chore: publish new package versions (#4607)

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 the packages will
be published to npm automatically. 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.


# Releases
## @electric-ax/agents@0.6.2

### Patch Changes

- 9d063e4: Add desktop-configurable skill directories and common
Claude/Codex skill locations so Markdown command files load recursively
as slash-command skills without LLM metadata extraction.
- 16c8339: Use Anthropic's adaptive thinking API for models that require
it (Opus 4.6/4.7,
Sonnet 4.6). The builtin model catalog previously injected budget-based
`thinking: { type: "enabled", budget_tokens }` for every
reasoning-capable
  Anthropic model, which Opus 4.7 rejects with a 400 — it needs
`thinking: { type: "adaptive" }` + `output_config.effort`. Those models
now emit
the adaptive shape (reasoningEffort mapped to the effort level); older
models
  keep budget-based thinking.
- ac2391d: Capture OpenAI Codex reasoning summaries by default when
built-in reasoning is enabled while preserving explicit reasoning
summary settings from caller payloads.
- 8f4368d: Delegate built-in reasoning configuration to pi-ai instead of
rewriting provider request payloads in the model catalog. Built-in
OpenAI, OpenAI Codex, and Anthropic reasoning models now pass
`reasoning` through the runtime adapter, preserving existing `auto`
defaults and Anthropic thinking budgets while letting pi-ai own
provider-specific payload shapes.
- 6dbb0bb: Require meaningful slugs when spawning workers and append a
random suffix so worker URLs are easier to identify while remaining
generally unique.
- ee0da19: Fix Electric Agents quickstart startup by authenticating the
built-in pull-wake runner with the same principal it registers as,
registering built-in agent types with the local runner as their default
dispatch target, and aligning the CLI's default principal with the local
quickstart user. Pin the CLI-launched agents-server Docker image to the
matching released agents-server version, improve registration fetch
errors so startup failures include the endpoint and underlying cause,
avoid the CLI observe live-query Collection boundary, and clarify
browser-only credentials settings copy.
- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
## @electric-ax/agents-runtime@0.6.2

### Patch Changes

- 9d063e4: Add desktop-configurable skill directories and common
Claude/Codex skill locations so Markdown command files load recursively
as slash-command skills without LLM metadata extraction.
- 0418d90: Batch queued child completion wakes into a single wake
payload so parent agents receive every child result without extra
handler runs.
- ac2391d: Capture OpenAI Codex reasoning summaries by default when
built-in reasoning is enabled while preserving explicit reasoning
summary settings from caller payloads.
- 8f4368d: Delegate built-in reasoning configuration to pi-ai instead of
rewriting provider request payloads in the model catalog. Built-in
OpenAI, OpenAI Codex, and Anthropic reasoning models now pass
`reasoning` through the runtime adapter, preserving existing `auto`
defaults and Anthropic thinking budgets while letting pi-ai own
provider-specific payload shapes.
- 354b01b: Enable provider retries for agent model calls by default so
transient LLM errors are retried. Set `modelMaxRetries: 0` to preserve
the previous no-retry behavior.
- 37b9922: Fix agent wake handling so concurrent sessions do not
invalidate each other's claim write tokens, and retry same-stream wakes
after the active wake drains instead of dropping pending work.
- ee0da19: Fix Electric Agents quickstart startup by authenticating the
built-in pull-wake runner with the same principal it registers as,
registering built-in agent types with the local runner as their default
dispatch target, and aligning the CLI's default principal with the local
quickstart user. Pin the CLI-launched agents-server Docker image to the
matching released agents-server version, improve registration fetch
errors so startup failures include the endpoint and underlying cause,
avoid the CLI observe live-query Collection boundary, and clarify
browser-only credentials settings copy.
- c14a886: Remove the skill content `max` cap so `use_skill` loads
complete skill instructions instead of truncating them.
## @electric-ax/agents-server@0.6.2

### Patch Changes

- 37b9922: Fix agent wake handling so concurrent sessions do not
invalidate each other's claim write tokens, and retry same-stream wakes
after the active wake drains instead of dropping pending work.
- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
## electric-ax@0.2.22

### Patch Changes

- ee0da19: Fix Electric Agents quickstart startup by authenticating the
built-in pull-wake runner with the same principal it registers as,
registering built-in agent types with the local runner as their default
dispatch target, and aligning the CLI's default principal with the local
quickstart user. Pin the CLI-launched agents-server Docker image to the
matching released agents-server version, improve registration fetch
errors so startup failures include the endpoint and underlying cause,
avoid the CLI observe live-query Collection boundary, and clarify
browser-only credentials settings copy.
- Updated dependencies [9d063e4]
- Updated dependencies [16c8339]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [6dbb0bb]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
  - @electric-ax/agents@0.6.2
## @electric-ax/example-agents-chat-starter@0.1.13

### Patch Changes

- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
## @electric-ax/example-agents-walkthrough@0.1.10

### Patch Changes

- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
## @electric-ax/example-deep-survey@0.1.29

