Skip to content

chore: release v0.10.0 - #80

Open
sw-release-bot[bot] wants to merge 2 commits into
mainfrom
release-plz-2026-10-25T11-34-55Z
Open

sw-release-bot[bot] wants to merge 2 commits into
mainfrom
release-plz-2026-10-25T11-34-55Z

Conversation

@sw-release-bot

@sw-release-bot sw-release-bot Bot commented Oct 25, 2026 •

Copy link
Copy Markdown
Contributor

New release

coordinode 0.9.0 -> 0.10.0 (all 22 workspace crates share this version)

Added

  • (schema) declare type, implied and auxiliary labels
  • (cluster) replace a member whose vote no copy proves
  • (storage) collect orphan blobs and return their space
  • (storage) give blobs owners, a catalog and a guarded lifetime
  • (storage) per-table pressure series that go with the table
  • (vector) exact scores and domain errors end to end
  • (storage) hold the database to a strict total footprint limit
  • (query) change a node's type with RETYPE NODE
  • (storage) judge, repair and drop columnar tables as partitions
  • (embed) physical backup of the whole database
  • (vector) score vectors exactly in binary64 with one reduction
  • (operation) apply each client write once
  • (storage) store index generations under local installations
  • (migrate) add the node sizes report module
  • (migrate) catalog record migrations, check and journal stats
  • (migrate) offline tool to bring a data directory current

Documentation

  • (cluster) describe a replacement refused by a changed group
  • (query) drop the type-keyed index reason from relabel
  • (storage) name a private helper without linking it

Fixed

  • (query) keep rows without a vector in a vector ordering
  • (storage) refuse a hard limit below the supported minimum
  • (bench) read git state without optional locks
  • (query) charge exact dedup memory before allocating it
  • (query) defer only the retyped nodes' index entries in RETYPE
  • (embed) outwait the commits in flight before re-running a statement
  • (vector) draw each HNSW level once and finite
  • (raft) refuse a member without its vote with a typed error
  • (storage) serve a rebuilt tree's history above its copy's floor
  • (storage) refuse a rebuild from a damaged checkpoint copy
  • (storage) rebuild a partition at its original seqnos
  • (storage) rebuild a table only from provable state
  • (embed) finish an interrupted table rebuild on open
  • (query) plan subqueries with the index access paths
  • (query) order by variables the projection drops
  • (storage) capture the journal tail after a checkpoint's trees
  • (query) subqueries see the outer row; one index search per top-k
  • (query) keep each edge's orientation in undirected hops
  • (replication) wait for the leader's address, not pauses
  • (query) three-valued AND/OR, simple CASE, correlated keys
  • (cluster) take a replacement's membership only once committed
  • (cluster) check both members of a replacement in one view
  • (cluster) fence a replacement on the membership it checked
  • (cluster) apply each change to a joint config left behind
  • (cluster) start every membership change from a single config
  • (cluster) resume a change left in its joint configuration
  • (cluster) refuse a joiner on another member's store as a join
  • (raft) bind promises to a member, write copy durably
  • (raft) keep the purge point and a loss with the vote; replace a member
  • (raft) keep a second copy of the member's vote
  • (storage) decide a schema read from the current state
  • (storage) save the consensus state below half the space reserve
  • (storage) name the key a pending-commit conflict is about
  • (storage) close a dropped table's tree before removing its folder
  • (storage) admit promise copy writes as recovery work
  • (storage) let blob collection run below the free-space reserve
  • (storage) flush and capture a table only while it lives
  • (storage) no table file is made in a folder being dropped
  • (storage) a second drop of a table drops the same table
  • (storage) hold a table name until its drop finishes
  • (storage) open a checkpoint on its clock, never compacted
  • (storage) roll back a torn manifest tail instead of repairing
  • (storage) sync a new journal segment's directory
  • (vector) keep every HNSW node reachable when lists fill
  • (vector) keep cosine exact for every finite vector
  • (storage) never copy a checkpoint over a file it holds
  • (migrate) find every marked store, Raft snapshot captures too

