stack-583 Release Notes

Release Date: April 5, 2026

Summary

Stack-583 advances two major strategic investments: maturing the Synapse Search experience and expanding the Curator / RecordSet data submission tooling. On the search side, the upgraded Search V2 is now the default experience for all users — the experimental flag has been removed and legacy search code retired. On the infrastructure side, new platform APIs have been added to support configurable text analyzers and per-column analyzer overrides, laying the groundwork for richer, more accurate full-text search across datasets. Curator users benefit from several data quality and usability fixes: array-type fields in CSV uploads now auto-coerce the same way the UI does, multi-value columns load correctly in the grid editor, and the inline cell editor now expands so long values can be read and edited without scrolling.

This release also resolves two infrastructure reliability issues: a high CPU spike on the table database cluster from the prior stack, and an unrelated authorization error that prevented some users from retrieving RecordSet validation results.

11 tickets completed across PLFM (8) and SWC (3).

Affected Users & Systems

Deprecated or Breaking Changes

Search V2 is now the only search experience. The legacy search presenter/place code has been removed from both synapse-web-monorepo and SynapseWebClient. The SearchV2 feature flag no longer exists. Any internal tooling or configuration that referenced this flag should be updated. There is no user-facing migration required — the new search has been in experimental mode for some time and is functionally equivalent or better. (SWC-7752)

New Features

Search V2 — Now Generally Available (SWC-7752)

The improved Synapse search experience is now enabled for all users by default. Experimental mode scaffolding has been fully retired, and the underlying legacy code removed. Users will notice a cleaner, more capable search interface across the platform.

Configurable Text Analyzers (PLFM-9509)

A new platform API allows Sage admins and organization managers to define and manage Text Analyzers — configurations that control how text fields are tokenized and analyzed for search. Six built-in system analyzers are now bootstrapped automatically at startup: SCIENTIFIC, STANDARD, IDENTIFIER, KEYWORD, AUTOCOMPLETE, and AUTOCOMPLETE_SEARCH. Custom analyzers can be created starting at ID 1000+. This is a foundational building block for improved search relevance across portals.

Per-Column Analyzer Overrides (PLFM-9511)

Sage admins and organization managers can now specify which Text Analyzer to use for specific columns in a dataset, overriding system defaults. This enables fine-grained control over how individual fields (e.g., identifiers vs. free-text descriptions) are indexed and searched. Read access is public; write access is restricted to Sage employees and org admins.

Curation Task v2 API (PLFM-9508)

A new version of the curation task API has been implemented on the platform, enabling more structured and extensible workflows for data curation tasks.

Fixes & Improvements

Download Cart — Confirmation Step Before Adding Files (SWC-7741)

When adding a large number of files to the Download Cart from search results (especially from folders, projects, or entity views), users now see a confirmation banner showing a count of files before the files are actually added. This prevents accidental bulk additions.

Download Cart — Refreshed UI (SWC-7757)

The Download List interface has been updated to match the latest design specifications, improving visual consistency and usability.

Curator Grid — Improved Inline Editor (SWC-7625)

The inline cell editor in the Curator grid now expands when selected, allowing users to view and edit the full content of a cell without scrolling. Previously, large cell values were difficult to work with.

CSV Upload: Array Fields Now Auto-Coerce Plain Strings (PLFM-9558)

When uploading a CSV to a RecordSet, fields defined as array types (e.g., HTAN_PARENT_ID) no longer require values to be manually wrapped in JSON array syntax (["value"]). Plain string values are now automatically handled, matching the behavior of the Curator UI. This removes a significant friction point for bulk data contributors.

RecordSet Validation Results — Authorization Error Fixed (PLFM-9563)

Users who encountered a SynapseAuthorizationError: You are not authorized to access fileHandleId error when retrieving validation results for a RecordSet will no longer see this issue. The underlying authorization check has been corrected.

Array Data Parsing Fixed in Curator Grid Sessions (PLFM-9357)

Multi-value array columns (e.g., DataUseModifiers) were loading incorrectly when creating a Curator grid session from a RecordSet — values like ["GRU", "COL"] were being displayed as ["[GRU, COL]"]. This has been resolved.

Migration Job Timeout Fixed (PLFM-9541)

An incorrect timeout setting for CodeBuild migration jobs was causing migrations to fail prematurely (after 10 hours instead of the intended 36 hours). The correct timeout has now been applied.

High CPU on Table Cluster Resolved (PLFM-9567)

A high CPU utilization event on the production table cluster (carried over from stack-582) has been investigated and resolved.

CloudWatch Cost Spike — Investigation Complete (PLFM-9565)

A 3× spike in AWS CloudWatch costs in March 2026 (from ~$644/month to ~$2,428) was traced to high-cardinality custom metrics being published by the logging service. The root cause has been identified and remediation is underway, including audit and optimization of custom metric publishing patterns.

For questions about this release, contact the Synapse Platform team.