What get-convex/convex-backend shipped
Written by FoxPlug from the repository's own releases; not affiliated with Convex. Paste your own repository in the box below and FoxPlug writes the same weekly update for your product. This page is an automatic summary of the releases, pull requests and commits in github.com/get-convex/convex-backend; Convex did not write it and does not use or endorse FoxPlug. Every line links to the change it describes.
Get a weekly update like this for your product, free
Week of November 2, 2026
What shipped
- S3 export tables now keep only their last 10 snapshots, so metadata.json no longer lists every snapshot since the export started. Release
- Persistence index scans are now attributed to their caller and index kind. Release
- Index cache intervals are now registered before inserting the entry, with registration undone when the entry already exists, and a concurrent population test was added. Release
- Added a gzip ConfigDecoder. Release
- Included usage from the ledger is now shown when it is not in the current billing period. Release
- The usage page now estimates monthly spend for active deployments. Release
- MetricPrice generation now leaves out allocations. Release
Why it matters
S3 exports stop accumulating every snapshot since the export began, so export metadata stays bounded instead of growing and being rewritten on each commit. Persistence index scan attribution and safer index cache registration give clearer measurements and fewer concurrency problems. The billing and usage pages gain clearer figures for included usage and estimated monthly spend.
Changelog entry
- S3 export tables keep their last 10 snapshots instead of listing every snapshot since the export started. Release
- Persistence index scans are attributed to their caller and index kind. Release
- Index cache intervals are registered before inserting the entry, with registration undone when an entry already exists, plus a concurrent population test. Release
- Added a gzip ConfigDecoder. Release
- Included usage from the ledger is shown when it is not in the current billing period. Release
- The usage page estimates monthly spend for active deployments. Release
- MetricPrice generation no longer includes allocations. Release
This week: S3 export tables keep only their last 10 snapshots, persistence index scans are attributed to caller and index kind, index cache intervals register before insert, gzip ConfigDecoder added, and the usage page shows included usage from the ledger plus estimated monthly s
This week's changes: S3 export tables now keep their last 10 snapshots, so metadata.json no longer lists every snapshot since the export started and each commit reads and rewrites less. Persistence index scans are attributed to their caller and index kind. Index cache intervals are registered before inserting the entry, with the registration undone when an entry already exists, and a concurrent population test checks every registered interval. A gzip ConfigDecoder was added. The usage page now estimates monthly spend for active deployments and shows included usage from the ledger when it is not in the current billing period. MetricPrice generation no longer includes allocations.
Week of October 19, 2026
What shipped
- Added a flag to distrust the compromised eMudhra certificates in Node. Commit
- Fixed the Brotli mitigation on Node.js 22 in the node executor. Commit
- The S3 export worker no longer holds a database transaction open while a page syncs. Release
- S3 export now reports Parquet bytes as egress through
bytes_written. Release - Raised the batch size for data sync to avoid unnecessary work. Release
- AWS credentials are now cached process-wide. Commit
- The dashboard shows the card hold note when a card is pre-authorized. Commit
- Published npm release version 1.48.0. Release
- Reapplied the eMudhra certificate distrust after the earlier revert. Commit
Why it matters
This week's changes affect sync and export behavior on busy deployments and tighten how compromised certificates are handled in Node. The S3 export transaction change and the egress reporting make large exports more predictable to run and measure.
Changelog entry
- Add a flag to distrust the compromised eMudhra certificates in Node. Commit
- Node-executor: fix the Brotli mitigation on Node.js 22. Commit
- Don't hold the S3 export checkpoint transaction across a page. Release
- Report Parquet bytes as S3 export egress. Release
- Up the batch size for data sync to avoid unnecessary work. Release
- Cache AWS credentials process-wide. Commit
- Dashboard: show the card hold note when card pre-authorizes. Commit
- Npm release version 1.48.0. Release
- Reapply the eMudhra certificate distrust. Commit
- Docs: update the AI Gateway model list. Release
This week: a flag to distrust compromised eMudhra certificates in Node, a Node.js 22 Brotli mitigation fix, S3 export no longer holding a transaction across a page, Parquet bytes reported as egress, larger data sync batches, and npm 1.48.0.
This week's changes: a flag to distrust the compromised eMudhra certificates in Node, with the distrust reapplied after its revert; a fix for the Brotli mitigation on Node.js 22 in the node executor. The S3 export worker no longer holds a database transaction open while a page syncs, and S3 export reports Parquet bytes as egress. Data sync uses a larger batch size, AWS credentials are cached process-wide, and npm release 1.48.0 is out.
Week of October 12, 2026
What shipped
- Precompiled release 2026-10-10-72ce13c shipped, with usage tracking now able to aggregate database bandwidth across requests. Release
- Deployment JWT max lifetime lowered to 1 hour. Commit
- The jti claim is now reserved in deployment JWTs. Commit
- Pagination can now be repeated within mutations. Commit
- The ai-files install and update CLI commands now exit 1 when they fail. Commit
- Verbose push output now redacts environment variable values. Commit
- Bundle analyzer gained an explore view. Commit
- Bundle analyzer gained insights. Commit
- Bundle analyzer now shows package and file details. Commit
- IO errors from mysql are now classified. Commit
Why it matters
This week tightened deployment token security with a shorter JWT lifetime and a reserved claim, and fixed pagination so it can be repeated in mutations. CLI output is safer because env var values are redacted in verbose push output, and ai-files install and update now fail with a nonzero exit code. The bundle analyzer also gained explore, insights, and package and file details views.
