This PR was opened by the [Changesets
release](https://github.com/changesets/action) GitHub action. When
you're ready to do a release, you can merge this and publish to npm
yourself or [setup this action to publish
automatically](https://github.com/changesets/action#with-publishing). If
you're not ready to do a release yet, that's fine, whenever you add more
changesets to main, this PR will be updated.
⚠️⚠️⚠️⚠️⚠️⚠️
`main` is currently in **pre mode** so this branch has prereleases
rather than normal releases. If you want to exit prereleases, run
`changeset pre exit` on `main`.
⚠️⚠️⚠️⚠️⚠️⚠️
# Releases
> The changelog information of each package has been omitted from this
message, as the content exceeds the size limit.
## @mastra/slack@1.9.0-alpha.0
## @mastra/factory@0.28.0-alpha.3
## @mastra/code-sdk@1.19.0-alpha.3
## mastracode@0.50.0-alpha.3
## @mastra/observability@1.24.0-alpha.2
## @mastra/core@1.83.0-alpha.3
## @mastra/mcp-apps-components@0.1.0-alpha.3
## @mastra/playground-ui@69.0.0-alpha.4
## @mastra/server@1.83.0-alpha.3
## @mastra/server-adapters-test-suite@0.3.0-alpha.1
## @mastra/platform-workspace@1.13.0-alpha.0
## @mastra/auth-better-auth@1.1.11-alpha.1
## @mastra/client-js@1.60.0-alpha.3
## @mastra/react@1.15.0-alpha.3
## @mastra/deployer-cloud@1.83.0-alpha.3
## @mastra/longmemeval@1.1.42-alpha.3
## @mastra/opencode@0.1.42-alpha.3
## create-factory@0.4.3-alpha.3
## @mastra/arize@1.3.30-alpha.2
## @mastra/arthur@0.4.30-alpha.2
## @mastra/braintrust@1.3.27-alpha.2
## @mastra/datadog@1.4.22-alpha.2
## @mastra/deepeval@0.1.24-alpha.2
## @mastra/laminar@1.3.32-alpha.2
## @mastra/langfuse@1.6.6-alpha.2
## @mastra/langsmith@1.4.3-alpha.2
## @mastra/otel-bridge@1.5.22-alpha.2
## @mastra/otel-exporter@1.5.2-alpha.2
## @mastra/posthog@1.4.7-alpha.2
## @mastra/sentry@1.2.32-alpha.2
## @mastra/agent-builder@1.2.1-alpha.1
## mastra@1.39.0-alpha.3
## @mastra/connect@1.3.2-alpha.2
## @mastra/deployer@1.83.0-alpha.3
## @mastra/editor@0.17.1-alpha.2
## @mastra/fastembed@1.3.7-alpha.0
## @mastra/mcp-docs-server@1.3.13-alpha.6
## @mastra/memory@1.41.1-alpha.1
## @mastra/elysia@0.1.23-alpha.3
## @mastra/express@1.5.27-alpha.3
## @mastra/fastify@1.5.27-alpha.3
## @mastra/hono@1.7.25-alpha.3
## @mastra/koa@1.7.27-alpha.3
## @mastra/nestjs@0.3.0-alpha.3
## @mastra/next@0.2.41-alpha.3
## @mastra/tanstack-start@0.2.41-alpha.3
## @mastra/convex@1.8.1-alpha.0
## @mastra/pg@1.36.0-alpha.2
## @mastra/inngest@1.11.2-alpha.2
## @mastra/temporal@0.6.0-alpha.3
## create-mastra@1.39.0-alpha.3
## @internal/playground@1.39.0-alpha.3
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
## Description
Durable and evented agents wrote two or three semantic recall vector
rows for each message. Two writers embedded the same messages:
- the `SemanticRecall` output processor, which keys rows by message id;
- `Memory.saveMessages`, which the save queue calls from the durable
input save and the finalize flush, and which wrote rows without ids, so
the store generated random ones.
The duplicates took up `topK` slots, and the embedder ran twice per
message.
The two writers also built rows differently. Only `saveMessages` chunked
long text, stripped working-memory tags and stored thread metadata.
