omo-ai 5.1.44 → 5.1.45
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/bin/lib/bun-bin-shim.js +53 -72
- package/bin/lib/doctor.js +2 -2
- package/bin/lib/dori.js +187 -0
- package/bin/lib/launcher.js +7 -6
- package/bin/lib/provider-map.json +1 -1
- package/bin/lib/service-auth.generated.js +4 -4
- package/bin/omo.js +3 -4
- package/bin/senpi-patch.mjs +3 -4
- package/package.json +2 -2
- package/plugin/CHANGELOG.md +155 -7
- package/plugin/extensions/dream-persona.md +1 -1
- package/plugin/extensions/gateway-rules-extension.mjs +1 -1
- package/plugin/extensions/gateway-store-worker.mjs +2 -2
- package/plugin/extensions/memory-run-supervisor.mjs +1 -1
- package/plugin/extensions/omo-claim-receipt.js +2 -2
- package/plugin/extensions/omo-computer-use.js +2 -2
- package/plugin/extensions/omo-device-account.js +2 -0
- package/plugin/extensions/omo-dori-migration.js +2 -0
- package/plugin/extensions/omo-feedback.js +2 -2
- package/plugin/extensions/omo-init-deep-advisor.js +2 -2
- package/plugin/extensions/omo-member.js +2 -2
- package/plugin/extensions/omo-memory-commands.js +11 -0
- package/plugin/extensions/omo-memory-doctor.js +3 -3
- package/plugin/extensions/omo-memory-memfs.js +3 -3
- package/plugin/extensions/omo-memory-palace.js +3 -3
- package/plugin/extensions/omo-memory-sync.js +2 -2
- package/plugin/extensions/omo-onboarding-scan.js +3 -3
- package/plugin/extensions/omo-recall-store.js +1 -1
- package/plugin/extensions/omo-session-history.js +2 -2
- package/plugin/extensions/omo-side-panel.js +3 -3
- package/plugin/extensions/omo-task.js +2 -2
- package/plugin/extensions/omo-telemetry-posthog.js +2 -2
- package/plugin/extensions/omo-telemetry-recovery.js +2 -0
- package/plugin/extensions/omo-thread-client.js +2 -2
- package/plugin/extensions/omo.js +3 -3
- package/plugin/extensions/reflection-persona.md +1 -1
- package/plugin/package.json +9 -2
- package/plugin/runtime/agent-toolkit-sdk/sdk.js +2 -2
- package/plugin/runtime/ast-grep-mcp/cli.js +6 -6
- package/plugin/runtime/ast-grep-mcp/manifest.json +2 -2
- package/plugin/runtime/category-coverage/index.js +39 -75
- package/plugin/runtime/dori/layer2.js +2 -0
- package/plugin/runtime/dori/replay.js +2 -0
- package/plugin/runtime/lsp-daemon/dist/.omo-runtime-manifest.json +4 -4
- package/plugin/runtime/lsp-daemon/dist/cli.js +4 -4
- package/plugin/runtime/lsp-daemon/dist/client.js +4 -4
- package/plugin/runtime/lsp-daemon/dist/index.js +4 -4
- package/plugin/runtime/rollback-migrate.js +2 -2
- package/plugin/runtime/task-config/index.js +1 -1
- package/plugin/runtime/thread-sdk/sdk.js +2 -2
- package/plugin/scripts/install.mjs +5 -2
- package/plugin/skills/browser/SKILL.md +21 -22
- package/plugin/skills/browser/references/commands.md +2 -9
- package/plugin/skills/browser/references/install.md +1 -2
- package/plugin/skills/browser/runtime/omowright/manifest.json +1 -1
- package/plugin/skills/browser/scripts/browser-doctor.mjs +2 -2
- package/plugin/skills/browser/scripts/browser-engine-guard.mjs +63 -487
- package/plugin/skills/browser/scripts/browser-install.mjs +1 -1
- package/plugin/skills/browser/scripts/omowright.mjs +1 -10
- package/plugin/skills/frontend/SKILL.md +79 -118
- package/plugin/skills/frontend/references/design/README.md +75 -221
- package/plugin/skills/frontend/references/design/stylegallery.md +32 -27
- package/plugin/skills/frontend/references/perfection/README.md +2 -2
- package/plugin/skills/frontend/scripts/perfection/check-accent-stripes.ts +8 -24
- package/plugin/skills/frontend/scripts/perfection/check-draft-guard.ts +40 -0
- package/plugin/skills/frontend/scripts/perfection/collect-files.ts +53 -0
- package/plugin/skills/frontend/scripts/perfection/draft-guard/findings.ts +16 -0
- package/plugin/skills/frontend/scripts/perfection/draft-guard/lint.ts +38 -0
- package/plugin/skills/frontend/scripts/perfection/draft-guard/rules.ts +69 -0