Performance

  • (query) read a node's schemas once per write
  • (query) hashed exact DISTINCT, grouping and UNION
  • (query) build only the node properties a statement reads
  • (bench) declare the label the label-scan bench adds
  • (bench) measure label kinds on the write path
  • (query) look the keys of a correlated index join up in batches
  • (query) look a correlated index key up once per distinct key
  • (query) read a property column without allocating
  • index partition trees by discriminant, share index definitions
  • (storage) leave a journaled tree's flush syncs to a barrier
  • (storage) export a generation's history by its own key range
  • (storage) bench flush and compaction of a table
  • (vector) reduce four AVX2 blocks vertically
  • (vector) score exactly only candidates that can rank
  • (vector) no per-evaluation clear of the reduction stack
  • (vector) SSE2 exact kernels and cached sums of squares
  • (query) bound property delta chains with whole writes

Refactored

  • (schema) store canonical labels and judge RETYPE by its result
  • (storage) keep the table repair record crate-private

Testing

  • (embed) cover expiry, merge, restore and failed declarations for label kinds
  • (embed) keep an operation's outcome across a replacement
  • (embed) accept a cascade the capacity scanner ran first
  • (embed) report a table rebuild the open cannot finish
  • (embed) refuse an unevaluable query vector in every top-k form
  • (embed) measure the index top-k with the vector not returned
  • (embed) measure the correlated join one row per key and as a search
  • (embed) wait for reports and rebuilds before index checks
  • (embed) cover the fallbacks of the batched correlated join
  • (embed) wait for the recorded suspicion before a failing check
  • (embed) measure a correlated join through a unique index
  • (embed) count allocations per vector statement
  • (embed) bench the index-served top-k shape, store beside the build
  • (embed) measure vector statements end to end
  • (embed) keep the cascade's data on hot until it moves it
  • (embed) lower a full group to the supported minimum
  • (embed) lower the limit below the data, not the transient charge
  • (query) tell a woken snapshot read from a timed-out one by margin
  • (query) measure the distributed top-k merge
  • (raft) let the refused bound lie above the read one
  • (raft) finish a conversion cut short at the marker
  • (raft) count journal syncs for an unchanged vote
  • (raft) catch an unchanged vote appended to the journal
  • (raft) resume a replacement on another leader, refuse old votes
  • (raft) run the lost-answer case around the actual leader
  • (storage) fail every matching op of a persist
  • (storage) fail only the persisting partition's files
  • (storage) wait for the drop to close the tree, not for a sleep
  • (storage) measure the copy's peak from a steady store
  • (storage) model clean-stale pages and torn files
  • (storage) fail each step of a promise copy write
  • (storage) compare table tree paths absolute, as the tree config does
  • (storage) compare table tree paths against the endpoints as configured
  • (storage) hold a slowed table's rows to the write rate
  • (storage) judge each table into its own verdict and one metrics series
  • (storage) check the tail-captured segment's bytes on every platform
  • (vector) choose the exact kernel set for a bench run
  • (vector) measure the production exhaustive scan, not the unused flat index
  • (vector) qualify exact vector evaluation before and after

Bench

  • (raft) observe the copy's footprint inside a write
  • (raft) wait for a leader that confirms its lease
  • (raft) give the recovery image a promised term above 1
  • (raft) measure recovery, the copy's peak footprint and reads
  • (raft) trace each commit against the device's I/O
  • (raft) separate leadership seen from confirmed
  • (raft) bound each promise cost to its own span
  • (raft) measure what keeping consensus promises costs
  • (storage) measure statement flushes on a fresh store per series

Where the changes landed

The list above is the folded view. Each crate below carries the entries written
to its own changelog; a crate with no commits since the last release is absent.

