Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 4 Next »

Squirro 3.7.3 was released on March 22, 2023.

What’s New

  • Concept search performance has been improved, including better identification of semantically-similar terms for searches.

  • Added a Feedback Mode toggle button to the Actions widget.

  • All new projects have significantly improved storage efficiency through changes to the mapped summary field, which is now just an abstract and not the fully body text. You can manually set the maximum size of this abstract using the new server setting lib.pipeline.content.max-summary-size. Truncation can be disabled by setting the value to 0. When you load data with an existing summary field, you can automatically map that field. Otherwise, the first paragraph up to the max summary size is used. This also means that PDF sub-items do not populate the sub_item.summary field anymore.

  • In the Ingestion Pipeline, you can now configure the amount of max-characters via the new server-level setting lib.pipeline.content.max-summary-size. By default, this is set to 600 characters.

  • Added the topic.item.allowed-reader-edit-keywords project-level configuration, which allows setting specific Item keywords that users with the Reader role are allowed to edit.

  • Added the configuration option topic.typeahead.facetvalue.configuration to allow simpler performance tuning for facet-value typeahead. For more details, see https://go.squirro.com/docs-typeahead

Improvements

  • A wider range of languages is now supported in the NLP Keyphrase Tagger (e.g. Chinese, Russian).

  • Previously, users were unable to clear filters applied through persistent links, if they pasted a url that contained filters for example. Now, users can reset those filters via the Reset Filters widget.

  • Now pipeline save error messages are shown to the user.

  • Added support for nested aggregations for merged entities setup.

  • The width of the Global Search bar was increased.

  • In the Dashboard Editor, moved the widget browser Close and Add Layer/Widget buttons to the top.

  • Removed the Excel and Alert options from the dashboard More button.

  • For those with the preview service-nlp enabled, added the possibility to accelerate QA models via ONNX runtime. This can be done in the configuration file by setting use_ort to True.

  • The Navigation Chips and Table widgets were migrated to React.

  • After favoriting a query, the toast will show a link to see all favorites if the link is configured in server settings.

  • Query endpoint search performance was improved by reducing the amount of fetched payload by loading only a specific subset of interesting facets client.query(fields=[keywords.facet_name, keywords.facet_name2]). To fetch all, provide fields=["keywords"] or fields=["keywords.*"]

  • Exposed facets aggregations collection to custom widgets via own factory.

  • Improved Groundtruth documentation via ml_groundtruth.py.

  • Improved nested entity highlighting and entity sorting for PDFs and non PDFs.

  • Entities for PDFs are now stored with a page_info property, which includes information about the size (width and height) of the page that the entity was found on.

  • Global Search configuration was simplified to a boolean checkbox.

  • Added support for adding custom widget context values inside of the main.js file.

  • Updated CommunityCard overrides to include CommunityCardComponent.

  • Improved the useWithoutCollection React hook to include props and config.

  • For searches utilizing advanced query syntax techniques, added support for nested, negated groups like NOT ( (tag:foo OR tag:bar) AND hello ).

  • Added a link to open Community 360 pages in a new tab from the Tabs widget.

  • Now, the Reset Filters widget accordion shows selection count.

  • Improved the styling for Rules table buttons in AI Studio.

Bug Fixes

  • Fixed a bug where PDF searches included the “Page” and page number fields, returning unwanted results. Now, searches only take place in the body of PDF Sub-Items. Sub-Item title (Page Nr) is not used for text-matching nor highlighting.

  • Changed the return value from id to rule_id to keep it uniform.

  • Fixed an issue with the early closing of entity highlights in situations where the end and beginning of highlights overlap.

  • Fixed an issue with the QA widget not opening the same answer twice.

  • Fixed styling issues with the Items widget sort dropdown menu.

  • Enabled full support of ACL filters within content-based typeahead as defined in query_templates (with logical operators).

  • Fixed an issue with the Add Rule button being hidden in AI Studio.

  • In Groundtruth Labeling, fixed an issue with labels not persisting when switching documents.

  • Fixed an issue with tabs not respecting url parameters when refreshing certain pages.

  • Fixed an issues with the Items widget fetching multiple pages at once.

  • Previously, items would be dropped during ingestion if they did not contain mapped field content (emails without attachments when the attachments field was mapped, for example.) Now, items are not dropped because they are missing configured non-mandatory mappings. The mandatory Squirro Item fields are body and title, which must be present to successfully ingest an item.

  • Fixed an issue with Groundtruths showing labels from different ones.

  • Fixed an issue with the Rules table not correctly adjusting when moving to another tab.

  • Fixed an issue with resetting dashboard selections on home in breadcrumbs.

  • Force the source edit to fetch sources collection.

  • Fixed an issue with Entities chips sometimes not showing.

  • Fixed an issue with the checkmark not showing correctly in subscribed communities for users.

  • Fixed errors with PDF entity sorting.

  • Fixed an issue where Go to Groundtruth disappears when the candidate set is on different page.

  • Fixed an issue with CandidateSetEdit crashing when an item is not found.

  • All of a user's subscriptions to communities now correctly show as subscribed.

  • Fixed an issue with widgets on the Community 360 dashboard not reacting to a Global Search reset.

  • Fixed an issue where clearing the global search on the Community 360 page removed the selected community query.

  • Improved Elasticsearch performance in edge cases by auto-closing scroll context when the client stops the iterator via break, closing properly when all returned items fit into a single batch, and extending client scan to explicitly close scroll context.

  • Fixed issue with the Groundtruth in AI Studio not showing the proper list of items.

Installation and Upgrade

You will have to resolve at least the following config files when upgrading from Squirro 3.3.0

  • /etc/nginx/conf.d/ssl.inc

  • /etc/squirro/common.ini

For new installations, please follow the Setup on Linux instructions.

To upgrade an existing installation, please consult the Upgrades for Squirro 3.5.3 and later guide.

  • No labels