- package/plugin/skills/frontend/scripts/perfection/draft-guard/text-nodes.ts +121 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/attr-lookalikes.tsx +10 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/bound-values.tsx +12 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/const-generics.tsx +3 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/generics.tsx +7 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/measured-stats.html +6 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/plain-copy.html +9 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/negative/state-copy.html +11 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/attribute-edges.html +7 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/components.tsx +16 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/decorative-badge.vue +7 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/emoji-icons.tsx +9 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/jsx-edges.tsx +1 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/promo-copy.html +6 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/promo-copy.tsx +8 -0
- package/plugin/skills/frontend/scripts/perfection/fixtures/draft-guard/positive/vanity-stats.html +5 -0
- package/plugin/skills/programming/SKILL.md +10 -7
- package/plugin/skills/programming/references/typescript/README.md +10 -7
- package/plugin/skills/programming/references/typescript/bootstrap.md +9 -23
- package/plugin/skills/programming/references/typescript/oxc-lint.md +130 -0
- package/plugin/skills/programming/references/typescript/tsconfig-strict.md +6 -62
- package/plugin/skills/programming/scripts/typescript/check-no-excuse-rules.ts +12 -12
- package/plugin/skills/programming/scripts/typescript/new-project.ts +20 -81
- package/plugin/skills/programming/scripts/typescript/oxc/.oxfmtrc.json +9 -0
- package/plugin/skills/programming/scripts/typescript/oxc/.oxlintrc.json +46 -0
- package/plugin/skills/programming/scripts/typescript/oxc/lint/oxlint-plugin/LICENSE +21 -0
- package/plugin/skills/programming/scripts/typescript/oxc/lint/oxlint-plugin/index.js +10 -0
- package/plugin/skills/programming/scripts/typescript/oxc/lint/oxlint-plugin/no-em-dash-copy.js +28 -0
- package/plugin/skills/programming/scripts/typescript/oxc/lint/oxlint-plugin/require-suppression-reason.js +62 -0
- package/plugin/skills/programming/scripts/typescript/oxc/tsconfig.json +29 -0
- package/plugin/skills/refactor/SKILL.md +13 -13
- package/plugin/skills/ulw-execute/SKILL.md +1 -1
- package/plugin/skills/ulw-research/SKILL.md +1 -1
- package/plugin/skills/ulw-research/references/deliverable-phase.md +7 -6
- package/plugin/skills/ulw-research/references/fact-ledger-grammar.md +116 -0
- package/plugin/skills/ulw-research/references/report-gates.md +51 -3
- package/plugin/skills/ulw-research/scripts/claims-index.mjs +125 -0
- package/plugin/skills/ulw-research/scripts/contracts.mjs +68 -3
- package/plugin/skills/ulw-research/scripts/fact-ledger-kinds.mjs +136 -0
- package/plugin/skills/ulw-research/scripts/fact-ledger-schema.mjs +204 -0
- package/plugin/skills/ulw-research/scripts/fact-ledger.mjs +103 -0
- package/plugin/skills/ulw-research/scripts/fidelity-brief.mjs +145 -0
- package/plugin/skills/ulw-research/scripts/fidelity-state.mjs +63 -0
- package/plugin/skills/ulw-research/scripts/gates-static.mjs +1 -1
- package/plugin/skills/ulw-research/scripts/gates-structure.mjs +1 -1
- package/plugin/skills/ulw-research/scripts/gates-text.mjs +1 -27
- package/plugin/skills/ulw-research/scripts/lexicon.mjs +47 -0
- package/plugin/skills/ulw-research/scripts/numerals.mjs +226 -0
- package/plugin/skills/ulw-research/scripts/outcome.mjs +2 -3
- package/plugin/skills/ulw-research/scripts/particles-ko.mjs +108 -0
- package/plugin/skills/ulw-research/scripts/rational.mjs +139 -0
- package/plugin/skills/ulw-research/scripts/report-tools-commands.mjs +2 -2
- package/plugin/skills/ulw-research/scripts/sentence-split.mjs +77 -0
- package/plugin/skills/ulw-research/scripts/text-normalize.mjs +82 -0
- package/plugin/skills/ulw-research/scripts/text-similarity.mjs +28 -0
- package/plugin/skills/ulw-research/scripts/units.mjs +105 -0
- package/plugin/skills/visual-qa/SKILL.md +70 -63
- package/plugin/skills/visualize/SKILL.md +1 -1
- package/plugin/telemetry-schema.json +3857 -0
- package/plugin/telemetry-schema.omo-cloud.json +3850 -0
- package/bin/lib/gateway.js +0 -166
package/bin/omo.js
CHANGED
|
@@ -14,10 +14,9 @@ try {
|
|
|
14
14
|
// too old stops with an upgrade message, any other old bun leaves the launch on node (#9563).