Changelog entry
- Released precompiled 2026-10-10-72ce13c; database bandwidth usage events can now omit
request_idso summaries can span multiple requests. Release - Lowered the deployment JWT max lifetime to 1 hour. Commit
- Reserved the jti claim in deployment JWTs. Commit
- Allowed repeated pagination in mutations. Commit
- Ai-files install and update now exit 1 on failure. Commit
- Verbose push output now redacts environment variable values. Commit
- Bundle analyzer: added explore. Commit
- Bundle analyzer: added insights. Commit
- Bundle analyzer: added package and file details. Commit
- Bundle analyzer: added modules. Commit
- Classified IO errors from mysql. Commit
- Reserved the TokenMatch.distance proto field. Commit
New release 2026-10-10-72ce13c. Deployment JWT lifetime is now capped at 1 hour, jti is reserved in deployment JWTs, and database bandwidth usage can be aggregated. Pagination can repeat in mutations, verbose push output redacts env var values, and the bundle analyzer has explore
A new precompiled release, 2026-10-10-72ce13c, is out. Deployment JWT lifetime is now capped at 1 hour and the jti claim is reserved in deployment JWTs. Pagination can be repeated inside mutations, and the ai-files install and update commands now exit 1 on failure. Verbose push output redacts environment variable values. Usage tracking can aggregate database bandwidth events without a request id. The bundle analyzer adds explore, insights, and package and file details. IO errors from mysql are now classified.
Week of October 5, 2026
What shipped
- The database bandwidth aggregate now allows events that summarize multiple requests to omit the request id, while individual execution events keep it, and a redundant empty-batch check in audit-log delivery was removed. Release
- The dashboard health page now shows schema validation and index backfill progress, with a line for each job while a deploy is validating a schema or building indexes. Release
- A missing Glue table is now treated as missing when loaded lazily, fixing S3 exports that failed with EntityNotFoundException when a sync page touched a table not yet in Glue. Release
- The ai-sdk-provider package was released at 0.3.0 for the stable release of voice support and opt-in ZDR routing. Release
- Transcription failures in the ai-sdk-provider now use APICallError so transient HTTP errors honor maxRetries, with status and response headers preserved for retry handling. Release
- The S3 integration was renamed to AWS S3 and its setup steps were collapsed so only one of the Destination, AWS access and Frequency steps is open at a time. Release
- The logs page gained a search query argument so pages with filter params and component, function or level selection can be linked to directly. Release
- Metrics were added to the S3 export worker. Release
- The documentation on function bundle size limits was updated. Release
- The documentation AI Gateway model list was updated. Release
Why it matters
This week's changes touch core data paths and the deploy experience. Aggregate database bandwidth events can now omit the request id, and S3 exports no longer fail when a sync page meets a Glue table that does not exist yet. The dashboard health page also shows schema validation and index backfill progress, so a deploy that is still validating or backfilling is visible. The ai-sdk-provider reached version 0.3.0 for stable voice support and opt-in ZDR routing, and transcription failures now retry on transient HTTP errors.
Changelog entry
- Database bandwidth aggregate events may omit the request id; individual execution events retain it, and the redundant empty-batch check in audit-log delivery was removed. Release
- Dashboard health page shows schema validation and index backfill progress per job while a deploy is in progress. Release
- Missing Glue tables are treated as missing when loaded lazily, fixing S3 export failures with EntityNotFoundException. Release
- Released ai-sdk-provider 0.3.0 for stable voice support and opt-in ZDR routing. Release
- Ai-sdk-provider retries transient transcription failures using APICallError with status and headers preserved. Release
- Renamed the S3 integration to AWS S3 and collapsed its setup steps so one is open at a time. Release
- Added a search query argument to the logs page for linking to filter params and component, function or level selection. Release
- Added metrics to the S3 export worker. Release
- Updated docs for function bundle size limits. Release
- Updated the AI Gateway model list in the docs. Release
This week: aggregate database bandwidth events can omit the request id, missing Glue tables no longer break S3 exports, and the dashboard health page shows schema validation and index backfill progress during deploys. The ai-sdk-provider hit 0.3.0 with stable voice support and op
Here is what shipped this week. The database bandwidth aggregate now allows events covering multiple requests to omit the request id, while individual execution events keep it. S3 exports no longer fail when a sync page touches a Glue table that is not there yet, because a missing table is treated as missing when loaded lazily. The dashboard health page now shows schema validation and index backfill progress, with a line per job while a deploy is validating a schema or building indexes. The ai-sdk-provider reached 0.3.0 for the stable release of voice support and opt-in ZDR routing, and transcription failures now use APICallError so transient HTTP errors honor maxRetries. The logs page gained a search query argument for linking to filtered views, the S3 integration was renamed to AWS S3 wi
Week of September 28, 2026
What shipped
- Added support for two new deployment regions: aws-ap-southeast-2 (Sydney) and aws-ca-central-1 (Montreal), available for deploymentRegion, defaultRegion, and
list_deployment_regions. Release - WebAssembly OutOfMemory errors now labeled by phase (importing vs executing) for better debugging. Release
- Index backfill writes now retry on MySQL deadlocks to improve reliability during index operations. Release
- Request and Response implementations improved for better spec compliance. Release
- URLSearchParams initialization now reads @iterator only once, reducing redundant operations. Release
- Npm dependencies updated to latest versions. Release
- Fixed log streaming disable transaction and flag ordering. Release
- Documentation typos fixed across 14 pages, correcting doubled words and missing/incorrect terms. Release
Why it matters
This week brings expanded regional availability for deployments and improvements to reliability and correctness across WebAssembly execution, database operations, and web standards compliance. Fixes to error labeling and documentation make the platform easier to debug and understand.
