v0.13.0

Release v0.13.0 prevents taking snapshot from different database and if you already did you can delete the snapshot.

Full Changelog: v0.12.2...v0.13.0

v0.12.2

Full Changelog: v0.12.0...v0.12.2

v0.12.0

DryRun format is now bumped to v2

• capture planner related GUCs
• normalize GUCs for correct replay
• settings aware hash
• storage-param baselines - can now track storage params and GUCs changes (primary for VACUUM)

MCP Improvements

• Input/output schema validation and structured tool results
• Stats-injection tripwires

Compliance/Security

• Read-only connections with timeouts
• CLI Role protection for production connections
• Better panic recovery in handlers

Database flavours

• Detect database flavour
• AlloyDB autovacuum advice and columnar_report MCP tool

Full Changelog: v0.11.1...v0.12.0

v0.11.1

Various bug fixes

Full Changelog: v0.11.0...v0.11.1

v0.11.0

Bloat and vacuum health are now lint findings

Table and index bloat, plus vacuum/freeze health, used to be ad-hoc outputs. They now flow through the lint-findings pipeline as everything else, so they show up with a severity, a stable key, and an explanation.

Other refactorings

• Preparation for #3
• database_id falls back to the project id; current database name is surfaced
• improved the snapshot hashing
• snapshot push is from oldest to newest

Full Changelog: v0.10.0...v0.11.0

v0.10.0

snapshot diff now does really diffing

It's doing real contextual diffing, instead of comparison of JSON. It correlates schema, planner, and activity across two snapshots through a single shared snapshot-ref resolver, and joins planner/activity rows by exact schema_ref so the numbers line up node-to-node.

• rename/identity detection via OID, i.e. renamed table reads as a rename, not a drop + create
• planner sizing diff and column-stats drift, made reset-aware so a stats reset doesn't masquerade as change
• index drift that survives the table filter
• multi-node activity correlation

At the same time old MCP tool schema_diff is gone, replaced by snapshot_diff.

Bloat estimation

Both on heap and indexes.

Remote HTTP Store

Not doing much, but let's say .... boringSQL Hindsights are coming ;-)

Full Changelog: v0.9.0...v0.10.0

v0.9.0

Snapshots go remote

The headline feature of this release is support for snapshots being pushed to and pull from any OCI registery. The same you already might be using for container images.

# manage registries
dryrun remote add/list/rm

# capture snapshot and push it in one step
dryrun snapshot take --push

# pull with latest / --full / --since selection
dryrun snapshot pull

Wraparound visibility
Snapshots now capture relfrozenxid and relminmxid, so dryrun can reason about transaction-ID and multixact freeze age and provide early detection of possible problems.

MCP improvements

• detect and vacuum_health output is now capped to the 50 most pressing issues, sorted worst-first
• dropped compare_nodes tool
• MCP now reports the dryrun version and schema version

dryrun setup
A new command to wire dryrun into an agent's config.

v0.8.3

What's Changed

• fix: index validity not being captured by @radim in #34

Full Changelog: v0.8.2...v0.8.3

v0.8.2

What's Changed

• feat: distribute dryrun via npm with prebuilt platform binaries by @radim in #33

Full Changelog: v0.8.1...v0.8.2

v0.8.1

Full Changelog: v0.8.0...v0.8.1