|
|
15
15
|
// Only an explicit OMO_RUNTIME=node, or a machine without a usable bun, stays on node.
|
|
16
16
|
const scriptPath = fileURLToPath(import.meta.url)
|
|
17
|
-
//
|
|
18
|
-
//
|
|
19
|
-
//
|
|
20
|
-
// above and the bin mapping in package.json are all Windows shims ever read, and both stay.
|
|
17
|
+
// Prepare this package-owned entrypoint for a direct Bun boot on POSIX Bun-global installs.
|
|
18
|
+
// Keep Bun's bin symlinks, restoring only our legacy marked wrappers before commands run.
|
|
19
|
+
// Windows/npm retain the shipped Node shebang and package.json bin mapping.
|
|
21
20
|
ensureBunBinShim({ scriptPath })
|
|
22
21
|
const reexeced = await maybeReexecUnderBun({ scriptPath })
|
|
23
22
|
if (!reexeced) {
|
package/bin/senpi-patch.mjs
CHANGED
|
@@ -7,10 +7,9 @@ import { prepareInstalledEngine, preparePluginLaunchSpec, writeEnginePreparedSta
|
|
|
7
7
|
|
|
8
8
|
const packageRoot = dirname(dirname(fileURLToPath(import.meta.url)))
|
|
9
9
|
|
|
10
|
-
//
|
|
11
|
-
//
|
|
12
|
-
//
|
|
13
|
-
// fail-open, and a no-op outside a POSIX bun-global install, so it runs before the engine work.
|
|
10
|
+
// Prepare the package-owned entrypoint before a Bun-only machine's first launch (#9293).
|
|
11
|
+
// Bun owns the global bin symlinks; this restores only our legacy marked regular wrappers.
|
|
12
|
+
// It cannot undo a link error Bun recorded before postinstall began (#11019).
|
|
14
13
|
ensureBunBinShim({ scriptPath: join(packageRoot, "bin", "omo.js") })
|
|
15
14
|
const require = createRequire(join(packageRoot, "package.json"))
|
|
16
15
|
let senpiRoot = process.env.OMO_SENPI_PATCH_ROOT
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "omo-ai",
|
|
3
|
-
"version": "5.1.
|
|
3
|
+
"version": "5.1.45",
|
|
4
4
|
"type": "module",
|
|
5
5
|
"description": "OmO Native - the standalone omo command with the OMO extension built in: bun add -g omo-ai",
|
|
6
6
|
"bin": {
|
|
@@ -17,7 +17,7 @@
|
|
|
17
17
|
"postinstall": "node bin/senpi-patch.mjs"
|
|
18
18
|
},
|
|
19
19
|
"dependencies": {
|
|
20
|
-
"@code-yeongyu/senpi": "2026.10.