Neither writer can simply be removed: the processor serves agent runs,
while `saveMessages` serves direct callers (the HTTP route, the client
SDK, user code) and the save queue. So this PR gives both writers one
row builder and one id scheme.
Changes:
- Core: `buildMessageEmbeddingRows` and `upsertMessageVectors`
(internal, in `@mastra/core/memory`) now decide the row format: text
extraction, chunking, metadata and row ids. A message's first chunk
keeps its message id, which is the id the processor already wrote, so
existing rows get replaced rather than duplicated. Later chunks get a
UUID derived from the message id and chunk index. Every message-vector
upsert now sends ids. Each writer passes its own embed function, so
embedding calls, retries and caches work as before.
- Core: `MessageList` keeps a per-run record of embedded messages (index
and message id to text hash) and serializes it with the list. The save
queue passes the record to `saveMessages`, and the durable input save
copies it back from its cloned list, so the second writer skips what the
first one embedded into the same index. If a message's text has changed
since, it is embedded again. A processor writing to a custom
`semanticRecall.indexName` and `Memory` writing to its default index
each still get every message.
- Core: new `semanticRecall.maxChunkChars` option, default 16,384
characters (about 4,096 English tokens). The limit is in characters
because each provider uses its own tokenizer. The processor now chunks
long messages the way `saveMessages` already did, instead of embedding
them as one input, which failed past the embedder's limit.
- Memory: `saveMessages`, `updateMessages`, cloned-message embedding and
`updateThreadResourceId` all write through the builder. Edits and
resource moves embed before deleting the old rows, so a failed embedding
call leaves them in place. Thread metadata is read only when rows will
be written. The `@mastra/core` peer range is now `>=1.83.0-0`.
- FastEmbed: models declare `maxInputTokens: 512`, so recall chunks fit
their window.
Existing rows are not rewritten. Rows `saveMessages` wrote before this
change keep their random ids. Rows the processor wrote already use the
message id, so new writes replace them. An evented or Inngest finalize
that runs on another worker doesn't see the input save's record, so it
embeds the user message once more into the same row.
```ts
const memory = new Memory({
options: {
semanticRecall: { topK: 3, messageRange: 1, maxChunkChars: 6000 },
},
})
```
Tests:
`packages/memory/integration-tests/src/semantic-recall-embedding-rows.test.ts`
runs plain, durable and evented agents against a LibSQL vector index,
with thread scope, resource scope and a custom index. It checks:
- each persisted message has exactly one row under its own id;
- no row is written twice in a run;
- the fact is recalled once.
On `main` the durable and evented cases fail with 30 rows for 12
messages. With the record not passed to `saveMessages`, the write-once
check fails on durable and evented. Memory cases cover:
- record skips;
- chunk size taken from the embedder's input limit;
- an edit that shortens a long message;
- an embedder failure during an edit.
Unit tests cover the builder and the `MessageList` record.
Six cases in `agent-memory.test.ts` fail because their recorded fixtures
no longer match the requests. They fail the same way on `main`.
## Related issue(s)
Related to #29352. Agent runs now store thread metadata on vector rows,
which the read-side fix in that issue needs. This PR does not change the
read side.
## Type of change
- [x] Bug fix (non-breaking change that fixes an issue)
- [x] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [x] Documentation update
- [ ] Code refactoring
- [ ] Performance improvement
- [x] Test update
## Checklist
- [x] I have linked the related issue(s) in the description above
- [x] I have made corresponding changes to the documentation (if
applicable)
- [x] I have added tests that prove my fix is effective or that my
feature works
- [ ] I have addressed all Coderabbit comments on this PR
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
<!-- coderabbit-ai-generated:v1 -->
## ELI5
The PR helps agents avoid saving the same message to search more than
once. It also splits long messages into smaller pieces so search can use
more of their content.
## Summary
- Shared helpers extract message text, split it into chunks, build
vector rows, and upsert them. The first chunk uses the message ID when
it is a UUID or small integer; other IDs and later chunks use stable
derived IDs.
- `SemanticRecall` and `Memory.saveMessages()` use the shared
row-building path. Agent runs track embedded text by message and index,
so writers can skip unchanged messages.
- `semanticRecall.maxChunkChars` sets the chunk limit. It defaults to
16,384 characters and can use the embedder’s declared `maxInputTokens`
when unset. FastEmbed models declare a 512-token input limit.
- Memory updates and thread-resource moves build replacement rows before
deleting existing vectors. Re-saving shorter text can leave older chunk
rows searchable; the documentation describes this behavior.
- Tests cover row counts, scopes, custom indexes, chunking, repeated
saves, edits, and embedding failures. The supplied description reports
six fixture-based `agent-memory.test.ts` failures that also occur on the
base branch. Completed check results are not available.
- Documentation and peer dependency minimum versions were updated.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Enables all eleven requested Oxlint rules as errors in the root config
and migrates the 115 existing findings across 55 files. JSX curly-brace
simplification is limited to props (`props: 'never', children:
'ignore'`), boolean props use shorthand, and empty elements become
self-closing.
The manual edits preserve the exact continued-string and template
values, remove redundant assignment-side non-null assertions, and retain
the indexed-read assertion in fastembed's compound assignment. The
BigInt and octal conversions use native globals and exact values
supported by the existing compilation targets. No lint suppressions were
added.
Validation: the eleven-rule scan reports zero diagnostics; before/after
emitted-JavaScript comparison passes for all 55 migrated files,
including 31 JSX files, after normalizing only the reviewed syntax
rewrites. Formatting and whitespace checks pass. Core utility/clone
tests (146), fastembed pooling tests (9), the changed A2A signature test
(1), and fastembed build-config typechecking pass. The root lint scan
retains the same 1,244 pre-existing diagnostics with no new findings.
GitHub CI now also passes the full lint/format job, build, Mastra Code
typecheck, generated-file check, and bundle validation. Local UI/evals
runs were limited by missing dependencies; the CodeDiff folding failure
was reproduced on unchanged baseline code. A second audit confirmed
native unshadowed globals for all five literal conversions, pure targets
for all twelve assignment rewrites, no references or direct eval in all
eight simplified catches, and ordinary semantics for both computed keys.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
<!-- coderabbit-ai-generated:v1 -->
## ELI5
This PR turns on more code-style checks and updates existing code to
pass them. The changes simplify syntax while preserving the intended
behavior.
## Changes
- Enabled eleven Oxlint rules as errors in the root configuration.
- Updated 115 findings across 55 files. Changes include simpler string
and JSX syntax, compound assignments, native BigInt and octal literals,
and catch clauses without unused error variables.
- Configured JSX curly-brace simplification for props only. JSX children
remain unchanged.
- Removed redundant assignment-side non-null assertions while retaining
the indexed-read assertion needed for fastembed’s compound assignment.
- Added no lint suppressions.
## Validation
The author reports zero diagnostics from the eleven-rule scan and no new
findings in the root lint scan, which retains 1,244 pre-existing
diagnostics. Before-and-after emitted-JavaScript comparisons reportedly
pass for all 55 changed files after normalizing the reviewed syntax
rewrites. Formatting and whitespace checks also pass.
Reported tests include 146 core utility and clone tests, 9 fastembed
pooling tests, 1 changed A2A signature test, and fastembed build-config
typechecking. Full workspace lint and the UI/evals suites were limited
by missing dependencies in the checkout. The CodeDiff folding test fails
on unchanged baseline code.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Description
`SparseTextEmbedding.init()` fails with `ENOENT` on the first run when
`cacheDir` contains missing parent directories, such as `parent/models`.
Create those directories recursively, matching dense retrieval and
preserving mode `0o755`.
```diff
- fs.mkdirSync(cacheDir, { mode: 0o755 });
+ fs.mkdirSync(cacheDir, { recursive: true, mode: 0o755 });
```
The regression exercises public initialization and embedding with a
local HTTP server and real tokenizer/ONNX fixtures. It checks downloaded
bytes, known sparse-vector values, and cache reuse without additional
HTTP requests. Includes a patch changeset for `@mastra/fastembed`.
