Changelog

This update includes bug fixes and new features.

• Fix Resolved an issue where Deep Research could prevent the Strategy Planner from generating content when Google Gemini was the configured provider. Gemini rejects requests that combine its search tool with forced JSON output, which the Strategy Planner requires.
• Changed: Content Gap Suggestions now insert topics directly into the Content Pipeline with publish dates distributed across the next 30 days, instead of creating unscheduled drafts in the Content Library.
• New Added a \\\\\\\"Let articles link out to the sources Deep Research finds\\\\\\\" setting under Deep Research. Disabled by default, it allows users to enable direct links to research sources in generated articles instead of citing sources by name only.

This update includes improvements, new features, and bug fixes.

• Improvement Added support for paid OpenRouter models, improved Pro AI provider reliability, and restored Google Gemini to the Pro provider list.
• New Added an \\\"Also show paid models, with pricing\\\" option to the OpenRouter connection panel, allowing licensed sites to browse the full model catalog and view pricing per million tokens.
• New Added live model discovery for Anthropic, Groq, Mistral, and DeepSeek, ensuring current model lists are fetched dynamically instead of relying on potentially outdated defaults.
• Changed: Restored Google Gemini as a Pro provider alongside Anthropic, Groq, Mistral, DeepSeek, and Ollama. Existing licensed Gemini connections continue to work without requiring any action.
• Changed: When a Pro license expires while a paid OpenRouter model is configured, generation continues using a free model, and the OpenRouter card explains the restriction. Activating or upgrading Pro does not automatically switch generation to a paid model.
• Fix Improved Gemini reliability by replacing a blocked default model, translating HTTP 429 quota errors into clear messages, and resolving a key-testing issue that returned \\\"Empty response from provider.\\\"
• Fix Replaced four unavailable or deprecated model defaults across Groq, Mistral, DeepSeek, and GPT Image thumbnails with the providers\\\' current recommendations.
• Fix Improved Ollama reliability by routing localhost requests through 127.0.0.1, increasing the timeout to 300 seconds for slower hardware, and resolving a timeout race condition that could cause a fatal error.
• Fix Corrected the SEO minimum-score retry loop so genuine generation failures during regeneration are reported accurately instead of being discarded.

This update includes new features.

• New Moved the full implementation of Real AI Thumbnails, Content Gap Suggestions, Finer Automation Control (multi-time/day-of-week scheduling), SEO enforced-minimum retry loop, Deep Research, Custom Prompts, and Content Refresh into Pro. These features were previously included in the free SynaPost plugin and have been moved to keep the WordPress.org-submitted Lite package limited to genuinely free functionality. No behavior changes for licensed Pro users; all features continue to work as before.