|
|
20
|
+
"@code-yeongyu/senpi": "2026.10.23-1"
|
|
21
21
|
},
|
|
22
22
|
"repository": {
|
|
23
23
|
"type": "git",
|
package/plugin/CHANGELOG.md
CHANGED
|
@@ -7,6 +7,161 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
7
7
|
|
|
8
8
|
## [Unreleased]
|
|
9
9
|
|
|
10
|
+
## [5.1.45] - 2026-10-24
|
|
11
|
+
|
|
12
|
+
**Reading the end of a long session takes 42 ms and 68 MB instead of 40 s and 1.8 GB.** `session_read`, `session_search` and `session_info` page through OpenCode's message API instead of loading the whole session: on a 23,000-message session, `session_read` with `limit: 2, from_end: true` dropped from 39.6 s and 1,848 MB peak memory to 42 ms and 68 MB, and a search that runs out its 60-second timeout now stops reading when it returns, where it used to read another 228 MB in the background ([#10738](https://github.com/code-yeongyu/oh-my-openagent/issues/10738), [#10741](https://github.com/code-yeongyu/oh-my-openagent/pull/10741)).
|
|
13
|
+
|
|
14
|
+
**OmO's extension no longer carries an LSP engine it never runs.** The native LSP tools and post-edit diagnostics already go through the packaged LSP daemon, but the formatter's import pulled lsp-core's whole in-process engine into `omo.js`; importing only the helpers it uses shrinks the extension by about 82 KB, with no change in behavior. ([#10659](https://github.com/code-yeongyu/oh-my-openagent/pull/10659))
|
|
15
|
+
|
|
16
|
+
**Every session start reads about 33 KB less of `omo.js`: eleven memory slash commands and memory sync's device account load on first use ([#10902](https://github.com/code-yeongyu/oh-my-openagent/pull/10902), [#10860](https://github.com/code-yeongyu/oh-my-openagent/issues/10860)).** The commands (`/memory`, `/remember`, `/init`, `/recompile`, `/memory-repository`, `/sleeptime`, `/reflect`, `/dream`, `/search`, `/facts`, `/agreements`) still appear at startup with the same names, descriptions and hints; their handlers load from `extensions/omo-memory-commands.js`. The signed-in device account loads from `extensions/omo-device-account.js` the first time a session resolves its sync, and if that file cannot load, memory sync reports `sync-unavailable` with the reason instead of looking signed out.
|
|
17
|
+
|
|
18
|
+
**ulw-research ships a fact ledger, a fidelity state file and a claim index for research reports.** They give each research run a machine-readable record of its facts and verified claims; nothing in the CLI uses them yet. ([#11046](https://github.com/code-yeongyu/oh-my-openagent/pull/11046), part of [#11009](https://github.com/code-yeongyu/oh-my-openagent/issues/11009); batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
19
|
+
|
|
20
|
+
**Windows dori sidecar paths pass the extension's path checks.** The omo-senpi tests accept the Windows form of the sidecar path, so the Windows suite no longer fails on a correct path. ([#11111](https://github.com/code-yeongyu/oh-my-openagent/pull/11111), fixes [#11110](https://github.com/code-yeongyu/oh-my-openagent/issues/11110))
|
|
21
|
+
|
|
22
|
+
**Child-task, workflow and workpool wake sources report kind `children`.** ([#11114](https://github.com/code-yeongyu/oh-my-openagent/pull/11114), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782))
|
|
23
|
+
|
|
24
|
+
**An unreadable workpool store no longer drops or falsely clears the children wake source.** ([#11118](https://github.com/code-yeongyu/oh-my-openagent/pull/11118), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782))
|
|
25
|
+
|
|
26
|
+
**The children wake source reads the workpool store once per snapshot.** ([#11124](https://github.com/code-yeongyu/oh-my-openagent/pull/11124), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
27
|
+
|
|
28
|
+
**Child results waiting in the wake queue keep the parent from reading as done.** ([#11115](https://github.com/code-yeongyu/oh-my-openagent/pull/11115), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782))
|
|
29
|
+
|
|
30
|
+
**The wake-queue count can no longer cost a delivery receipt and is re-published per session generation.** ([#11121](https://github.com/code-yeongyu/oh-my-openagent/pull/11121), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782))
|
|
31
|
+
|
|
32
|
+
**A receipt that throws no longer leaves the wake-queue count up.** ([#11122](https://github.com/code-yeongyu/oh-my-openagent/pull/11122), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782))
|
