SparkPlus

बदलांची यादी

1.1.4

• Architecture: full client-side AI orchestration. All AI API calls (text and image, for all providers) are now made directly from the browser. PHP no longer proxies any AI request. The server is a pure data layer: it supplies settings, field configs, existing content, and the linking pool on demand, then receives and saves the results the browser sends back.
• Removed the server-side generation pipeline entirely: SparkPlus_Generation_Runner (formerly Generation_Cron), SparkPlus_Prompt_Builder, SparkPlus_OpenAI_API_Caller, SparkPlus_API_Manager, and all three provider classes (SparkPlusOpenAIProvider, SparkPlusAnthropicProvider, SparkPlusGeminiProvider) have been deleted from PHP. None of the flush-early, transient polling, cron, or loopback-request infrastructure they relied on exists any longer.
• The flush_early / fastcgi_finish_request() approach introduced in 1.1.3 and the WP-Cron background dispatch introduced in 1.1.1 are gone. There is nothing left to time out on the server side.
• content-generator.php / SparkPlus_Content_Generator split into three focused files: generation-helpers-trait.php (shared field-map and settings helpers), generation-meta.php / SparkPlus_Generation_Meta (data provider: supplies all generation inputs to the client), and generation-saver.php / SparkPlus_Generation_Saver (result saver: persists AI output to the database).
• New JS module sparkplus-providers.js: browser-side provider classes (SparkPlusOpenAIProvider, SparkPlusAnthropicProvider, SparkPlusGeminiProvider) and SparkPlusProviderFactory mirror the deleted PHP provider layer.
• New JS module sparkplus-prompt-builder.js: browser-side SparkPlusPromptBuilder mirrors the deleted PHP prompt builder, constructing text and image prompts entirely in the browser.
• OpenAI model quirks handled client-side: gpt-5.5 omits the temperature parameter (API only accepts the default); gpt-5.5-pro routes to the /v1/responses endpoint with the correct input / text.format shape and its response is parsed from output[].content[].text.
• Debug panel tabs (Last Text Prompt, Last Image Prompt, Last Text API Response, Last Image API Response) now populate correctly: generation.js emits build_text_prompt, build_image_prompt, and full_api_response entries at the right points in the client-side flow.
• Last Image API Response tab displays the saved WordPress media URL as a visual image preview above the raw JSON, using wp_get_attachment_url() returned by the save endpoint.

1.1.3

• Replaced WP-Cron background dispatch with a direct flush_early approach: the AJAX handler sends the job ID to the browser immediately via fastcgi_finish_request(), then runs the API call inline. This eliminates the server-to-itself loopback HTTP request that WP-Cron required, which was blocked on some hosting environments (including SiteGround).

1.1.2

• Added changelog entries for all previous versions.
• Fixed js/css caching problem by increasing the SPARKPLUS_VERSION constant.

1.1.1

• Moved AI generation work into WP-Cron background jobs to eliminate server-side timeout errors on long API calls.
• Generation cron callbacks are now registered on every request type (including wp-cron.php) via a dedicated generation-cron.php class.
• AJAX handlers for text and image generation now return a job ID immediately and trigger cron asynchronously.

1.1.0

• Debug log entries are now streamed to the client continuously during generation rather than delivered all at once at the end.
• The server writes each debug entry to the transient in real time; the polling client reads and displays new entries incrementally on every poll.

1.0.9

• Generation requests (generate_text, generate_image) now return a job ID immediately and close the HTTP connection early.
• The actual API work runs after the response is flushed, with the result stored in a WordPress transient.
• The JavaScript client polls a new check_job_status endpoint every 2 seconds until the job completes.
• This eliminates 504 gateway timeout errors caused by long-running API calls holding open HTTP connections.

1.0.8

• Added URL slug generation: a new “URL Slug” field can be enabled in the SEO settings tab and will appear in the CPT field list.
• The AI generates a keyword-rich URL slug and applies it as the post permalink (post_name).
• Added reference image support for image fields: a media picker allows attaching a reference image per image field (used by Gemini as inline image data in the prompt).
• Image settings layout in CPT tab reorganised into two rows (Aspect Ratio + WebP % on row 1; Generation Quality + Resolution on row 2).
• Resolution dropdown labels simplified (Low / Medium / High).
• Tooltip icons added to all four image settings.

1.0.7

• Added multi-AI-provider support: Google Gemini and Anthropic Claude can now be used alongside OpenAI.
• New abstract provider base class with concrete implementations for OpenAI, Gemini, and Anthropic.
• API Settings tab redesigned to show separate API key fields per provider, grouped by provider section.
• Deprecated model detection: a warning is shown in settings if a previously saved model is no longer supported.
• Added util.php with shared helper functions including sparkplus_get_supported_models() and sparkplus_get_model_provider().
• Content generator refactored to route all API calls through the new SparkPlus_API_Manager factory.

1.0.6

• Version bump / compatibility update (no functional changes).

1.0.5

• Added optional parent post selection to the keyword loader: when a published post of the selected post type exists, a dropdown allows setting all newly created posts as children of a selected parent.
• Draft parents are shown in the parent post dropdown.
• Prompt builder updated with minor SEO title instruction improvements.

1.0.4

• Added SEO tab in settings with RankMath integration: enables AI generation for RankMath title and description meta fields.
• Added Taxonomies tab in settings (foundation for taxonomy term generation).
• Added post_object ACF field type support in the CPT field list, with a “Nr. of Posts” option instead of word count.
• Warning shown in CPT settings when a post_object field is enabled but the linking pool is not active.
• true_false ACF fields now correctly hide the options column in the CPT field table.

1.0.3

• Added post_object ACF field type detection inside group sub-fields in the CPT settings UI.
• true_false sub-fields inside groups now correctly hide the options column.
• Linking pool enabled state is checked and shown as a warning for post_object sub-fields when the pool is disabled.
• Prompt builder improvements for field instruction generation.

1.0.2

• Improved debug panel: added source parameter (client / server) to all debug entries and debug UI.
• Centralised error handling: all generation errors now show a WordPress admin notice with the post keyword/ID in addition to logging to the debug panel.
• Error notices from previous runs are cleared when a new generation starts.
• Debug panel CSS improvements.

1.0.1

• Internal test release / version bump (no functional changes).

1.0.0

• Initial release.
• Bulk keyword loading with duplicate detection.
• AI text generation for post titles, content, excerpts, and ACF fields.
• AI image generation with automatic WebP conversion.
• Support for all public post types and ACF field groups.
• Three-tier context system (site-wide, per-CPT, per-post).
• Configurable WYSIWYG formatting rules.
• Queue-based generation with debug mode.
• Post editor meta box for keyword and context editing.
• Full settings management with reset capability.