coordinode-core

Added

  • (schema) declare type, implied and auxiliary labels
  • (storage) collect orphan blobs and return their space
  • (storage) give blobs owners, a catalog and a guarded lifetime
  • (storage) hold the database to a strict total footprint limit

Fixed

  • (query) charge exact dedup memory before allocating it

Performance

  • (query) read a node's schemas once per write
  • (query) hashed exact DISTINCT, grouping and UNION

Refactored

  • (schema) store canonical labels and judge RETYPE by its result
coordinode-storage

Added

  • (schema) declare type, implied and auxiliary labels
  • (storage) collect orphan blobs and return their space
  • (storage) give blobs owners, a catalog and a guarded lifetime
  • (storage) per-table pressure series that go with the table
  • (storage) judge, repair and drop columnar tables as partitions
  • (storage) hold the database to a strict total footprint limit

Documentation

  • (storage) name a private helper without linking it

Fixed

  • (storage) decide a schema read from the current state
  • (storage) save the consensus state below half the space reserve
  • (storage) name the key a pending-commit conflict is about
  • (storage) close a dropped table's tree before removing its folder
  • (storage) admit promise copy writes as recovery work
  • (raft) bind promises to a member, write copy durably
  • (raft) keep the purge point and a loss with the vote; replace a member
  • (raft) keep a second copy of the member's vote
  • (storage) let blob collection run below the free-space reserve
  • (storage) flush and capture a table only while it lives
  • (storage) no table file is made in a folder being dropped
  • (storage) a second drop of a table drops the same table
  • (storage) hold a table name until its drop finishes
  • (storage) serve a rebuilt tree's history above its copy's floor
  • (storage) open a checkpoint on its clock, never compacted
  • (storage) refuse a rebuild from a damaged checkpoint copy
  • (storage) rebuild a partition at its original seqnos
  • (storage) rebuild a table only from provable state
  • (storage) roll back a torn manifest tail instead of repairing
  • (storage) capture the journal tail after a checkpoint's trees
  • (storage) refuse a hard limit below the supported minimum
  • (storage) sync a new journal segment's directory

Performance

  • (query) read a node's schemas once per write
  • (storage) leave a journaled tree's flush syncs to a barrier
  • (storage) export a generation's history by its own key range
  • (storage) bench flush and compaction of a table
  • index partition trees by discriminant, share index definitions

Refactored

  • (schema) store canonical labels and judge RETYPE by its result
  • (storage) keep the table repair record crate-private

Testing

  • (storage) fail every matching op of a persist
  • (storage) fail only the persisting partition's files
  • (storage) wait for the drop to close the tree, not for a sleep
  • (storage) measure the copy's peak from a steady store
  • (storage) model clean-stale pages and torn files
  • (storage) fail each step of a promise copy write
  • (storage) compare table tree paths absolute, as the tree config does
  • (storage) compare table tree paths against the endpoints as configured
  • (storage) hold a slowed table's rows to the write rate
  • (storage) judge each table into its own verdict and one metrics series
  • (storage) check the tail-captured segment's bytes on every platform

Bench

  • (storage) measure statement flushes on a fresh store per series
  • (raft) separate leadership seen from confirmed
coordinode-cluster

Added

  • (storage) hold the database to a strict total footprint limit
coordinode-vector

Added

  • (vector) exact scores and domain errors end to end
  • (vector) score vectors exactly in binary64 with one reduction
  • (storage) hold the database to a strict total footprint limit

Fixed

  • (vector) draw each HNSW level once and finite
  • (vector) keep every HNSW node reachable when lists fill
  • (vector) keep cosine exact for every finite vector

Performance

  • (vector) reduce four AVX2 blocks vertically
  • (vector) score exactly only candidates that can rank
  • (vector) no per-evaluation clear of the reduction stack
  • (vector) SSE2 exact kernels and cached sums of squares

Testing

  • (vector) choose the exact kernel set for a bench run
  • (vector) measure the production exhaustive scan, not the unused flat index
  • (vector) qualify exact vector evaluation before and after
coordinode-modality

