Changelog
What’s new in Yazgan
Updates arrive through the WordPress updates screen, from temorya.com with your license.
0.3.1Unlicensed campaign limit
- Without a license, posts generated by campaigns (including scheduled runs and WP-CLI) now count toward the same 3 free drafts a day; previously only the Generate article screen was limited. When the allowance is used up, campaign posts are not deleted: they wait in the queue and continue the next day. No change on licensed sites.
- Yazgan now runs on a temorya.com license: enter and activate your key on the License screen. A license is active on one site; 1 year or lifetime.
- License screen: plan (Yearly / Lifetime), validity, support and update end date, this domain, last verification; "Verify again", "Manage in my account", "Renew / go lifetime", "Move to another site" and "Remove from this site".
- Security: license responses are signed by temorya.com and re-verified on every load; a record written into the database by hand or copied from another site is not a license. Demo mode only turns on at local development addresses (localhost, .test); no setting or constant can turn it on.
- If the server is temporarily unreachable, features stay on for up to 7 days ("Awaiting verification") and the license renews itself once the connection is back. A server decision to revoke or expire is applied immediately.
- Autonomous publishing, the bulk row limit and image generation follow the features included in your license. Unlicensed use is unchanged: 3 manual drafts a day, autonomous publishing off. Your posts and your site's front end are never affected by any license state.
- Updates come from temorya.com through the WordPress update screen, with a license.
- When a free Gemini key reaches its daily limit (20 requests per model) a clear message: "daily limit of 20 requests reached; it resets today/tomorrow at ~HH:MM, or enable billing" (time in the site time zone; Google resets the daily quota at midnight Pacific time).
- When the quota is used up the article waits in the queue until the reset time and then continues from the same step; waiting does not use up an attempt. Article generation shows "Waiting for quota" instead of polling.
- Article generation: with Gemini selected, "about 6 articles a day with a free key (one article = 3 requests)"; if the quota is used up, when it resets.
- Test no longer uses quota: it validates the key by reading the model list only (Gemini, GPT, Claude). A separate "Try generation" button sends one real, short request.
- Optional (off by default): AI › Gemini › "When quota runs out, switch to another flash model" (default fallback gemini-3.1-flash-lite).
- Fix: the queue attempt counter is reset per stage; previously the first temporary error in the writing stage could fail an article without retrying.
- Fixed a "critical error" on Campaigns › New campaign when a provider key is saved (regression from 0.2.5).
- Screen tests now also run with keys saved (this kind of error is caught before packaging).
0.2.6Image generation error cards
- The same billing/quota error no longer repeats on every card: one notice above the preview (explanation + Google AI Studio link + provider message), cards show a short "Billing required". Error text inside cards wraps instead of being cut.
- Google images are generated with Gemini image models (generateContent); default gemini-3.1-flash-image, the key's -preview name is picked automatically. Imagen (predict) is not offered to new keys and was removed. The image model can be changed under AI › Gemini › Image model.
- Free keys have an image quota of 0 (429): "Google image generation requires billing; article generation keeps working" + a Google AI Studio link. No extra setting once billing is on.
- Model suggestions come from the key's real model list (text + image, cached 12 h; Test refreshes).
- Clear image errors: billing required, model not available with this key, quota; the provider's original message is folded under details.
- The campaign wizard shows the model next to the provider.
- Gemini invalid key (HTTP 400, API_KEY_INVALID) now shows the translated "API key is invalid" message.
- Image generation provider cards show the same state as the AI screen: connected / saved · not tested / test failed.
- Gemini default is gemini-3.5-flash (works with a free key); suggestions refreshed with current models. gemini-2.5-* is closed to new users.
- Clear provider errors: 404 "this model is not available with this key", 429 "no quota / Pro closed on free keys"; Test suggests models to try (recommended first).
- Each provider (Claude, GPT, Gemini) has its own connection panel: key, model, test. Selecting a provider card switches the panel (no JavaScript needed).
- If Test reports "model not found", the models available with this key are listed (Gemini ListModels, OpenAI models).
- Keys and models are stored per provider; GPT/Gemini get model suggestions plus free entry (defaults gpt-5, gemini-2.5-pro). A Claude model ID is never stored for another provider.
- Image and Article generation: native radio circles no longer overlap the labels when no key is set or on "Coming soon" providers (one shared fix for all segment, choice-card and switch controls).
- Bulk generation: clear "opens once a key is entered" card with a link to the AI screen.
- License badge readable at every width: state-coloured background, short "Demo" label in the collapsed sidebar and on phones.
- Promo and demo links come from settings (no hard-coded address).
- Article generation (live step bar, asynchronous), review queue, "Edit here" (WordPress editor).
- Writing guide: research notes, type templates, editor pass, internal links; quality gate (fact mapping, similarity, clichés, cannibalization, YMYL), publish threshold 80.
- Image generation (OpenAI GPT Image, Google Imagen; WebP, alt text, SEO file name), bulk generation (up to 50 rows, 2 at a time).
- Mergen and SEO plugin fields; demo license mode.
0.1.0development (Phase 1a)
- Modular skeleton: schema-based settings, encrypted key storage, masked logging, database migrations, capabilities.
- Safe HTTP client: internal network addresses blocked, TLS verification always on, every redirect re-checked.
- robots.txt support and per-site request spacing (minimum 2 seconds).
- AI providers: Claude (recommended), OpenAI, Gemini; monthly budget cap and usage log.
- RSS / Atom source; duplicate check by URL + content similarity.
- Content pipeline: draft → quality gate (blocks copies, unverified links, banned words, fake experience) → WordPress draft (review queue).
- WP-CLI:
wp yazgan run, wp yazgan source-test, wp yazgan status.