SEO Copilot v1.1.4

Fixed

• Applied changes now flush the SEO plugin's own cache so they show up immediately. After writing, apply() calls rank_math_clear_cache() (and fires a new seocp_after_apply action). On object-cache hosts like Kinsta (Redis), Rank Math serves derived/rendered meta from cache, so a raw postmeta write could stay hidden in the editor and front-end until the cache expired — the exact "the data is generated but the product shows nothing" symptom. This matches the cache-clear step a known-working reference plugin performs.

Added

• seocp_after_apply action ($post_id, $written_field_ids) for flushing page caches / CDNs after a write.

SEO Copilot v1.1.3

Fixed

• Bulk writes silently dropped under cron because there was no current user. Individual writes from Smart Optimizer / Pending Review worked because they run inside an authenticated REST request, but bulk processing runs from WP-Cron with no logged-in user — so the auth_callback registered by Rank Math / Yoast on their meta keys rejected the write. update_post_meta() returned truthy, the Logs read applied, and the product stayed empty. The Bulk Runner and OpenAI Batch dispatcher now restore the originating user before each apply() call (with a safe fallback to any admin who can edit the post if the originating user was deleted).
• Schema v1.4.0 adds queue.created_by (auto-upgraded on plugin load) so cron knows which user enqueued the batch.

SEO Copilot v1.1.2

Fixed

• Bulk apply now verifies each write actually persisted instead of trusting update_post_meta()'s return value. On large catalogs with a persistent object cache (Redis/Memcached), a DB read-replica, or a plugin that filters postmeta, update_post_meta() can return success while the value never becomes visible — so the Logs showed applied but the product stayed empty. apply() now flushes the post cache, re-reads each field from the database, counts only verified writes (so a silent failure shows as noop, not a false applied), and logs the exact field/post that failed with a diagnostic hint. The cache flush also fixes the case where the write did persist but a stale cache was hiding it from the editor/front-end.
• New filter seocp_verify_writes (default true) to disable the read-back if needed.

SEO Copilot v1.1.1

Fixed

• "Apply entire batch" in Pending Review now applies the whole batch. It was capped at ~500 segments per click (list_pending limit), so large review batches needed hundreds of clicks. The button now drains the batch in chunks until nothing is pending, with live progress, and auto-resolves segments that can't be written so the loop always terminates.
• "Missing focus keyword" preset now actually filters. It used relation => OR across the Rank Math / Yoast / AIOSEO keys, so on a single-plugin site the two inactive plugins' keys (always absent) matched every post — the filter returned the whole catalog, including products that already had a focus keyword. Both "missing" presets now use AND across only the active SEO plugins' keys (with empty-string checks), so they return just the posts genuinely missing the field. This is what was causing bulk runs to spend tokens on already-optimized products.
• OpenAI Batch mode no longer fails most chunks on large catalogs. Previously every chunk was submitted near-simultaneously (cron + browser poll), blowing past OpenAI's per-model enqueued-token limit so all but the first chunk failed 0/N. Submissions are now serialized behind a concurrency gate (seocp_openai_batch_concurrency, default 1): a new chunk is only submitted once earlier chunks free up batch-queue capacity.
• Transient batch failures are retried instead of permanently killing a whole chunk. Token/rate-limit failures and 24h expiries reset the chunk to draft and re-enqueue its rows (up to seocp_openai_batch_max_retries, default 5). Validation errors in our own request file are still failed fast.
• Partially-completed batches now also ingest OpenAI's error file and resolve any items missing from both files, so a batch can't hang below 100% forever.
• Bulk Wizard chunk pills now show a retry N state and surface the OpenAI failure reason on hover.

Changed

• seocp_openai_batches gained an attempts column (Schema v1.3.0, auto-upgraded on plugin load).

SEO Copilot v1.1.0

Added

• OpenAI Batch API dispatch in the Bulk Wizard — submit large jobs (e.g. 50k+ products) to OpenAI's Batch API for 50% lower token cost and no per-minute rate-limit ceiling. Results return within 24h and apply automatically as they download. Pick "OpenAI Batch API" in Step 4 → Processing mode.
• New seocp_openai_batches table tracking per-chunk lifecycle (build → submit → poll → apply).
• BatchDispatcher service orchestrates the OpenAI Files + Batches APIs end-to-end.
• Browser-driven /bulk/tick now also nudges the batch dispatcher so progress advances without waiting for WP-Cron.
• Filters: seocp_openai_batch_chunk_size (default 5000), seocp_batch_apply_size (default 500).

Changed

• Settings → Connection: raised the Requests per minute cap from 600 to 10,000 so users on higher OpenAI tiers can configure realistic limits. Sub-text now points to Batch mode for 500+ post runs.
• Queue table gained dispatch, openai_custom_id, and payload_response columns (Schema v1.2.0, auto-upgraded on plugin load).
• Bulk Wizard's "Recent batches" panel surfaces a batch badge for Batch-API runs.

v1.0.0

SEO Copilot — Initial Release

AI-powered SEO content for any WordPress post type. Per-field, per-template control. Fluent 2 UI.

Highlights

• Smart Optimizer — single-post wizard with Google SERP preview, character-budget counters, and side-by-side diff of current vs. proposed values.
• Bulk Wizard — paginated post picker with 'Select all matching' for large catalogs, per-batch cost estimate, and WP-Cron background processing.
• Templates — reusable prompt blueprints with per-field, per-post-type toggles.
• Pending Review — review and apply AI proposals before they touch your content.
• Multi-keyword — generates 3-5 focus keywords per field, split into Rank Math / Yoast / SEOPress / AIOSEO formats.
• Local SEO — baked into every default template with configurable city, region, and country.
• Bricks Builder — reads Bricks content into the prompt context when present.
• Fluent 2 UI — hand-authored design system with WCAG-AA contrast. Light by default, dark opt-in.
• OpenAI — gpt-4.1 / gpt-4o family via JSON mode.

See CHANGELOG.md.

Install: upload the attached ZIP via Plugins > Add New > Upload Plugin.