Added

  • (storage) collect orphan blobs and return their space
  • (storage) give blobs owners, a catalog and a guarded lifetime
  • (vector) exact scores and domain errors end to end
  • (storage) hold the database to a strict total footprint limit

Performance

  • (query) look the keys of a correlated index join up in batches
coordinode-search

Added

  • (storage) hold the database to a strict total footprint limit
coordinode-query

Added

  • (query) change a node's type with RETYPE NODE
  • (schema) declare type, implied and auxiliary labels
  • (storage) give blobs owners, a catalog and a guarded lifetime
  • (storage) judge, repair and drop columnar tables as partitions
  • (vector) exact scores and domain errors end to end
  • (storage) hold the database to a strict total footprint limit

Documentation

  • (query) drop the type-keyed index reason from relabel

Fixed

  • (query) defer only the retyped nodes' index entries in RETYPE
  • (query) plan subqueries with the index access paths
  • (query) keep rows without a vector in a vector ordering
  • (query) order by variables the projection drops
  • (query) subqueries see the outer row; one index search per top-k
  • (query) keep each edge's orientation in undirected hops
  • (query) three-valued AND/OR, simple CASE, correlated keys
  • (query) charge exact dedup memory before allocating it

Performance

  • (query) build only the node properties a statement reads
  • (query) read a property column without allocating
  • (query) read a node's schemas once per write
  • (query) look the keys of a correlated index join up in batches
  • (query) look a correlated index key up once per distinct key
  • index partition trees by discriminant, share index definitions
  • (query) hashed exact DISTINCT, grouping and UNION

Refactored

  • (schema) store canonical labels and judge RETYPE by its result

Testing

  • (query) tell a woken snapshot read from a timed-out one by margin
  • (query) measure the distributed top-k merge
coordinode-wire

Added

  • (storage) hold the database to a strict total footprint limit
coordinode-raft

Added

  • (storage) hold the database to a strict total footprint limit

Fixed

  • (cluster) take a replacement's membership only once committed
  • (cluster) check both members of a replacement in one view
  • (cluster) fence a replacement on the membership it checked
  • (cluster) apply each change to a joint config left behind
  • (cluster) start every membership change from a single config
  • (cluster) resume a change left in its joint configuration
  • (cluster) refuse a joiner on another member's store as a join
  • (raft) bind promises to a member, write copy durably
  • (raft) keep the purge point and a loss with the vote; replace a member
  • (raft) refuse a member without its vote with a typed error
  • (raft) keep a second copy of the member's vote
  • (replication) wait for the leader's address, not pauses

Performance

  • (storage) leave a journaled tree's flush syncs to a barrier

Testing

  • (raft) let the refused bound lie above the read one
  • (raft) finish a conversion cut short at the marker
  • (raft) count journal syncs for an unchanged vote
  • (raft) catch an unchanged vote appended to the journal
  • (raft) resume a replacement on another leader, refuse old votes
  • (raft) run the lost-answer case around the actual leader

Bench

  • (raft) observe the copy's footprint inside a write
  • (raft) wait for a leader that confirms its lease
  • (raft) give the recovery image a promised term above 1
  • (raft) measure recovery, the copy's peak footprint and reads
  • (raft) trace each commit against the device's I/O
  • (raft) separate leadership seen from confirmed
  • (raft) bound each promise cost to its own span
  • (raft) measure what keeping consensus promises costs
coordinode-swarm

Added

  • (storage) hold the database to a strict total footprint limit
coordinode-replicate

Added

  • (storage) hold the database to a strict total footprint limit
coordinode-embed

Added

  • (query) change a node's type with RETYPE NODE
  • (schema) declare type, implied and auxiliary labels
  • (cluster) replace a member whose vote no copy proves
  • (storage) collect orphan blobs and return their space
  • (storage) give blobs owners, a catalog and a guarded lifetime
  • (storage) judge, repair and drop columnar tables as partitions
  • (embed) physical backup of the whole database
  • (vector) exact scores and domain errors end to end
  • (storage) hold the database to a strict total footprint limit

