Could Not Generate a Question

"Could not generate a question. Please try again."

This is the widget's fallback message. The JavaScript shows it when the request failed and the server did not supply a more specific message it could display.

Because it is a catch-all, the useful diagnostic step is to find out what the server actually said.

Step 1: read the real error

Open your browser's developer tools, go to the Network tab, and reload the page. Look for the request to:

/wp-json/ai-fun-questions/v1/question

Click it and read the response. It will contain a code and a message far more specific than the fallback. Common ones:

Code	Meaning
ai_fq_openai_config, ai_fq_ollama_config, ai_fq_hf_config	Settings incomplete for the selected provider
ai_fq_provider_error	The call was attempted and failed
ai_fq_invalid_json	The model's reply was not usable JSON
ai_fq_missing_field	A required field was missing from the reply
ai_fq_invalid_content	The reply was too long or empty once sanitised
ai_fq_rate_limited	Rate limited — see the dedicated page
ai_fq_invalid_widget	The request had no valid widget token

Look each up in Error Messages Reference.

Step 2: if the request never happened

If there is no request to that endpoint at all, the problem is on the page, not the provider:

• The widget's JavaScript did not load. Check the Console tab for errors. An optimisation or minification plugin that defers, combines or rewrites scripts is the usual culprit — exclude the plugin's script and retest.
• The REST API is blocked. Some security plugins and hosts restrict /wp-json/. The plugin's two endpoints are intentionally public, because the widget must work for logged-out visitors. If your security plugin blocks anonymous REST access, allow these two routes.
• The shortcode did not render. If you see the literal text [ai_fun_question] on the page, you have put it in a Custom HTML block, which does not run shortcodes. Use a Shortcode block.

Step 3: the common causes, in order

Settings incomplete. The most frequent cause by far. Remember that only the selected provider's panel matters — a fully-filled Hugging Face panel does nothing if OpenAI-compatible is selected.

The server cannot reach the provider. The AI call is made from your server, not the visitor's browser. A host that blocks outbound HTTP, or a firewall in the way, produces this even though the credentials are perfect. Test from the server, not your laptop.

Bad credentials or no credit. An expired key, a revoked token, or an account out of credit all surface as ai_fq_provider_error, because the plugin will not show you the provider's own error text.

The model is not suitable. A model that cannot reliably return JSON produces ai_fq_invalid_json or ai_fq_missing_field, often intermittently — some questions work, some do not. Try a larger instruction-tuned model.

Ollama unreachable from PHP. A classic: Ollama answers fine in your terminal but PHP-FPM runs as another user or in a container with a different network view. Test from the web server's perspective.

Step 4: the debug log

If the network response is ai_fq_provider_error and you still cannot tell why, the upstream error is written to the PHP error log when WP_DEBUG is enabled.

Do this on staging, not production. The plugin deliberately keeps upstream errors away from visitors, and debug logging is a separate path that can capture request detail.

define( 'WP_DEBUG', true );
define( 'WP_DEBUG_LOG', true );
define( 'WP_DEBUG_DISPLAY', false );

Reproduce the failure, read wp-content/debug.log, then turn it off again.

What will not help

• Reinstalling the plugin. This message is nearly always configuration or connectivity.
• Clearing your cache. The widget fetches at runtime; cached HTML is not the issue.
• Switching provider blindly. Find the real error first — you will usually hit the same wall.