT license.

  1. Install and activate the plugin.
  2. Go to Settings Connectors and locate “OpenAI Compatible”.
  3. Click “Set up”, enter your server’s base URL (including /v1), and an API key if needed.
  4. Click “Test Credentials” to verify the connection and detect models, then Save.
Which servers are supported?

Anything that implements the OpenAI-compatible /v1/models and /v1/chat/completions endpoints: Ollama, LM Studio, vLLM, llama.cpp server, LocalAI, and many others.

Do I need an API key?

Only if your server requires one. Local servers like Ollama usually don’t; leave the field empty or use a placeholder value.

Can my server run on a different machine?

Yes. Enter any base URL reachable from your WordPress host. The plugin allows requests to the configured base URL even when it resolves to a local/private address.

Does this send my content to a third party?

No. Requests go only to the server base URL you configure.

1.0.0

  • Initial release.
Back to top