Documentation

  • (cluster) describe a replacement refused by a changed group

Fixed

  • (query) defer only the retyped nodes' index entries in RETYPE
  • (embed) outwait the commits in flight before re-running a statement
  • (vector) draw each HNSW level once and finite
  • (raft) refuse a member without its vote with a typed error
  • (storage) serve a rebuilt tree's history above its copy's floor
  • (storage) refuse a rebuild from a damaged checkpoint copy
  • (storage) rebuild a partition at its original seqnos
  • (storage) rebuild a table only from provable state
  • (embed) finish an interrupted table rebuild on open
  • (query) plan subqueries with the index access paths
  • (query) keep rows without a vector in a vector ordering
  • (query) order by variables the projection drops
  • (storage) capture the journal tail after a checkpoint's trees
  • (query) subqueries see the outer row; one index search per top-k
  • (query) keep each edge's orientation in undirected hops
  • (replication) wait for the leader's address, not pauses
  • (query) three-valued AND/OR, simple CASE, correlated keys
  • (storage) refuse a hard limit below the supported minimum

Performance

  • (query) build only the node properties a statement reads
  • (bench) declare the label the label-scan bench adds
  • (bench) measure label kinds on the write path
  • (query) look the keys of a correlated index join up in batches
  • (query) look a correlated index key up once per distinct key

Refactored

  • (schema) store canonical labels and judge RETYPE by its result

Testing

  • (embed) cover expiry, merge, restore and failed declarations for label kinds
  • (embed) keep an operation's outcome across a replacement
  • (embed) accept a cascade the capacity scanner ran first
  • (embed) report a table rebuild the open cannot finish
  • (embed) refuse an unevaluable query vector in every top-k form
  • (embed) measure the index top-k with the vector not returned
  • (embed) measure the correlated join one row per key and as a search
  • (embed) wait for reports and rebuilds before index checks
  • (embed) cover the fallbacks of the batched correlated join
  • (embed) wait for the recorded suspicion before a failing check
  • (embed) measure a correlated join through a unique index
  • (embed) count allocations per vector statement
  • (embed) bench the index-served top-k shape, store beside the build
  • (embed) measure vector statements end to end
  • (embed) keep the cascade's data on hot until it moves it
  • (embed) lower a full group to the supported minimum
  • (embed) lower the limit below the data, not the transient charge
coordinode-s3

Added

  • (storage) collect orphan blobs and return their space
  • (storage) give blobs owners, a catalog and a guarded lifetime
coordinode-bench

Fixed

  • (bench) read git state without optional locks
coordinode-migrate

Added

  • (operation) apply each client write once
  • (storage) store index generations under local installations
  • (migrate) add the node sizes report module
  • (migrate) catalog record migrations, check and journal stats
  • (migrate) offline tool to bring a data directory current

Fixed

  • (storage) never copy a checkpoint over a file it holds
  • (migrate) find every marked store, Raft snapshot captures too

Performance

  • (query) bound property delta chains with whole writes

This PR was generated with release-plz.

@coderabbitai

coderabbitai Bot commented Oct 25, 2026 •

Copy link
Copy Markdown

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Organization UI
  • Review profile: ASSERTIVE
  • Plan: Advanced
  • Run ID: 967b658c-01ac-49df-ae5a-c1facec29429

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

  • 🔍 Trigger review
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@sw-release-bot
sw-release-bot Bot force-pushed the release-plz-2026-10-25T11-34-55Z branch 27 times, most recently from 06a45e0 to 9531071 Compare October 30, 2026 13:19
@sw-release-bot
sw-release-bot Bot force-pushed the release-plz-2026-10-25T11-34-55Z branch from e499b56 to 578befa Compare October 30, 2026 13:44
@sonarqubecloud

Copy link
Copy Markdown

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants