3.10.2 - Release Notes#

Squirro 3.10.2 was released May 10, 2024.

Reference: Learn more about the Squirro Release Process.

Known Issues#

  • My Communities and Recommended Communities tabs in the Cognitive Search projects don’t work properly.

What’s New#

  • Add endpoint for acquiring user session information with a caching mechanism.

  • Restrict access to sessions that have been flagged as deleted.

  • Add a response checker chain. This will flag if the generated answer from the GenAI service was giving a satisfactory answer or not.

  • Add a new chain that now also does keyword-based query formulation. This will help in cases where semantic search will not be enough to retrieve documents. Especially when specific terms are searched.

  • Add new toggle to switch between editing a model with JsonForms or with the ACE editor in the server and project configuration menus (only available for configurations with a defined schema).

  • The pipeline step Thumbnail Extraction has a new config option for setting the format of the generated thumbnail image. It supports JPEG and PNG formats. The default format has changed from PNG to JPEG.

  • Allow MUI theme customization.

  • The plumber service can now be deployed as a Docker container. For detailed instructions, refer to the official Squirro documentation on “Deploying the Plumber service in a Docker container”.

Improvements#

  • Parse new editable chips into separate chips if applicable.

  • Respect open state in item detail sidepanel icon.

  • Auto-focus modal search input in item detail.

  • Add action button to open list of all favorites from the create favorite modal success snackbar.

  • Add support for Unicode filenames in the X-Filename HTTP header. The filenames are now properly encoded using percent-encoding standard for compatibility with various browsers and HTTP clients.

  • Typeahead suggestions redesign.

  • Add caching to fetching user grants for SQGPT. This can lead to an increase in performance to about 30%, if a project is affected by grant leakage.

  • Parse SSO information from the proxy to GenAI service.

  • Do not pass None values as a query to query transformers, but instead pass an empty string. It allows query transformer to return the original query without explicitly changing None to an empty string.

  • Use byte embeddings type by default for both Paragraph Embedding pipeline step and Semantic Search Scoring plugin. The change is backward compatible so projects that already use semantic search will use the float type by default.

  • When building the “sparse” query, use the active query processor workflow if available. If not, default to the query_default query processor workflow.

  • Add configuration option labels widget to disable dropdown when empty.

  • Add support for Unicode filenames in the X-Filename HTTP header. The filenames are now properly encoded using percent-encoding standard for compatibility with various browsers and HTTP clients.

  • Store Elasticsearch credentials using the Filebeat keystore. References to these credentials have been updated in the Filebeat configuration file to access the stored values. This change eliminates the need to store sensitive information in plaintext.

  • SquirroGPT - Company Data and Squirro GPT - Mixtral project templates have been updated to include Cognitive Search dashboards as well as Chat with Results and Chat with Document widgets.

  • New expiration handling of user sessions is now enabled by default (security.expiring-sessions).

  • Expose an option to specify what item fields to export when exporting Squirro project.

  • Fix an issue where multiple processes for PDF conversion were not terminating correctly, leading to orphaned instances.

  • Indicate active search in ItemDetail

  • Exclude semantic_chunks field from the payload sent to ML service.

  • Remove card padding on communities list widget when card styling is disabled in config.

  • Add information about the server used for the GPU vs CPU experiment described in the Semantic Search documentation (https://docs.squirro.com/en/latest/technical/search/features/semantic-search.html#efficiency-comparison).

  • Support the Retry Failed Items and Rerun from Raw Data features for the [QFSS](https://docs.squirro.com/en/latest/technical/data-processing/content-streamers.html#queuefilesystemstreamer) system.

  • Reset search query when changing items in item detail.

  • Convert the whole query to favorite chip after creating it from global search.

  • Add clamping to the idetail title.

  • Expos GET endpoint of /sessions/<session_id> in SquirroClient via get_session().

