WordPress Ollama Provider for Local AI (OpenAI-Compatible)

Connect your WordPress AI plugin to Ollama in minutes and run a private, self-hosted AI workflow with your own local LLM models.

This plugin is a production-minded AI provider for WordPress that uses Ollama's OpenAI-compatible API while keeping prompts and responses in your own environment.

Keywords: WordPress Ollama, Local AI, Self-hosted AI, OpenAI compatible, Offline AI, Private AI, Ollama provider.

What This Is

AI Provider for Local Ollama registers ollama_local as a provider in the WordPress Connectors ecosystem so your site can use local models through Ollama instead of sending traffic to cloud-only APIs.

Why Developers Use It

• Private by default: keep content in your local network or self-hosted infrastructure.
• Lower recurring cost for routine generation tasks.
• OpenAI-compatible request shape (/v1/chat/completions) for familiar tooling.
• Dynamic model discovery from Ollama (/api/tags).
• Endpoint health validation to reduce false "connected" states.
• Automatic default model selection from discovered local models.

Screenshots

Connector setup and connected state visuals:

Replace these placeholders with real screenshots as your UI evolves.

Animated Demo (GIF)

• Planned demo path: assets/demos/local-run.gif
• Suggested flow:

1. Open wp-admin/options-connectors.php
2. Enter http://localhost:11434
3. Save connector settings
4. Show successful connected state
5. Generate AI Content

Quick Start (Under 5 Minutes)

1. Install and activate the WordPress AI plugin.
2. Install and run Ollama.
3. Pull a model locally:

ollama pull qwen2.5:7b

4. Confirm Ollama is reachable:

curl http://localhost:11434
curl http://localhost:11434/api/tags

5. Activate this plugin.
6. In WordPress admin, open Settings > Connectors (wp-admin/options-connectors.php).
7. Configure endpoint as http://localhost:11434 and save.

Local Ollama Setup Guide (Step-by-Step)

Use this if you are new to Ollama or want to verify your local setup before connecting WordPress.

1. Start Ollama runtime:

ollama serve

2. Check Ollama server health:

curl http://localhost:11434

Expected output should include Ollama is running.

3. Pull a model:

ollama pull qwen2.5:7b

4. Verify model list from Ollama API:

curl http://localhost:11434/api/tags

5. Confirm your model appears in models.
6. In WordPress Connectors, set endpoint to http://localhost:11434 and save.

If Ollama is hosted on another machine, use that reachable host and port instead (example: http://192.168.1.10:11434).

Installation

1) WordPress Plugin Setup

• Copy this repository into your plugins directory.
• Activate AI Provider for Local Ollama in WordPress admin.

2) Ollama Setup

macOS:

brew install ollama
ollama serve
ollama pull qwen2.5:7b

Linux:

curl -fsSL https://ollama.com/install.sh | sh
ollama serve
ollama pull qwen2.5:7b

Windows:

• Install from ollama.com/download
• Launch Ollama and run model pull in terminal:

ollama pull qwen2.5:7b

Supported Models

Any Ollama model that appears in:

curl http://localhost:11434/api/tags

Examples:

• qwen2.5:7b
• llama3.1:8b
• mistral:7b

Use Cases

• Private editorial drafting in WordPress.
• Internal knowledge and content generation with offline AI workflows.
• Cost-controlled experimentation with local model swaps.
• Development environments requiring no external AI API keys.

Why Local AI Matters

Cloud APIs are excellent for many workloads, but local/self-hosted AI is often preferred when teams need:

• Stronger data locality and privacy guarantees.
• Predictable cost for high-frequency usage.
• Independence from third-party API outages.
• Control over model choice and upgrade timing.

Local Ollama vs Cloud API Providers

Capability	Local Ollama Provider	Cloud OpenAI-Style APIs
Data control	High (self-hosted/private)	Lower (external service)
Marginal usage cost	Usually lower after setup	Per-token billing
Setup effort	Moderate	Low
Latency	Depends on local hardware	Depends on network/region
Internet dependency	Optional	Required

Architecture Overview

WordPress AI Plugin
  -> Connectors API
    -> ollama_local provider (this plugin)
      -> Ollama endpoint (http://localhost:11434)
        -> /api/tags (model discovery)
        -> /v1/chat/completions (generation)

Core files:

• ai-provider-for-local-ollama.php
• includes/class-ollama-provider.php
• includes/class-ollama-model-metadata-directory.php
• includes/class-ollama-text-generation-model.php
• includes/admin-connectors-ui.js

Reliability and Safety Notes

• Endpoint is normalized to http(s)://host[:port].
• Path/query/fragment/user-info are dropped.
• Health checks gate connected state.
• Settings updates are capability-gated.

Troubleshooting

Connector connected, but generation fails

• Confirm Ollama process is running.
• Verify endpoint in connector settings: http://localhost:11434
• Confirm at least one model exists:

curl http://localhost:11434/api/tags

No model supports text_generation

• Ensure /api/tags returns non-empty models.
• Re-save connector endpoint to refresh default model selection.

UI still shows API-key wording

• Hard refresh browser.
• Reopen connectors page.
• Confirm plugin is active and admin JS loaded.

Contributing

Contributions are welcome. For now:

1. Open an issue with reproduction details.
2. Keep pull requests focused and minimal.
3. Include before/after behavior notes for connector and model discovery flow.

(Planned: dedicated CONTRIBUTING.md, templates, and CI checks.)

Roadmap

• Add dedicated docs pages for architecture and development.
• Add issue/PR templates and contributor guidelines.
• Add Markdown/documentation lint workflow.
• Add richer demo assets (real screenshots + setup GIF).
• Add release checklist and semantic versioning guidance.

Star History

If this plugin helps your local WordPress AI workflow, a star helps discovery and signals demand for continued maintenance.

Version

Current plugin header version: 0.2.0