bPlugins AI Provider Connectors
| 开发者 |
bplugins
abuhayat |
|---|---|
| 更新时间 | 2026年10月11日 18:43 |
| PHP版本: | 7.4 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPLv2 or later |
| 版权网址: | 版权信息 |
详情介绍:
- 11 ready-to-use AI providers. Connector cards for OpenRouter, Mistral AI, Cohere, Groq, xAI (Grok), DeepSeek, Perplexity, Together AI, Fireworks AI, Hugging Face and Replicate. Paste an API key, save it, and every feature that uses the WordPress AI Client (such as the AI plugin's excerpt and summary tools) can route through that provider — exactly like the built-in Anthropic, OpenAI and Google providers.
- Test Connection button. A "Test Connection" button next to the "Connected" badge — and inside the expanded edit panel — for every supported provider. Clicking it re-validates the stored API key against the provider's live API and shows the result in a tooltip.
- Custom providers. Add up to five OpenAI-compatible endpoints of your own — a hosted API that is not bundled, or a local model server such as Ollama, LM Studio or vLLM — and they behave exactly like the bundled ones.
- AI Providers dashboard. A Settings → AI Providers screen where you switch providers on or off, see at a glance which ones are connected, run connection tests, manage custom providers, and drag providers into the order the AI Client should try them. When a feature needs an AI provider, the first connected provider in your list that supports the task is used.
- OpenRouter
- Mistral AI
- Cohere
- Groq
- xAI (Grok)
- DeepSeek
- Perplexity
- Together AI
- Fireworks AI
- Hugging Face
- Replicate
- Anthropic, OpenAI, Google — when their respective AI Provider plugins are active.
- All connection tests run server-side; API keys are never sent to the browser.
- Tests and the priority page are gated by the
manage_optionscapability, the standard REST nonce, and admin nonces. - Keys can be supplied via the Connectors screen, via a PHP constant, or via an environment variable — checked in that order. The names follow the WordPress convention
{PROVIDER_ID}_API_KEY, for exampleOPENROUTER_API_KEY,HUGGINGFACE_API_KEYorREPLICATE_API_KEY; custom providers use their own ID the same way. - Connectors and API keys are managed entirely by WordPress core; this plugin only registers providers with the AI Client, just like the official AI Provider plugins.
安装:
- Upload the
b-ai-provider-connectorsfolder to the/wp-content/plugins/directory, or install it from the Plugins screen. - Activate the plugin through the Plugins screen.
- Visit Settings → Connectors. The 11 bundled providers appear in the list alongside any AI Provider plugins you already have installed. Paste an API key into any of them, save, then click Test Connection to verify.
- Visit Settings → AI Providers to switch providers on or off, add custom OpenAI-compatible providers, and choose which provider is tried first.
常见问题:
Which WordPress version do I need?
WordPress 7.0 or newer — the plugin builds on the Connectors API and the AI Client that ship with core since 7.0.
Who can run the test or change the priority?
Only users with the manage_options capability.
Does the test send my API key anywhere new?
No. Each test re-uses the API key already stored for the connector (or supplied via an environment variable / PHP constant) and calls the same provider endpoint a normal client would use.
How does the test know if the key is valid?
Each provider exposes a lightweight endpoint that returns 200 for a valid key and 401/403 for an invalid one (Replicate uses Authorization: Token, all others use Authorization: Bearer). Perplexity has no public /models endpoint, so the test sends an intentionally-malformed chat/completions request and treats 400/422 (auth passed, payload invalid) as a valid key, 401 as invalid.
How do I hide a bundled provider I do not use?
Switch it off on Settings → AI Providers. Disabled providers disappear from the Connectors screen and are never used by AI features; any key you saved is kept in case you switch it back on.
Can I use a local model server?
Yes. Add a custom provider with the server's OpenAI-compatible base URL (for example http://localhost:11434/v1 for Ollama). If the server does not need a key, enter any placeholder on the Connectors screen.
Why do OpenRouter and Hugging Face verify keys differently?
Both publish their model list without authentication, so the AI Client's standard "list models" check would accept any key. This plugin verifies those two against an authenticated endpoint instead.
更新日志:
- Initial release: 11 bundled AI providers, custom OpenAI-compatible providers, Test connection button on the Connectors screen, and the AI Providers settings screen with enable/disable and priority.