  • Add client kwards to squirro_retriever. This will enable parsing information such as user_information, auth_service, auth_user to the squirro client in the squirro_retriever in the genai service.

  • Add Grid mode to the Items widget.

  • Add search button control to the item detail toolbar.

  • Mark split view icon in item detail toolbar active if open.

  • Overlay sidepanel in item details on small screens 900px.

  • Re-work how SquirroGPT project and server-level configurations behave. The genai.sqgpt.default-settings configuration was introduced on the server level to set defaults for all the existing SquirroGPT projects. These values are overridden by the values of the genai.sqgpt.settings configuration of each project. Values on the project-level configuration that are set to null or to “Default to Server Setting” will fallback to the server-level configuration.

  • Make tooltips consistent with item cards in item detail.

  • Item Detail Rework.

  • Improve the performance of retrieving communities. This should be visible in many places, mainly when doing a search.

  • Update icon fonts to the latest version and change all icons to the rounded type.

  • GenAI will now persist code change in the container when restarted. However, these changes will be overwritten once a new image is being used. The volume will be mounted on /var/lib/squirro/docker/volumes.

  • Improve performance of closing modal.

  • Change MySQL driver from the PyMySQL to the MariaDB Connector. The PyMySQL driver is built completely in Python and has a huge bottleneck when retrieving a bigger number of rows. The bottleneck blocks the parallel requests, which results in much slower responses. The MariaDB Connector uses C connector and thus is much faster.

  • Include nltk punkt and spacy model dependencies in the docker image. This will stop downloading those dependencies after startup.

  • Add autocomplete for facet values.

  • Disable prev/next buttons in idetail on first/last item.

  • Remove entity_query from item model to improve performance (It is not doing anything meaningful at this moment).

Bug Fixes#

  • Fix issue where create favorite filter query was not updated after opening modal a few times.

  • Fix communities list widget sending duplicated requests on query change.

  • Ensure searchbar is above search results in item detail sidepanel.

  • Ensure proper sizing of item detail content.

  • Fix issue where the server-level config option pdfconversion.pdf-conversion-timeout was not respected. Use this option to set the maximum number of seconds that a single PDF conversion can take (by default, it is set to 60 seconds).

  • Add dedicated close button for the idetail split view mode.

  • Fix recursive loading in spacy model.

  • Ensure tooltips also trigger from icons in idetail toolbar.

  • Fix greenlet RPM missing from Squirro Platform and throwing error when using Squirro Toolbox CLI (e.g., squirro_asset).

  • Fix the GenAI Studio Plugin embed_token endpoint always generating a grant, even if there is a valid grant for the user.

  • Add tooltips to idetail source title and typography display.

  • Never request entities for items collection.

  • Remove entities display inside of the items cards (leftover from the two-column mode).

  • Resolve an issue causing sessions to not be marked as deleted when an access token was not provided.

  • Fix entity highlights not working for non-pdf.

  • Remove tab indicator transition in item detail.

  • Fix expiring Flask session every time access token expires.

  • Fix not being able to log in to Squirro Platform frontend app when using a refresh_token in the URL.

  • Fix community 360 not showing navigation when opening in a new tab.

  • Sometimes we would run into cases where the user information cannot be fetched from the GenAI studio plugin. Instead of breaking, we will now log the error and continue without user information.

  • Fix raise_404 invocation in the SingleProjectExportResource.GET endpoint.

  • Fix item detail tooltips Chicago style.

  • Fix NOT query facets being shown in the labels widget as selection.

  • Fix AI Studio Ground Truth Labeling not loading.

Breaking Changes#

  • The SquirroGPT custom widget will no longer be supported by the GenAI Studio Plugin. To get the full functionality, please change the relevant dashboards to use the Chat widget that is shipped with the platform.

Installation and Upgrade#

For new installations, find step-by-step instructions in Install and Manage Squirro with Ansible (recommended) or Installing Squirro on Linux.

To upgrade an existing installation, see Upgrading Squirro.