Changelog entry
- Deployment regions expanded: aws-ap-southeast-2 (Sydney) and aws-ca-central-1 (Montreal) now supported Release
- WebAssembly OutOfMemory errors now labeled by phase (importing vs executing) Release
- Index backfill writes retry on MySQL deadlocks Release
- Request/Response spec compliance improved Release
- URLSearchParams initialization optimized to read @iterator only once Release
- Npm dependencies updated Release
- Log streaming disable transaction and flag ordering fixed Release
- Documentation typos corrected across 14 pages Release
New regions live: Sydney (ap-southeast-2) and Montreal (ca-central-1). Plus improvements to WASM error reporting, index backfill reliability, and web spec compliance.
This week we expanded Convex's regional footprint with Sydney and Montreal now available for deployments. We also shipped reliability improvements to index backfill operations with MySQL deadlock retries, better WebAssembly error labeling to distinguish import vs execution failures, and enhanced Request/Response spec compliance. These changes support better debugging and more robust database operations for production deployments.
Week of September 21, 2026
What shipped
- Added a 1.0/storageStore async syscall for storage operations. Release
- Fixed document revision replay after interrupted migration index scans to reconcile V6 deletion markers. Release
- Skipped index retention worker for v6 persistence to prevent deployments from loading below retained log buckets. Release
- Improved async stack trace formatting in error reporting. Release
- Fixed URLSearchParams.set() to replace the first match in place. Release
- Added validation to refuse team handovers that would empty the team or remove the caller. Release
- Added Canada and Australia regions to the dashboard behind feature flags. Release
- Fixed 404 errors after creating a deployment by awaiting revalidation in the dashboard. Release
- Clipped the globe visualization to match the rounded corners of the provision deployment card. Release
- Added coupon and plan migration tooling to big-brain-tool. Release
Why it matters
This week focused on stability improvements for v6 persistence migrations and storage operations. Fixes to index scanning recovery and error reporting improve reliability for deployments managing large datasets. New regional options expand deployment choices.
Changelog entry
- Add a 1.0/storageStore async syscall Release
- Replay document revisions after interrupted migration index scans Release
- Skip the index retention worker for v6 persistence Release
- Improve async stack trace formatting in error reporting Release
- Make URLSearchParams.set() replace the first match in place Release
- Refuse team handovers that empty the team or remove the caller Release
- Add Canada and Australia regions to the dashboard Release
- Fix 404 after creating a deployment by awaiting revalidation Release
- Clip the globe to the provision deployment card's rounded corners Release
- Add a coupon and plan migration tool to big-brain-tool Release
This week: improved async error reporting, fixed URLSearchParams behavior, added 1.0 storage syscalls, and resolved v6 migration index scanning issues. Canada and Australia region support now available.
Convex updates this week address database reliability and developer experience. We fixed async stack trace formatting for better error visibility, resolved v6 persistence migration edge cases, and implemented the 1.0/storageStore syscall. The dashboard now supports Canada and Australia regions and improved validation for team operations. These changes strengthen deployment stability and expand regional availability.
More from get-convex/convex-backend on GitHub
Releases
- precompiled-2026-09-26-27ef234 (Precompiled 2026-09-26-27ef234), a pre-release published September 26, 2026. (source)
- precompiled-2026-09-26-7c0c8a4 (Precompiled 2026-09-26-7c0c8a4), a pre-release published September 26, 2026. (source)
- precompiled-2026-09-25-72fad67 (Precompiled 2026-09-25-72fad67), a pre-release published September 25, 2026. (source)
- precompiled-2026-09-25-278951b (Precompiled 2026-09-25-278951b), a pre-release published September 25, 2026. (source)
- precompiled-2026-09-25-f7d96aa (Precompiled 2026-09-25-f7d96aa), a pre-release published September 25, 2026. (source)
- precompiled-2026-09-24-d1770b1 (Precompiled 2026-09-24-d1770b1), a pre-release published September 24, 2026. (source)