- Fix registry SDK detection: use 'workspace_operations' not in RunnerPermissions.model_fields instead of hasattr
- Remove raw_manifest fallback from declared_workspace_operations and resource_builder so 0.7.19 safely disables declared grants
- Unify is_local_agent check across descriptor.py and resource_builder.py, fail-closed when descriptor is None
- Add unmocked tests for real installed 0.7.19 SDK and real new SDK
- Add tests/demo_e2e_workspace_runner.py for reproducible end-to-end verification
Co-authored-by: multica-agent <[email protected]>
A processor granted pipeline.extensions.update / pipeline.update / agent.update
derives resource.manage, which let it add any resource.manage operation to its
own grants. Granting is a human authorization act, so processor principals may
only keep or remove grants, and the catalog marks nothing grantable for them.
LANG-226
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Co-authored-by: multica-agent <[email protected]>
LANG-231: Fix SDK 0.7.19 input compatibility regressions by accepting optional
cursor in ExecutionRuntimesInput and float | None stale_after_seconds in
ExecutionRuntimeReconcileInput. Update test_all_69_operations_route_through_execution_api
to drive real SDK 0.7.19 payloads through handler and assert raw execution dispatch.
Add autouse fixture in plugin tests to isolate transfer root environment leaks.
Co-authored-by: multica-agent <[email protected]>
- Add test_referenced_operation_ids_exist asserting all 'get it from X' and 'Call X' targets exist in the catalog
- Add assertions for model_uuid, server_uuid, mcp.server.* risk and exposure
Ref: LANG-229
Co-authored-by: multica-agent <[email protected]>
* feat(plugin): upload debug plugin to LangBot Space with optional GitHub sync
- SDK: reusable in-memory .lbpkg packaging + git sync helper (local git by default, optional repo/token override)
- routing: space-upload config/upload/git-sync endpoints; debug plugins resolved via OSS bridge without touching get_plugin_info
- frontend: debug-card entry + upload dialog with multi-language metadata, icon upload and changelog
- i18n across 8 locales; operation-trace + action allowlist entries
* fix(plugin-space-upload): stable dialog width, single-line locale tabs, broader error surfacing
- fixed dialog width so toggling the GitHub section no longer resizes it
- language tabs render on one line (horizontal scroll when needed)
- surface the real upload/git error message (msg field) instead of [object Object]
- hint that a missing GitHub repo is auto-created when a token is provided
* fix(plugin-space-upload): narrow dialog and reset sm:max-w clamp
The base DialogContent applies sm:max-w-lg at the sm breakpoint, which overrode the unprefixed max-w-none and clamped the upload card to 32rem on desktop. Set an explicit sm:max-width and size the card to ~36rem. Locale tabs now wrap instead of scrolling horizontally.
* fix
* fix(plugin-space-upload): address CI and UX feedback
- narrow the upload dialog to ~36rem and reset the sm:max-w clamp
- guide the publisher to the Space review page after a successful upload
- drop the redundant null comparison flagged by code quality
- update the operation-action expectation and Space/LangBot Account wording
- ruff format the touched API modules
* feat(plugin-space-upload): public repo, editable author/license, card red dot
* fix(market): replace yellow download count with update available badge
* style(market): unify installed and update indicators into badges
* feat(plugin): redesign debug card banner, style debug icon, and prioritize debug plugins
* feat(plugin-space-upload): simplify upload dialog layout, add creator verification and market redirection
* fix(i18n): use LangBot Account terminology in Space upload dialog
Co-authored-by: multica-agent <[email protected]>
* style(i18n): reflow reworded locale strings for prettier
Co-authored-by: multica-agent <[email protected]>
* test(utils): stop timing out the pattern-count cap test on loaded runners
Co-authored-by: multica-agent <[email protected]>
---------
Co-authored-by: Hyu <[email protected]>
Co-authored-by: RockChinQ <[email protected]>
Co-authored-by: multica-agent <[email protected]>
Drop intermediate stream frames that land inside the pacing window so a fast
runner cannot flood the WeCom wire. The first frame of a stream is released
immediately because there is no prior timestamp to compare against, and final
frames always bypass the throttle.
Frames carry complete snapshots since a2b7f6b7, so dropping an intermediate
frame loses no content: the next frame already carries everything.
Stdio and Windows launches always bound the SDK default debug port 5401,
so changing plugin.display_plugin_debug_url only updated the console URL
and a second Core instance hit Address already in use. Pass the configured
debug (and Windows control) ports into the Runtime CLI, and document the
behavior.
Fixes#2713
Co-authored-by: Li-john1021 <[email protected]>
Persist the v4 plugin runtime state so recreating the runtime container no
longer discards installed plugins' artifacts, dependency environments and
installation-private data, and stop a transient Runner registry miss during
steering from dropping the message.
- docker-compose: bind-mount ./data/plugin-runtime for the plugin runtime
alongside ./data/plugins.
- kubernetes: add the langbot-plugin-runtime-state claim (5Gi) mounted at
/app/data/plugin-runtime, kept separate from the plugin data claim so
runtime state does not land inside the plugin directory.
- chat handler: treat RunnerNotFoundError while claiming steering as
"steering unavailable" and continue with a normal run; every other error
keeps propagating. The runner error imports move to module scope, which
the new early reference requires.
- tests: cover the steering fallback, the claimed-steering interrupt, and
the lasting-gap path that still reports runner-not-found.
The steering fallback buys one extra uncached registry lookup while a
runtime reconnects; it does not recover a gap that outlasts the message.
Classify event cleanup queries by selected columns instead of CORRELATED
plan text, preserving covering-index, primary-key, and no-sort checks.
Validated all 15 tests on ARM64 with SQLite 3.53.4 and 3.46.1.
- Replace per-session mocks with lightweight shutdown counters
- Preserve the one-second timeout and 100-workspace coverage
- Assert that every session shuts down exactly once
- Add migration 0042 to enable incremental vacuum with verified backup and restore
- Enable incremental vacuum before creating tables in new SQLite databases
- Adapt reclamation batches to elapsed time and lock contention
- Cover migration recovery, WAL mode, and concurrent writes with regression tests
Emit and persist a Host cancellation event before closing the runner stream.
Preserve existing terminal events and record cancelled runs consistently.
Serialize Host event sequence allocation to prevent conflicts with admin
cancellation audits, and notify observers with the persisted sequence.
Add deterministic cancellation and concurrent event allocation regressions.
- Deduplicate message and history resource URLs using stable catalog references
- Persist renewed URLs with concurrent refresh protection
- Refresh OneBot images via get_msg intervals or NapCat rkey TTL
- Migrate historical payloads and cover SQLite and PostgreSQL behavior
- Refresh images in inputs, outputs, deliveries, and conversation history
- Reuse OneBot refresh logic without rewriting persisted records
- Cover existing image URLs across agent and pipeline execution views
