Roddyka Local AI Connector for Ollama

Connect WordPress 7.0 AI to your own Ollama server — no API key, no cloud, no cost.

WordPress 7.0 ships with a native AI integration layer. The problem? It only comes with connectors for cloud providers that require an API key. This plugin bridges that gap, wiring the new WordPress AI Client directly to your local (or self-hosted) Ollama instance.

Run Llama 3, Mistral, Gemma, Phi, Qwen — any model Ollama supports — and use it to power WordPress AI features without sending a single token to the cloud.

---

✨ Features

Feature	Details
Zero API key	Ollama uses its own local auth; no credentials needed
Dynamic model list	Auto-discovers every model installed in Ollama via /v1/models
OpenAI-compatible	Uses Ollama's /v1/chat/completions endpoint
wp-admin UI	Custom connector card under Settings → Connectors
Runtime diagnostics	Card shows connection status and available models live
Flexible base URL	Option → constant → env var → sensible default
WP AI preferred models	Injects your default model at the top of the preference list
GPL-2.0-or-later	Free as in freedom

---

Requirements

Requirement	Minimum
WordPress	7.0 (AI Client API introduced)
PHP	7.4
Ollama	Any recent release with OpenAI-compatible endpoint

WordPress 7.0 is required. This plugin depends on WordPress\AiClient, the wp_connectors_init hook, and @wordpress/connectors — all introduced in WP 7.0.

---

Installation

From this repository (recommended for testing)

# Clone directly into your wp-content/plugins directory
cd /path/to/wordpress/wp-content/plugins
git clone https://github.com/roddyka/api-provider-for-ollama.git roddyka-local-ai-connector-for-ollama

Then activate it from Plugins → Installed Plugins in wp-admin.

Manual upload

1. Download the .zip from Releases
2. In wp-admin go to Plugins → Add New → Upload Plugin
3. Choose the zip, install and activate

---

Quick Start

1. Install Ollama → ollama.com/download
2. Pull a model (example):

   ollama pull llama3

3. Verify Ollama is running at http://127.0.0.1:11434
4. Activate this plugin in WordPress
5. Go to Settings → Connectors and open the Ollama Local card
6. The card will auto-detect your running models — pick a default and save

That's it. WordPress AI features (Block Editor AI, AI assistants, etc.) will now use your local Ollama instance.

---

Configuration

The base URL is resolved in this priority order:

Source	Example
wp-admin Connector card	Saved to wp_options
PHP constant	define('OLLAMA_API_BASE_URL', 'http://my-server:11434');
Environment variable	OLLAMA_API_BASE_URL=http://my-server:11434
Built-in default	http://127.0.0.1:11434/v1

The plugin automatically appends /v1 if you omit it.

Remote Ollama

If Ollama runs on a different machine or Docker container, just set the base URL to that host. Make sure the Ollama port (default 11434) is accessible from your WordPress server.

---

Architecture

roddyka-local-ai-connector-for-ollama.php  ← entry point, hooks, REST endpoint, settings
src/
  autoload.php           ← lightweight PSR-4 autoloader (no Composer needed)
  Provider/
    OllamaProvider.php   ← extends AbstractApiProvider; registers with WP AI Client
  Models/
    OllamaTextGenerationModel.php  ← OpenAI-compatible chat completions
  Metadata/
    OllamaModelMetadataDirectory.php  ← dynamic model discovery via /v1/models
  Support/
    OllamaConfig.php     ← centralised URL / model config with priority chain
assets/
  connector-card.js      ← React UI for the wp-admin Connectors screen

How the WP 7.0 AI integration works

WordPress AI Client (core)
       │
       ├─ asks registry for available providers
       │
       └─ OllamaProvider (this plugin)
              │
              ├─ OllamaModelMetadataDirectory  →  GET /v1/models
              └─ OllamaTextGenerationModel     →  POST /v1/chat/completions

---

Contributing

Contributions are welcome and encouraged! Here's how to get started:

1. Fork & clone

git clone https://github.com/roddyka/api-provider-for-ollama.git
cd api-provider-for-ollama

2. Set up your environment

• PHP 7.4+ with php in your PATH
• A local WordPress 7.0 install (e.g. wp-env, LocalWP, or plain LAMP/MAMP)
• Ollama running locally

3. Install into WordPress

Symlink or copy the plugin folder into wp-content/plugins/roddyka-local-ai-connector-for-ollama/ and activate.

4. Make your changes

The codebase is intentionally small and dependency-free (no Composer). Key extension points:

Want to…	Look at…
Add image/multimodal support	OllamaModelMetadataDirectory — extend $textCapabilities
Support streaming responses	OllamaTextGenerationModel — override request handling
Add vision/embedding models	New model class extending the base
Improve the admin UI	assets/connector-card.js (plain React via wp.element)
Add more config options	OllamaConfig.php + roddyka-local-ai-connector-for-ollama.php REST callback

5. Open a Pull Request

• Keep PRs focused on one thing
• Describe what and why in the PR description
• If it's a new feature, open an issue first to discuss

Reporting bugs

Open an issue with:

• WordPress version
• PHP version
• Ollama version & model(s) in use
• Steps to reproduce

---

Roadmap

• Streaming response support
• Image / multimodal model support (LLaVA, Moondream)
• Embedding model support
• WordPress.org plugin directory submission
• Unit test suite (PHPUnit)
• GitHub Actions CI

---

FAQ

Q: Does this work with Ollama on Docker / a remote server?
A: Yes. Set the base URL in the connector card or via the OLLAMA_API_BASE_URL constant/env var.

Q: Does it conflict with other AI providers (OpenAI, Anthropic)?
A: No. It registers as a separate provider (ollama-local) alongside any other connectors.

Q: What models are supported?
A: Any text generation model you have pulled in Ollama. The plugin discovers them automatically.

Q: I see "unreachable" in the card — what do I do?
A: Make sure ollama serve is running and accessible from the WordPress server. Try curl http://127.0.0.1:11434/v1/models.

---

License

GPL-2.0-or-later — free and open source.

Example:

define('OLLAMA_API_BASE_URL', 'http://127.0.0.1:11434/v1');

Usage

use WordPress\AiClient\AiClient;

$result = AiClient::prompt('Say hello in Portuguese.')
    ->usingProvider('ollama-local')
    ->generateTextResult();

echo $result->toText();

Notes

• This first version focuses on text generation.
• The provider relies on Ollama's OpenAI-compatible APIs (/v1/models and /v1/chat/completions).
• The Connectors screen lists this provider as keyless (authentication: none), which is the normal setup for local Ollama.
• The current WordPress Connectors UI only supports built-in auth fields (api_key or none), so custom fields like base URL/model are not first-class form inputs there yet.