Opensolr Search

1.0.39

• Added: PDF results are shown page by page. The Opensolr Web Crawler and Data Ingestion now index every PDF as one document per page, so a search lands on the page that actually answers it. Each page result carries a Page N pill that opens the PDF at that page.
• Added: pages of the same PDF that follow each other in the results are grouped under the document’s title, with how many of its pages matched. The five best pages are shown, the rest open with one click, and the group header stays on screen while scrolling.
• Added: Read Page summarises the page that matched; Read Document summarises the whole PDF, every page in order.
• Added: page results show the PDF’s page count and creation date.

1.0.38

• Fixed: the +word and -word search operators did nothing when AI search was on. They are ordinary keyword-search syntax, so they only ever bound the keyword half of a hybrid search, and every result the keyword half excluded came back in through the semantic half. Exclusion failed twice over, because the query was embedded with the minus sign still in it and an embedding model has no concept of negation, so -Ruben read as one more word and pulled results towards Ruben. Operators are now applied as filters over the whole result set.
• Fixed: quoted phrases were stripped out of hybrid queries, so "machine learning" searched for the two words separately. Phrases now survive, and +"machine learning" / -"open box" work as required and excluded phrases.

1.0.37

• Fixed: the ingestion cron now retries a batch that failed instead of skipping it, so documents are no longer silently left out of the index.
• Fixed: the WPML language filter now uses the language’s full locale, so results are scoped to the right language.
• Fixed: the plan badge on the settings screen showed a meaningless 0.00 price on free accounts. It now reads “free forever”, and paid plans still show their price.
• Added: the “did you mean” spelling suggestion now also appears when AI search is on; only image search hides it.
• Changed: on the search page the sort dropdowns are easier to read, and the result count sits on the controls row.
• Changed: wording updated throughout — the Opensolr free plan no longer expires after 15 days, it is free forever with no credit card.

1.0.36

• Security: hardened the search-as-you-type (autocomplete) so a specially crafted query can no longer change how the search is parsed on the server.
• Fixed: the search-as-you-type suggestions now line up directly under the search box.

1.0.35

• Added: Search by image. A camera button on the search box turns a photo into a search — it reads the picture as CLIP labels, OCR text or a barcode, keeps the thumbnail on screen, and offers a “search by picture meaning” alternative. A spinner shows while the picture is read.
• Added: result thumbnails grow on hover and open full size in a lightbox with the page title linked underneath; the searched image opens the same viewer with the words and codes it was read as.
• Added: an AI (hybrid) search toggle on the results page, so a visitor can switch between hybrid vector search and pure keyword search per search.
• Added: a Duration facet widget that renders an h:mm:ss min/max slider for integer-seconds fields such as video length.
• Added: the facet Filter box now searches the complete value list of a field, not just the values already shown; the same values power the “Show all” dialog with a single fetch.
• Added: “Use an existing index” during setup, so you can attach the plugin to any index your Opensolr account already owns instead of only creating a new one.
• Added: richer document results — page count, content type, detected language, author and modification date.
• Fixed: the search box no longer overflows on narrow phones, and the search UI now uses the full width on mobile.
• Fixed: spellcheck suggestions are suppressed on an image search and when AI search is on.

1.0.34

• Security: password-protected posts are no longer indexed. They are published, so the previous check let them through, and the indexed body was the expanded content rather than the password form — their full text was readable from the search results page.
• Security: WooCommerce products marked Hidden or Exclude from search are no longer indexed. Visibility is stored in the product_visibility taxonomy, which the plugin never consulted.
• Fixed: product price, sale price, stock and SKU changes now reach the index. WooCommerce stores them in post meta and does not fire save_post, so an order decrementing stock, a scheduled sale starting or a bulk price edit never updated the index and price sorting and filtering used stale values.
• Fixed: Save & Start Fresh no longer deletes the existing index before its replacement has been created, so a transient API error can no longer leave the site with neither.
• Fixed: four settings could never be saved — Browse mode, Infinite scroll, a custom Minimum Match expression, and any crawler Request Delay below one second, which was stored as zero and told the crawler not to wait at all.
• Fixed: the Hierarchical (drill-down) facet widget was silently downgraded to List on save, and the per-facet minimum count was discarded on every save.
• Fixed: the facet order set by dragging rows was scrambled on the next page load and then saved in that scrambled order.
• Fixed: the embedded search widget rendered below the page footer and only occupied part of the viewport; it now fills the content area.
• Fixed: the Data Crawler tab always reported 0 URLs in sitemap.
• Fixed: the AI panel could render an error page as though it were the answer.
• Added: the ingest job reports how many items were skipped because they are not publicly viewable.
• Added: links to the embed parameter documentation from the Search Page tab.