Before/after evidence from the fix agent's logs:
- Command: `timeout 3m pnpm --filter @mastra/fastembed test
sparse-cache.test.ts --maxWorkers=4`
- Before: 1 failed, 8 passed. The new nested-cache case failed with
`ENOENT` in `SparseTextEmbedding.retrieveModel`.
- After: 9 passed.
- Full suite: `timeout 5m pnpm --filter @mastra/fastembed test
--maxWorkers=4` passed 121 tests across 6 files after building workspace
dependencies, and passed again after merging main at `399cb80de9`.
Production and targeted regression typechecks, changed-file lint, and
formatting passed. Broader package lint still reports five existing
import errors in untouched files; the expanded all-tests typecheck
reports existing errors in untouched tests.
External pretrained downloads, the full pretrained-model test project,
and Windows/ARM64 were not tested because the local fixtures exercise
this filesystem regression without external downloads. No standalone
package build was run; source APIs were exercised and production
declarations typechecked.
## Related issue(s)
Fixes#28237.
Implementation and regression adapted from [sdivyanshu90's contributor
commit](f1c9086e40),
with co-author credit in both commits. No related open PR found. #28236
concerns separate tokenizer metadata behavior and is outside this
change.
## Type of change
- [x] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Code refactoring
- [ ] Performance improvement
- [x] Test update
## Checklist
- [x] I have linked the related issue(s) in the description above
- [ ] I have made corresponding changes to the documentation (if
applicable)
- [x] I have added tests that prove my fix is effective or that my
feature works
- [ ] I have addressed all Coderabbit comments on this PR
Created with GPT-6.1 Sol in T3 Code.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
<!-- coderabbit-ai-generated:v1 -->
## ELI5
Sparse model downloads now create any missing folders in the cache path.
This lets the model initialize on a first run without requiring callers
to create those folders first.
## Changes
- Create missing cache directories recursively while preserving the
existing `0o755` mode.
- Add a regression test that checks downloaded files, sparse embeddings,
and cache reuse without extra HTTP requests.
- Add a patch changeset for `@mastra/fastembed`.
## Validation
The linked issue reports that all 9 targeted tests pass after the fix
and that the full package suite passes 121 tests across 6 files. It also
reports passing typechecks, changed-file lint, and formatting. Broader
package lint and the all-tests typecheck report existing errors in
untouched files. External pretrained downloads, Windows/ARM64, and a
standalone package build were not tested.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: sdivyanshu90 <divyanshu74.80@gmail.com>
## Description
Custom tokenizers that omit `model_max_length` or set it to null can
silently discard every token, even with a positive initialization limit.
This produces empty dense and sparse embeddings and makes
`countTruncated()` report `maxLength: 0`.
The shared loader now preserves the requested limit, or the existing
512-token default, when that metadata is missing or null. Valid metadata
still uses the smaller model/requested limit.
```diff
-maxLength = Math.min(maxLength, tokenizerConfig['model_max_length']);
+maxLength = Math.min(maxLength, tokenizerConfig['model_max_length'] ?? maxLength);
```
Adds 10 regression and control cases using public dense/sparse APIs, a
real native tokenizer, and local ONNX fixtures, plus a patch changeset
for `@mastra/fastembed`. Other malformed metadata and invalid caller
limits remain outside this fix.
Before/after evidence from the implementation logs, on Linux x64 CPU:
```sh
timeout 10m pnpm --filter @mastra/fastembed test tokenizer-length.test.ts --maxWorkers=2
```
Before: 6 failed, 4 passed, with a zero-token window, undefined dense
components, and empty sparse indices. After: all 10 passed.
```sh
timeout 10m pnpm --filter @mastra/fastembed test --maxWorkers=2
```
All 130 tests across seven files passed, including a repeat after
syncing main. Changed-file strict typecheck, lint, formatting, and diff
checks also passed. These are recorded implementation runs; this PR
preparation did not rerun tests or modify the branch.
