Private Cloud v2.3.1 — Release Notes

warning

The current version of this product is exclusively supported for deployment on GCP Native support for AWS is in development and is targeted for in v2.3.2 shortly.

Major Improvements

• Financial Data:
  • M&A Screener, Company Fundamentals, Funding Screener, and Canalyst Models bring financial data capabilities.
• Snippet Explorer:
  • Snippet exploration and discovery tool for finding and working with relevant content excerpts across your transcripts has been redeployed.
• Channel Checks
  • Updated Channel Checks templates to add all 170+ Channel Checks for users with access
• Company Profile
  • Launched new company profile experience with structured data into this view(M&A history, Company Fundamentals, Funding data)
  • Company specific Workflow Agents like our Company Primer agent
  • Improved access to Industry Comps
• GenSearch as the Homepage
  • The default landing page for AlphaSense interaction is now our Generative Search. Dive directly into GenAI analysis!

Additional Improvements

• Update reference architecture to improve auto-scaling for search, image processing, and AI embedding services.
• Improved compute resources scaling in order to handle load better.
• Redis migration via Crossplane across 30+ services
• ScyllaDB manager upgrade to 3.2.8, enabling seemless upgrade of scylladb data nodes in next release
• Pod-level rate limiting for content processing.
• Improved monitoring with mini-eos support in GCP.
• Search Summary Page with enhanced modules and layout.
• Ask in Doc / Multi Ask in Doc — ask questions directly about document content.
• Fast GenSearch model upgrade for latency improvements
• New Notification Delivery Service.
• Drop the need for gpu-operator in GCP, use native cloud provider nvidia-device-plugin
• Ticker indexer will sync entity metadata within 15min interval

Bug Fixes / Resolved Issues

• Incorrect deep research answer link via email fixed.
• Fixed Solr router NPE causing UI results mismatch, intermittent docSearch failures, cross-site query failures due to undefined schema fields, searchById fix for external users, frontend doc list missing.
• Hybrid-search slowness resolved.
• Selecting additional analyst no longer overrides previous selections.
• Document export email source configuration.
• Misconfigured liveness probe in content-enrichment service resolved.
• GICS-industry data.
• Team tags saving.
• Onenote connector timeout in GCP.

Upgrade note

• GCP: 2 redis instances in integrations namespaces upgraded to use transit encryption:

  • for integration-oauth-service--integrations and recrawler-service--integrations
  • run kubectl get instance.redis | grep False during the upgrade to figure out
  • go to GCP console and delete the Falses redis instances
  • wait for new redis instance created sucessfully
  • run kubectl -n integrations rollout restart deploy recrawler-service integration-transformer integration-oauth-service integration-identity-sync integration-downloader-medium integration-downloader-low integration-downloader-high

• GCP: 3 subscriptions has immutable filter change, need manual deletion then crossplane will recreate:

  • subscriptions:
    • feature-storage-document-vectors-v1
    • vector-indexer-delete-v1
    • image-search-caption-userdocs
  • go to GCP console and delete those subscriptions during the upgrade
  • wait for crossplane to automatically recreate the subscriptions

• Frontend feature flags:

  • kubectl -n applications edit configmap frontend-features-overrides and change the data

    features-overrides.json: |-
    {
      "genSearchV3": true,
      "assistantDocList": true,
      "genSearchMultiAskInDoc": true,
      "genSearchSourceFilter": true,
      "genSearchWatchlist": true,
      "descriptiveCitation": true,
      "genSearchDocSelector": true,
      "askInDocPrompts": true,
      "genSearchDeepResearch": true,
      "citationController": true,
      "genSearchHome": true,
      "focusMode": true,
      "workspaces": true,
      "persistentHistoryPanel": true,
      "gqlSubscription": true,
      "knowledgeWorkflow": true,
      "enforceThinkLonger": true,
      "enableUnlimitedRowsInGenGrid": true,
      "genSearchDeepResearch": true,
      "search2": true,
      "docViewerMicroApp": true,
      "allowBrokerDocuments": true,
      "impersonationV2": true,
      "enableChannelCheckPhase2": true,
      "CompanyProfileV2": true,
      "CompanyProfileV3": true,
      "CoProAIWorkflow": true,
      "CoProMandAData": true,
      "CoProIntradayPricing": true,
      "CoProShareholders": true,
      "CoProWorkforcePage": true,
      "verticalFilters": true,
      "enableSlideGeneration": true,
      "deepResearchSlideGeneration": true,
      "onboarding-genai": true,
      "disableFinancials": true,
      "CoProIntradayPricing": true,
      "gridSummarisation": true,
      "streamAid": true,
      "workflow-agents": true
    }

  • then restart frontend pods with kubectl -n applications rollout restart deploy frontend

Known errors

• Several connectors not included in this release: Web Clipper, SharePoint, OneDrive, Notion, Google Drive, Evernote, EmailedIn, Egnyte, DropBox, Box, Amazon S3
• Search doesn't return bookmarked docs or manually shared user docs with corresponding filters
• Precompiled Nvidia driver is not enabled for gpu-operator in AWS, outbound connection to archive.ubuntu.com (download linux kernel headers) is required for AWS
• The rms-bulk-activity-service may intermittently indicate liveness probe failures, even though the service is functioning as expected. This behavior is attributed to the current liveness probe configuration and does not reflect an actual service disruption.
• saved-searches-user-update subscription in GCP may accumulate messages, but this is not expected to impact functionality
  • will be included to v2.3.2 hotfix shortly
• Sentiment score calculating for user documents is known to be slow.