chore: finalize kv structure for queue, workflows, traces, sqlite

[NathanFlurry]

Description

Please include a summary of the changes and the related issue. Please also include relevant motivation and context.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

How Has This Been Tested?

Please describe the tests that you ran to verify your changes.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[railway-app]

🚅 Deployed to the rivet-pr-4284 environment in rivet-frontend

Service	Status	Web	Updated (UTC)
frontend-cloud	😴 Sleeping (View Logs)	Web	Feb 24, 2026 at 7:28 am
website	😴 Sleeping (View Logs)	Web	Feb 24, 2026 at 7:25 am
frontend-inspector	😴 Sleeping (View Logs)	Web	Feb 24, 2026 at 7:25 am
mcp-hub	✅ Success (View Logs)	Web	Feb 24, 2026 at 7:15 am
ladle	❌ Build Failed (View Logs)	Web	Feb 24, 2026 at 7:15 am

[NathanFlurry]

• fix(rivetkit): stabilize workflow teardown timing test and restore queue limit test coverage #4287
• chore: finalize kv structure for queue, workflows, traces, sqlite #4284 👈 (View in Graphite)
• chore: increase actor storage size to 10 gib #4283
• chore(rivetkit): remove legacy behavior #4282
• chore: move otel to experimental flag #4280
• main

---

How to use the Graphite Merge Queue

Add the label merge-queue to this PR to add it to the merge queue.

You must have a Graphite account in order to use the merge queue. Sign up using this link.

An organization admin has enabled the Graphite Merge Queue in this repository.

Please do not merge from GitHub as this will restart CI on PRs being processed by the merge queue.

This stack of pull requests is managed by Graphite. Learn more about stacking.

[claude]

PR Review: chore: finalize kv structure for queue, workflows, traces, sqlite

This PR restructures the actor KV keyspace by adding storage versioning bytes and renumbering top-level namespaces. The intent is clearly to finalize the structure before it becomes immutable in production. The approach is directionally correct — versioned namespaces are the right design — but there are some issues worth flagging.

---

Breaking Change Without Migration

This is the most significant concern. The PR changes key paths for every major subsystem:

Subsystem	Old key prefix	New key prefix
Queue messages	[5]{message_id}	[5][1][2]{message_id}
Queue metadata	[6]	[5][1][1]
Workflow	[7]{key}	[6][1]{key}
Traces	[8]{key}	[7][1]{key}
SQLite	[9][0/1, ...]	[8][1][0/1, ...]

Any actor with existing data under the old scheme will silently lose access to it after this change. There is no migration path. If this is intentional (e.g., actors don't yet have meaningful production data at the old paths), that should be stated explicitly in the PR description to make the intent clear to future readers.

---

Documentation Inconsistency

The new ACTOR_KV_STRUCTURE.md tree view has this comment:

sqlite (8)/    # SQLite VFS namespace.
  v1 (1)/    # SQLite data version. Legacy pre-v1 SQLite keys are still resolved in sqlite-vfs/src/vfs.ts.

However, sqlite-vfs/src/vfs.ts contains no legacy key resolution logic. It only imports and uses the new getMetaKey / getChunkKey functions from kv.ts. This claim is inaccurate and should be removed or corrected.

---

Workflow Field VERSION Removal

WORKFLOW_FIELD.VERSION (previously [3, 4]) is removed, and INPUT is renumbered from [3, 5] to [3, 4]. Any workflow that previously stored version or input data at the old indices will silently read null or misinterpret the stored bytes after this change. Like the namespace renumbering, this is a hard break. Same comment applies: if intentional, state it explicitly.

---

rebuildMetadata Will Miss Old Queue Messages

In queue-manager.ts, #rebuildMetadata is called as a fallback when queue metadata can't be decoded. It scans for messages using QUEUE_MESSAGES_PREFIX (new path [5][1][2]). Old messages stored at [5]{message_id} won't be discovered during a rebuild, meaning the fallback recovery path does not actually recover pre-migration data.

---

Minor: Duplicate concatPrefix Implementations

concatPrefix is defined in both keys.ts and traces-driver.ts with identical implementations. The one in keys.ts could be exported and reused in traces-driver.ts.

---

No Tests

No new tests are added to cover the restructured key paths, the new versioned prefix functions (queueMetadataKey, queueMessagesPrefix, etc.), or the decodeQueueMessageKey prefix validation. The CLAUDE.md testing guidelines specify that driver tests should cover new functionality. This is especially important given that key encoding bugs would silently lose actor data.

---

Positive Notes

• Adding explicit versioning bytes to all namespaces is the right design and makes future migrations possible.
• The decodeQueueMessageKey prefix validation added in this PR is a good defensive check.
• Reducing maxQueueMessageSize from 1 MiB to 64 KiB is the correct fix — it was previously possible to configure a value that would fail at KV write time without any error at configuration time.
• Increasing KV storage limit from 1 GiB to 10 GiB and the simplified docs are reasonable.
• The new ACTOR_KV_STRUCTURE.md tree view is much cleaner and more readable than the previous version.