What was not tested, and check limitations:
- Package-wide typecheck and type-aware lint were attempted but did not
pass. `tsc --noEmit -p packages/fastembed/tsconfig.json` reports TS6059
because `tsdown.config.ts` is outside `rootDir: src`. `oxlint
--type-aware packages/fastembed` reports that config error and five
existing import-order/type-specifier errors in untouched `src/index.ts`
and `tsdown.config.ts`. The config and both source files are unchanged
from main. Widening `rootDir` exposes additional existing model-test and
types-builder JavaScript diagnostics. Passing changed-file checks do not
establish a clean package-wide check.
- Pretrained model download suites (`test:models`), GPU providers,
non-Linux platforms, full monorepo checks, and a published-package
consumer build were not tested. Validation focused on the affected
dense/sparse paths using local fixtures and native CPU inference,
without pretrained downloads.
## Related issue(s)
Fixes#28236.
Implementation and regression tests adopted from [sdivyanshu90's commit
ddcdd193d711c04aa9fbd6c2915e3288435fa943](ddcdd193d7),
with co-author credit retained in the fix commits. No related open PR
was found.
## Type of change
- [x] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Code refactoring
- [ ] Performance improvement
- [x] Test update
## Checklist
- [x] I have linked the related issue(s) in the description above
- [ ] I have made corresponding changes to the documentation (if
applicable)
- [x] I have added tests that prove my fix is effective or that my
feature works
- [ ] I have addressed all Coderabbit comments on this PR
Created with GPT-6.1 Sol in T3 Code.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
<!-- coderabbit-ai-generated:v1 -->
## ELI5
FastEmbed could lose its token limit when a custom tokenizer did not
include one. This fix keeps the requested limit, or the 512-token
default, so dense and sparse embeddings can process input correctly.
## Summary
- The shared tokenizer loader now uses `model_max_length` when it is
present and non-null.
- When metadata is missing or null, the loader preserves the requested
limit or the default of 512.
- When metadata provides a limit, the loader continues to use the
smaller of the metadata and requested limits.
- Regression tests use local ONNX fixtures and a WordLevel tokenizer.
They cover dense embeddings, sparse embeddings, truncation counts,
missing and null metadata, and cases where either limit is smaller.
## Validation
The author reports that all 10 targeted regression and control cases
passed, and the full FastEmbed test suite passed all 130 tests across
seven files on Linux x64 CPU. Changed-file typecheck, lint, formatting,
and diff checks also passed. Package-wide typecheck and type-aware lint
did not pass because of reported existing configuration and
untouched-file errors. Pretrained model downloads, GPU providers,
non-Linux platforms, full monorepo checks, and a published-package
consumer build were not tested.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: sdivyanshu90 <divyanshu74.80@gmail.com>
## Description
Cold model downloads wrote progress messages to stdout, contaminating
JSONL output. Send FastEmbed's own notice to stderr and use Hugging
Face's streaming downloadFile API so its cache helper cannot emit
additional stdout messages. Both dense and sparse downloads use the
quiet path; dense atomic publication and cache cleanup are preserved.
Stack 5/6. Review against `stack/mastra-04-jsonl-display`; merge
bottom-up. Only this fix is added relative to the preceding branch.
## Related issue(s)
Refs #26877, finding 5.
## Validation
13 retrieval/cache tests pass using the real Hugging Face Hub 2.12.0
library with mocked HTTP and real filesystem operations. Three stdout
regressions fail on the old helper. Native tokenizer/ONNX imports are
stubbed; no real model download or inference. Focused TypeScript
checking, formatting and whitespace checks pass. Full monorepo build has
not been run.
## Type of change
- [x] Bug fix
## Checklist
- [x] Linked the tracking issue
- [x] Added regression tests
- [ ] Full package build and typecheck
- [ ] Addressed all Coderabbit comments
## Stack
Review and merge in this order:
1. #26878 — fix(core): restore the default foreground shell timeout
2. #26879 — fix(core): force-kill commands that outlive their timeout
3. #26880 — fix(code-sdk): drain headless output before exiting
4. #26881 — fix(code-sdk): omit display snapshots from headless JSONL
5. #26882 — fix(fastembed): send model download notices to stderr
6. #26883 — fix(code-sdk): resume policy-answered headless suspensions
---------
Co-authored-by: Aamish <your@email.com>
> **Stacked PR (3 of 3).** Base: #25998 (`fix/om-commit-queue`), which
is stacked on #25906 (`fix/om-lossless-rollover`). Review this diff
against #25998. Draft until the stack below it lands.
