=== Sanaseppä Free ===
Contributors: wpplugi
Tags: writing, ai, editor, finnish, content
Requires at least: 6.0
Tested up to: 7.0
Requires PHP: 7.4
Stable tag: 1.0-rc10
License: GPLv2 or later
License URI: https://www.gnu.org/licenses/gpl-2.0.html

Finnish writing assistant for editing individual texts, posts and pages in WordPress.

== Description ==

Sanaseppä helps Finnish WordPress users improve text directly in the admin area. The free version focuses on practical single-content editing.

With Sanaseppä Free you can:

* Edit a single pasted text.
* Edit individual posts and pages.
* Compare original and rewritten content before saving.
* Save changes as a draft or update the original content.
* Use GPT-5 mini, GPT-5 nano, GPT-5 or GPT-4.1 through your own OpenAI API key.
* Review recent editing history.

A separate Pro add-on may provide advanced features in the future, such as bulk processing, WooCommerce product editing, extended history, custom prompts and brand voice tools.

== Installation ==

1. Upload the plugin folder to `/wp-content/plugins/` or install the ZIP file from the WordPress admin area.
2. Activate Sanaseppä from the Plugins screen.
3. Open Sanaseppä → Settings.
4. Add your OpenAI API key and choose a model.
5. Open Sanaseppä → Text tool and create your first rewritten version.

== Frequently Asked Questions ==

= Does Sanaseppä publish AI text automatically? =

No. You review the rewritten text before saving it.

= Is an OpenAI API key required? =

Yes. The free plugin uses your own OpenAI API key.

= Can I edit products or many posts at once? =

The free plugin focuses on individual texts, posts and pages. Product editing and bulk processing are intended for a separate Pro add-on.

== Changelog ==

= 1.0 RC10 =
* Moved the before/after and structure comparison below the actual edited result.
* Made the comparison section collapsible and secondary, opening automatically only when warnings need attention.
* Simplified preview messages so the edited content is the primary review target.

= 1.0 RC9 =
* Added automatic browser-side chunking for the single text tool so longer GPT-5 mini edits are processed as several short requests instead of one hosting-timeout-prone request.
* Added progress messages for chunked text editing and combines the rewritten chunks back into one suggestion.

= 1.0 RC8 =
* Route direct GPT-5 model requests through OpenAI Responses API first, with GPT-5 speed settings and dynamic output limits.
* Keep the older chat/completions route for non-GPT-5 models and as a fallback for clear Responses API request errors.

= 1.0 RC7 =
* Added GPT-5 speed settings for direct OpenAI requests so GPT-5 mini can answer faster on hosts with short request timeouts.
* Added an automatic retry without GPT-5 speed parameters if an API account rejects those parameters.
* Removed misleading connector guidance from direct API timeout messages.
* Tuned GPT-5 completion limits dynamically by prompt length.

= 1.0 RC6 =
* Changed the OpenAI timeout to finish before typical hosting request timeouts, so WordPress can show a controlled Sanaseppä error instead of a server 500 timeout page.
* Tuned the GPT-5 response cap to balance readable output with slower shared-hosting environments.
* Improved the admin error shown when a host returns an HTML Request Timeout page.
* Lowered new-install default text limits to reduce timeout risk on shared hosting.

= 1.0 RC5 =
* Increased the GPT-5 response cap so visible rewritten text is not starved by the completion limit.
* Improved empty-response handling by reading multiple OpenAI response shapes and showing a clearer message when the response stopped because of the length cap.
* Made the API connection test verify that OpenAI returned readable text, not only HTTP 200.

= 1.0 RC4 =
* Added model-aware response length limits to direct OpenAI requests so GPT-5 mini cannot run indefinitely inside the hosting timeout window.
* Added the same lightweight response cap to the API connection test.
* Kept GPT-5 mini as the default and did not force GPT-5 nano.

= 1.0 RC3 =
* Restored the older high-quality model choices GPT-5 and GPT-4.1 to avoid forcing users to GPT-5 nano when GPT-5 mini times out.
* Kept GPT-5 mini as the default model.
* Updated timeout guidance to suggest GPT-4.1 before the lighter GPT-5 nano model.

= 1.0 RC2 =
* Kept the cleaned Free-version interface and user guide from the newer Free build.
* Restored the older direct OpenAI chat/completions engine for text rewriting.
* Kept GPT-5 mini/nano compatible by not sending unsupported temperature values to GPT-5 models.
* Removed the newer GPT-5 rewrite path and restored the classic chat/completions path.
* Kept GPT-5 mini and GPT-5 nano as the Free-version model choices.
* Changed the direct API test back to the lightweight chat/completions diagnostic call.

= 0.77.25 =
* Changed the direct OpenAI API connection test to use a lightweight model reachability check instead of a text generation request.
* Improved timeout guidance so hosting/network issues are easier to distinguish from slow text generation.

= 0.77.24 =
* Increased the OpenAI HTTP timeout from 20 seconds to 45 seconds.

= 0.77.23 =
* Fixed request timeout risk by avoiding multiple OpenAI calls inside one post/page rewrite request.

= 0.77.22 =
* Cleaned the free plugin interface and user guide.
* Focused the free version on individual texts, posts and pages.