|
33
|
+
|
|
34
|
+
**Delivery receipts are isolated from each other, on retire too, and no receipt error is dropped.** ([#11123](https://github.com/code-yeongyu/oh-my-openagent/pull/11123), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
35
|
+
|
|
36
|
+
**A wake-queue flush settles each receipt and the child count once, and a throwing `onReceiptError` no longer stops the rest.** ([#11137](https://github.com/code-yeongyu/oh-my-openagent/pull/11137), part of [#10782](https://github.com/code-yeongyu/oh-my-openagent/issues/10782)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
37
|
+
|
|
38
|
+
**Windows helper tree kills are guarded against recycled parent PIDs.** A QA helper no longer stops an unrelated process tree that reused a parent's PID. ([#11007](https://github.com/code-yeongyu/oh-my-openagent/pull/11007))
|
|
39
|
+
|
|
40
|
+
**A task that a quitting session suspended says why it is still running.** Quitting the owning session parks every live child, but the parked record kept its bare `running` status with no reason, so a restart that never resumed that session read as a running task that nothing would ever move. The quit shutdown now records `session_quit` on the parked record, which `task_output` and the task snapshot present as "resuming (session quit)", and a later resume whose reconcile defers the revival replaces it with the deferral's own reason. ([senpi#3493](https://github.com/code-yeongyu/senpi/issues/3493), [#10533](https://github.com/code-yeongyu/oh-my-openagent/pull/10533), thanks [@Bearmancer](https://github.com/Bearmancer))
|
|
41
|
+
|
|
42
|
+
**A project's skills can no longer reach outside the project, and their MCP servers no longer get your whole environment.** A skill link in a project that points outside it is refused, and the reason shows in the skill tool's result and in OpenCode; links into skill folders you configured yourself still work. MCP servers that come from a project's skills start with a small safe base (paths, home, temp, locale, CA paths) plus the variables they declare. When one asks for a variable it didn't get, the result names it with the exact `/omo-mcp-env-grant` command that saves the grant for that server declaration; only names are stored, in the same file the engine uses. Skills you installed yourself, including paths in your global OpenCode configuration, keep today's environment. Task discovery keeps nested reference documents out of skills and allows links between packages in the same repository. Unreadable or missing skills report their actual file error rather than a security block. Replaced or evicted grant commands expire, and reopening the server provides a fresh command when needed. ([#10716](https://github.com/code-yeongyu/oh-my-openagent/pull/10716))
|
|
43
|
+
|
|
44
|
+
**Computer use explains a refused macOS launch context instead of reporting a closed engine.** Initial activation and later restarts now report `COMPUTER_LAUNCH_CONTEXT_REQUIRED` with recovery instructions when the actual engine child refuses to start. Permission errors, desktop handover, and the native launch gate keep their existing behavior. ([#11073](https://github.com/code-yeongyu/oh-my-openagent/issues/11073), [#11086](https://github.com/code-yeongyu/oh-my-openagent/pull/11086))
|
|
45
|
+
|
|
46
|
+
**Thread tools check that an RPC host is the one they selected before changing a session ([#10712](https://github.com/code-yeongyu/oh-my-openagent/pull/10712)).** The client proves the host's key with a challenge-response and attaches the selected session by its durable id on the same connection, so a host that reused a routing id cannot receive another session's mutation. A draining host answers with a retryable `host_draining` error instead of failing the call, and a change that took effect is never reported as failed.
|
|
47
|
+
|
|
48
|
+
**The build now records which files each bundled artifact is built from.** `packages/omo-senpi/bundle-inputs.json` lists every repo file the generators of the committed bundles read. The list is traced from the recipe itself, and CI checks that it is current. Merge tooling can now tell that a skill-only change leaves the bundles untouched ([#11079](https://github.com/code-yeongyu/oh-my-openagent/issues/11079), [#11084](https://github.com/code-yeongyu/oh-my-openagent/pull/11084)).