### Patch Changes

- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
## @electric-ax/agents-desktop@0.6.1

### Patch Changes

- 9d063e4: Add desktop-configurable skill directories and common
Claude/Codex skill locations so Markdown command files load recursively
as slash-command skills without LLM metadata extraction.
## @electric-ax/agents-mobile@0.6.2

### Patch Changes

- aa6ca85: Fix mobile chat timeline flashing when sending messages,
resizing the composer (multiline/attachments) and queuing messages, and
fix inconsistent auto-scroll as the composer grows. The timeline WebView
embed now receives dynamic updates (bottom inset, inline queued
messages, scroll) imperatively instead of via props that re-render and
flash it. Also fix the composer occasionally not clearing after send
when typing quickly.
- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [06a60fa]
- Updated dependencies [aa6ca85]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-server-ui@0.6.2
  - @electric-ax/agents-runtime@0.6.2
## @electric-ax/agents-server-ui@0.6.2

### Patch Changes

- 9d063e4: Add desktop-configurable skill directories and common
Claude/Codex skill locations so Markdown command files load recursively
as slash-command skills without LLM metadata extraction.
- 06a60fa: Fix reasoning section elapsed timers so each thinking block
measures from when that block starts, rather than from the beginning of
the run.
- aa6ca85: Fix mobile chat timeline flashing when sending messages,
resizing the composer (multiline/attachments) and queuing messages, and
fix inconsistent auto-scroll as the composer grows. The timeline WebView
embed now receives dynamic updates (bottom inset, inline queued
messages, scroll) imperatively instead of via props that re-render and
flash it. Also fix the composer occasionally not clearing after send
when typing quickly.
- ee0da19: Fix Electric Agents quickstart startup by authenticating the
built-in pull-wake runner with the same principal it registers as,
registering built-in agent types with the local runner as their default
dispatch target, and aligning the CLI's default principal with the local
quickstart user. Pin the CLI-launched agents-server Docker image to the
matching released agents-server version, improve registration fetch
errors so startup failures include the endpoint and underlying cause,
avoid the CLI observe live-query Collection boundary, and clarify
browser-only credentials settings copy.
- Updated dependencies [9d063e4]
- Updated dependencies [0418d90]
- Updated dependencies [ac2391d]
- Updated dependencies [8f4368d]
- Updated dependencies [354b01b]
- Updated dependencies [37b9922]
- Updated dependencies [ee0da19]
- Updated dependencies [c14a886]
  - @electric-ax/agents-runtime@0.6.2
## @core/sync-service@1.7.3

### Patch Changes

- 214e644: Store parsed UUID values as 16-byte binaries instead of
canonical text inside the sync-service filter evaluator. This reduces
memory use in where-clause filter indexes, especially equality and
subquery index keys, while converting UUIDs back to canonical strings at
API and SQL output boundaries. The shape metadata version is bumped so
persisted shape comparable terms created with the old UUID string
representation are ignored and rebuilt.
- 0632b1c: Stop the ShapeLogCollector from crashing when table
introspection fails on the hot replication path.

Previously the replication pipeline only handled `{:error,
:connection_not_available}` from the inspector; every other legal
inspector return crashed the ShapeLogCollector and took the whole shapes
supervision tree down with it, putting the replication client into a
10-minute noproc retry loop:
- `:table_not_found` (table dropped or renamed between the WAL record
and introspection) crashed `Partitions.handle_relation/2` and the
pk-column lookup with a `CaseClauseError`. Relation messages for dropped
tables are now ignored and changes for dropped tables are keyed on the
full record, the same fallback used for tables without a primary key.
- In-band database errors (e.g. out-of-memory, `statement_timeout`
cancelling the catalog query) crashed `Partitions.handle_relation/2`
with a `CaseClauseError` and made `Partitions.add_shape/3` raise. They
are now propagated as `{:error, reason}`; the collector logs a warning
and pauses replication via the existing retry path until introspection
succeeds.
- Connection-class errors returned in-band by Postgrex (e.g. `"ssl recv:
closed"`) are now classified as `:connection_not_available` instead of
leaking through as strings.
- Shape restore no longer crashes with a `MatchError` when the
connection pool isn't ready yet; it retries the introspection in place
and only gives up (with a descriptive error) after ~10 seconds.

- 0cf38e8: Honor an upstream head-sampling rate hint received via the
W3C `tracestate` header (`electric=rate:<N>`) on shape GET requests:

When the remote parent trace is sampled, the `Plug_shape_get` root span
(and the `shape_get.plug.stream_chunk` child spans) are stamped with the
`SampleRate` attribute — `N` for responses with status < 500 and `1` for
5xx responses — so tracing backends that understand sampling weights
scale aggregates over Electric's spans by the upstream sampling rate
instead of under-reporting traffic ~N-fold.

Requests without a remote trace context, with a not-sampled remote
parent, or with a missing/invalid rate hint, behave exactly as before.

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Author: github-actions[bot]