## Summary
Keep Observational Memory asynchronous between `messageTokens` and
`observation.blockAfter`. Today an agent that reaches `messageTokens`
waits on a synchronous Observer call, and crossing the threshold also
stops new background buffering from starting.
With async buffering enabled:
- **Between `messageTokens` and `blockAfter`:** background buffering
continues, and ready buffered observations activate without waiting on
an Observer call that is still running.
- **Multipliers:** `blockAfter` multipliers resolve against each
record's effective `messageTokens`. The default is `1.2`, which is
36,000 tokens for the default 30,000-token threshold.
- **At or above `blockAfter`:** activation runs first. If pending
messages still reach the limit afterwards, synchronous observation
covers only the messages that are still unobserved, keeping the
activation and cleanup behavior from #25060. Activation waits a bounded
time (60s) for an in-process buffer op that may still produce a chunk,
so that chunk activates instead of its messages being observed again.
Activations triggered by the idle TTL or a model change never wait for
that op.
- **Opting out:** with async observation disabled, synchronous
observation still runs at `messageTokens`. `blockAfter: 1` turns off the
async band.
`getStatus()` now returns `observationBlockAfter` and
`inAsyncObservationBand`.
## How it relies on the stack
#25906 makes storage keep these cases lossless on every adapter, and
#25998 serializes OM commits per thread or resource within a process:
- a chunk written after a reflection rollover lands on the current
generation;
- rollover carries buffered chunks to the new generation;
- activation swaps from the stored chunk list.
Because of that, this PR no longer carries its own guards for them.
Activation and sync observation no longer defer while a chunk write is
in flight, and the persist-time retarget is gone. For each removed
guard, the replacing storage mechanism was disabled and a test was
confirmed to fail; the runs are recorded in `ARCHITECTURE.md` (D3).
Kept from this PR:
- the async band policy;
- the chunk-write phase, tracked separately from indexing, so a stored
chunk can activate while its op is still indexing;
- the buffering-end marker emitted when the chunk is saved;
- the current turn's user input kept in the input bucket during step-0
buffering, so semantic recall still sees it;
- `getDynamicMemory().settled()` in `@mastra/code-sdk`, which the TUI
E2E teardown awaits;
- fastembed resolving its cache directory per model load.
A buffer op that outlives an activation, or is queued behind one, no
longer writes its pre-activation token boundary back over activation's
reset. Each op owns its boundary through a per-op token.
## Test plan
- [x] Memory unit suite: 96 files, 2,345 passed, 1 todo; typecheck,
lint, and build clean.
- [x] `ready-late-rollover.test.ts`: a ready chunk plus a late in-flight
chunk across a rollover, triggered either by step 0 or by another
writer. Every message behind the head cursor has its observation on the
head. Passes 4/4; the step-0 + ready-chunk case fails on this branch's
pre-stack head `4161e237`.
- [x] `buffer-write-generation.test.ts`: passes; fails on `ede145bf`
(before the retarget).
- [x] Async band and tail matrix (`threshold-activation-tail.test.ts`,
18 tests): default, absolute, and per-record gates; async disabled;
partial activation; reflected generations; no duplicate observation.
- [x] New regressions:
- a TTL activation doesn't wait on an in-flight buffer op;
- a hung op's chunk that lands during activation's swap survives;
- an activation's boundary reset survives an in-flight op and a queued
op.
- [x] Seeded lifecycle fuzz on the full stack: 500 InMemory seeds and
100 LibSQL file-database seeds, 0 failing.
- [x] Proof probes on the full stack:
- P1/P6: 2/2;
- P4 actor-context probe: 11/11 × 3, with PostgreSQL taking the
row-lock-wait path (the same probe fails 8/11 on `main`);
- durable agent: passes;
- cross-process LibSQL + PostgreSQL, 250 iterations × 3 seeds: no
violations.