|
|
49
|
+
|
|
50
|
+
**A session with telemetry off clears leftover session summaries as soon as it starts.** Summary files a crashed session left behind used to stay on disk until a later session with telemetry on picked them up. A session with telemetry off now deletes them at startup when `omo.json` turns telemetry off for their project; an opt-out set only through an environment variable such as `DO_NOT_TRACK` applies to that session alone and leaves other sessions' summaries in place. Turning telemetry off in `omo.json` takes effect on the very next turn instead of up to five seconds later. A summary whose project cannot be told apart is never sent. ([#10234](https://github.com/code-yeongyu/oh-my-openagent/issues/10234)) ([#10267](https://github.com/code-yeongyu/oh-my-openagent/pull/10267)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
51
|
+
|
|
52
|
+
**Session summaries survive a crash at any step and are never counted twice.** Each summary window now has its own id, sent as the event id, and is written to a local pending file before it is sent; the file is deleted only after delivery is confirmed, and a later OmO start re-sends anything left over under the same id. A window left open by a killed process is sent once, only after that process is gone and its once-a-minute heartbeat has stopped for 5 minutes, so a sleeping laptop keeps its own window. The session sequence survives restarts and counts within a random per-run `summary_epoch`, and a session continued or recovered by another process starts a new one, so two processes writing one session never send the same (epoch, sequence) pair even when one of them is wrongly taken for dead, costs are added as whole 1/10,000 USD units (`cost_1e4`) so summaries and raw turns agree exactly, and `dropped_events` counts what had to be discarded after 14 days. `telemetry.rawTurnSampleRate` (or `OMO_TELEMETRY_RAW_TURN_SAMPLE_RATE`) sets the share of sessions that also send raw turn rows, 0.05 by default. ([#10198](https://github.com/code-yeongyu/oh-my-openagent/issues/10198), [#10225](https://github.com/code-yeongyu/oh-my-openagent/pull/10225)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
53
|
+
|
|
54
|
+
**Every session event says whether it came from a session you opened or one OmO spawned.** `session_started`, `turn_completed` and `session_summary` carry `session_role` (`root`, `child` or `unknown`), so a delegated child's usage is not counted again on top of its parent. ([#10198](https://github.com/code-yeongyu/oh-my-openagent/issues/10198), [#10225](https://github.com/code-yeongyu/oh-my-openagent/pull/10225)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
55
|
+
|
|
56
|
+
**Raw turn rows keep the name of a public model reached through your own gateway.** `turn_completed` masked the model to `custom` whenever the provider was not one OmO ships, while `session_summary` and every other event already kept a published model id regardless of the routing provider. The gateway name itself is still never sent. ([#10198](https://github.com/code-yeongyu/oh-my-openagent/issues/10198), [#10225](https://github.com/code-yeongyu/oh-my-openagent/pull/10225)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
57
|
+
|
|
58
|
+
**Plan review keeps working after a restart or a restored session ([#10558](https://github.com/code-yeongyu/oh-my-openagent/pull/10558)).** plan-consultant and plan-reviewer only spawn after you asked for ulw-plan and a plan file exists. That state used to live only in memory, so after a restart, a desktop restore or a resumed session the gate was empty and plan review was refused until you asked again. The gate is now rebuilt from the session's own transcript when it starts: plan file reads and writes count again, and a request counts only when the engine recorded you typed it (senpi 2026.10.21-2 and later), so text injected by an extension still never opens it. State handed over by a reload is kept as well (#10378).
|
|
59
|
+
|
|
60
|
+
**The engine moves to senpi 2026.10.23-1: scheduled prompts and `/loop` ticks reach an open session mid-turn, the footer shows what subagents spent, and compaction works again on very long sessions and turns ([#11126](https://github.com/code-yeongyu/oh-my-openagent/pull/11126)).** A scheduled prompt now arrives in its open session, steered into the running turn when the agent is working, and a `/loop` tick that comes due mid-turn is read at the next step. Each session remembers the thinking level per model. Stop pauses the session until you type or `/resume`. An HTTP MCP server with no saved sign-in no longer freezes the process for up to 30 seconds, a background usage check no longer locks a healthy Claude login on a network error, and a `models.json` entry without a price keeps the built-in price instead of billing $0. The built-in `/llama` llama.cpp integration is removed upstream; a saved `llama.cpp` provider is ignored.
|
|
61
|
+
|
|
62
|
+
**Bun-installed `omo` keeps its command links intact so later upgrades can succeed.** Launches still start directly on Bun, and a new-version launch restores older generated wrappers to symlinks. This fixes recurring `EEXIST` failures on Bun 1.4.4; the first upgrade from an old wrapper may need the one-time removal described in the PR. ([#11024](https://github.com/code-yeongyu/oh-my-openagent/pull/11024))
|
|
63
|
+
|
|
64
|
+
**The engine moves to senpi 2026.10.21-3: long 1M-context conversations stop re-writing their cached prefix, Anthropic subscription context is counted once, and "No credential slots available" no longer locks every account out ([#10965](https://github.com/code-yeongyu/oh-my-openagent/pull/10965)).** When context reduction trims a long conversation on a 1M window, the reduced prefix now stays identical and keeps its cache checkpoint, so each step writes only what changed. Anthropic subscription turns with internal SDK re-requests no longer double the context percentage or start compaction early, and a configured compaction model now does threshold, speculative and idle compaction while your session provider is healthy. The subscription lane is the only layer that picks its accounts, so one revoked or briefly rate-limited account no longer blocks the rest until you log in again. Every turn end says what happens next, a turn-end report names the hold your queued message released, the trailing "Anything else? (optional)" card is gone from clients without native questions, and moving a session between machines runs through one implementation. Forced recovery after a context overflow no longer picks the same compaction cut again when an invisible entry leads the session; an overflow still never moves a task to its next fallback model.
|
|
65
|
+
|
|
66
|
+
**Tasks honour a `-fast` model alias that the provider config pairs with a model.** Kibitzer no longer stops at a model whose start is refused; it falls back to the next one. ([#10538](https://github.com/code-yeongyu/oh-my-openagent/pull/10538), fixes [#10536](https://github.com/code-yeongyu/oh-my-openagent/issues/10536))
|
|
67
|
+
|
|
68
|
+
**Child startup failures keep their cause.** When a task child exits while the parent writes to stdin, its own typed failure is reported instead of a broken-pipe error. ([#11097](https://github.com/code-yeongyu/oh-my-openagent/pull/11097), fixes [#11096](https://github.com/code-yeongyu/oh-my-openagent/issues/11096))
|
|
69
|
+
|
|
70
|
+
**The side panel says, once, that its column needs the fullscreen TUI.** `side_panel.enabled` renders its right-hand column only in the fullscreen TUI (`--tui-mode fullscreen`, or `tuiMode` in senpi's settings); in the default regular mode the sections stay a block above the editor. The docs now say so at the top of the `side_panel` table, and the first regular-mode run with the panel on shows a one-time notice naming both ways to switch. ([#10560](https://github.com/code-yeongyu/oh-my-openagent/issues/10560), [#10565](https://github.com/code-yeongyu/oh-my-openagent/pull/10565))
|
|
71
|
+
|
|
72
|
+
**A background task that wakes itself up, or reconnects after the task host restarts, keeps the right finished/running mark on the host ([#11094](https://github.com/code-yeongyu/oh-my-openagent/pull/11094)).** A task that started a new run on its own just as its previous turn finished could stay marked as finished, so a replaced task host might drop it while it was still working. A finished task whose host restarted could lose its finished mark, so the new host might reopen it. Both now keep the correct mark. ([#11092](https://github.com/code-yeongyu/oh-my-openagent/issues/11092))
|
|
73
|
+
|
|
74
|
+
**The side panel shows subscription usage when you are signed in but the engine is serving through an environment token ([#10564](https://github.com/code-yeongyu/oh-my-openagent/issues/10564)).** It used to show nothing in that case; it still never shows a different signed-in account's numbers. ([#11101](https://github.com/code-yeongyu/oh-my-openagent/pull/11101))
|
|
75
|
+
|
|
76
|
+
**The one-line Linux installer picks the glibc build on glibc hosts that also have the `musl` package installed.** `install.sh` used to treat any `/lib/ld-musl-*.so.1` loader as proof of a musl system, so Debian or Ubuntu hosts with `musl-tools` or `musl-dev` installed downloaded `omo-linux-x64-musl` and recorded it for later updates. A glibc host (one where `getconf GNU_LIBC_VERSION` answers) now always gets the glibc build. The musl check also reads `ldd --version` in a way that still matches under `pipefail`, since musl's `ldd` exits 1 after printing its version. ([#10562](https://github.com/code-yeongyu/oh-my-openagent/issues/10562), [#10612](https://github.com/code-yeongyu/oh-my-openagent/pull/10612))
|
|
77
|
+
|
|
78
|
+