- [x] OM TUI E2E scenarios: 5 passed.
---------
Co-authored-by: Mastra Code (anthropic/claude-fable-5) <noreply@mastra.ai>
Co-authored-by: Tyler Barnes <tylerdbarnes@gmail.com>
Co-authored-by: mastracode <284800079+mastra-platform[bot]@users.noreply.github.com>
This PR was opened by the [Changesets
release](https://github.com/changesets/action) GitHub action. When
you're ready to do a release, you can merge this and publish to npm
yourself or [setup this action to publish
automatically](https://github.com/changesets/action#with-publishing). If
you're not ready to do a release yet, that's fine, whenever you add more
changesets to main, this PR will be updated.
⚠️⚠️⚠️⚠️⚠️⚠️
`main` is currently in **pre mode** so this branch has prereleases
rather than normal releases. If you want to exit prereleases, run
`changeset pre exit` on `main`.
⚠️⚠️⚠️⚠️⚠️⚠️
# Releases
> The changelog information of each package has been omitted from this
message, as the content exceeds the size limit.
## @mastra/discord@1.1.0-alpha.0
## @mastra/client-js@1.50.0-alpha.1
## @mastra/connect@0.4.0-alpha.0
## @mastra/core@1.71.0-alpha.1
## @mastra/playground-ui@59.0.0-alpha.1
## @mastra/server@1.71.0-alpha.1
## @mastra/telegram@0.1.3-alpha.0
## @mastra/ai-sdk@1.10.5-alpha.0
## @mastra/react@1.6.3-alpha.1
## @mastra/deployer-cloud@1.71.0-alpha.1
## @mastra/longmemeval@1.1.30-alpha.1
## @mastra/opencode@0.1.30-alpha.1
## @mastra/factory@0.17.2-alpha.1
## create-factory@0.2.3-alpha.1
## @mastra/code-sdk@1.8.3-alpha.1
## mastracode@0.42.2-alpha.1
## @mastra/arize@1.3.19-alpha.0
## @mastra/arthur@0.4.19-alpha.0
## @mastra/braintrust@1.3.16-alpha.0
## @mastra/datadog@1.4.11-alpha.0
## @mastra/deepeval@0.1.13-alpha.0
## @mastra/laminar@1.3.21-alpha.0
## @mastra/langfuse@1.5.10-alpha.0
## @mastra/langsmith@1.3.21-alpha.0
## @mastra/observability@1.18.1-alpha.0
## @mastra/otel-bridge@1.5.11-alpha.1
## @mastra/otel-exporter@1.4.2-alpha.0
## @mastra/posthog@1.3.13-alpha.0
## @mastra/sentry@1.2.21-alpha.0
## @mastra/agent-builder@1.1.22-alpha.0
## mastra@1.31.3-alpha.1
## @mastra/deployer@1.71.0-alpha.1
## @mastra/editor@0.15.3-alpha.0
## @mastra/fastembed@1.3.2-alpha.0
## @mastra/mcp-docs-server@1.3.1-alpha.2
## @mastra/memory@1.32.1-alpha.0
## @mastra/elysia@0.1.11-alpha.1
## @mastra/express@1.5.15-alpha.1
## @mastra/fastify@1.5.15-alpha.1
## @mastra/hono@1.7.13-alpha.1
## @mastra/koa@1.7.15-alpha.1
## @mastra/nestjs@0.2.30-alpha.1
## @mastra/next@0.2.29-alpha.1
## @mastra/tanstack-start@0.2.29-alpha.1
## @mastra/clickhouse@1.21.1-alpha.0
## @mastra/duckdb@1.11.1-alpha.1
## @mastra/libsql@1.23.3-alpha.0
## @mastra/mysql@0.10.2-alpha.0
## @mastra/oracledb@0.4.1-alpha.0
## @mastra/pg@1.27.1-alpha.0
## @mastra/spanner@1.8.2-alpha.0
## @mastra/turso@0.1.9-alpha.0
## @mastra/temporal@0.4.9-alpha.1
## create-mastra@1.31.3-alpha.1
## @internal/playground@1.31.3-alpha.1
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>