**`bun run lint` runs the copy check again, and Bun no longer warns on every process in the repo.** The root `package.json` had two `"lint"` scripts after a merge, so the copy check was silently skipped and Bun printed a "Duplicate key" warning to stderr each time it started, which failed tests that expect a quiet child. One `lint` script now runs both checks, and a test rejects repeated keys in any `package.json`. ([#11108](https://github.com/code-yeongyu/oh-my-openagent/pull/11108))
|
|
79
|
+
|
|
80
|
+
**Memory notes keep their exact line endings on Windows.** Git for Windows converts line endings by default, so a note committed with LF or mixed endings came back with CRLF in a reflection worktree, and the description preflight that must keep every non-description byte failed there. Every memory repository now pins `core.autocrlf=false`, and opening an older repository restores only the working copies whose sole difference from the committed note is the converted line endings. ([#11112](https://github.com/code-yeongyu/oh-my-openagent/pull/11112), fixes [#11109](https://github.com/code-yeongyu/oh-my-openagent/issues/11109)) (batch [#11148](https://github.com/code-yeongyu/oh-my-openagent/pull/11148))
|
|
81
|
+
|
|
82
|
+
**Memory reflection and dream runs trim oversized note descriptions before completion validation.** The note body is preserved, instead of the run failing again and again and parking the identity. ([#10999](https://github.com/code-yeongyu/oh-my-openagent/pull/10999))
|
|
83
|
+
|
|
84
|
+
**The frontend skill now decides what a screen leaves out before drawing it, and visual QA proves a redesign improved and stops after a few rounds ([#11023](https://github.com/code-yeongyu/oh-my-openagent/pull/11023)).** Before code, every screen names one primary action, its information priorities, what it leaves out on purpose, and the content it must keep; after the draft renders, a subtraction pass removes what the user does not need and checks nothing required was dropped or merged. Small changes now read the project's design file and at most one reference instead of the whole library, and atmosphere and award-grade styling are built only when the brief asks. Visual QA gains a before/after mode that captures fixed cases before the first edit and compares them side by side, and its review loop now ends after 2-3 rounds: layout, function, accessibility and missing-content defects earn another round, taste notes are recorded and stop it.
|
|
85
|
+
|
|
86
|
+
**Computer use refuses input while the stop chord cannot be heard.** On macOS, a password field or Secure Keyboard Entry turns on Secure Event Input, which hides key presses from the stop chord's listener; computer use kept running anyway, so the chord could not stop it. Input is now refused with `StopPathUnavailable` (`secure-input-active`) until secure input ends, and resumes by itself. A `type` in progress now stops at the next character when the stop path stops authorizing input, instead of finishing the text, and the Linux X11 listener refuses to arm on an XInput 2.0 server, where a keyboard grab would swallow the chord. Plain Esc stays an ordinary key. ([#10534](https://github.com/code-yeongyu/oh-my-openagent/issues/10534), [#9651](https://github.com/code-yeongyu/oh-my-openagent/issues/9651), [#10537](https://github.com/code-yeongyu/oh-my-openagent/pull/10537))
|
|
87
|
+
|
|
88
|
+
**The frontend draft guard now catches promotional copy, emoji icons, hype badges and vanity stats, not only accent stripes ([#11022](https://github.com/code-yeongyu/oh-my-openagent/pull/11022)).** One command, `check-draft-guard.ts`, runs the accent-stripe rules plus four checks on visible markup text: promotional filler in English and Korean copy, an emoji used as an icon, a badge whose only content is a hype label such as "New" or "AI-powered", and a stat tile holding a marketing-form number such as "10x", "500+" or "24/7"; a measured figure such as "42%" or "1.2k" passes. Code (TypeScript generics and comparisons included), scripts, styles, comments, hidden elements, bound values and badges that carry a state attribute are not flagged. React component tags such as `<Badge>` and `<Stat value="10x">`, `cn()` class lists, and announced attributes such as `aria-label` and `placeholder` are checked too. The frontend and visualize skills now run it before visual QA in place of `check-accent-stripes.ts`, which still works on its own. Both commands now skip a dangling symlink in a scanned folder instead of crashing, and report an unreadable path given on the command line with exit 2.
|
|
89
|
+
|
|
90
|
+
**A memory sync that stops working now says so in the TUI, with what to do.** When a sync cycle failed, the error went only to the log, and the status kept reporting the last successful sync. Now the memory footer shows `sync blocked` while sync is failing, and one notice says why and what to do: `/memory unlock <path>` for a git lock left behind by a crashed writer, sign in again for an expired sign-in, or check your connection when the sync service cannot be reached. A single dropped request still retries quietly; the offline notice appears only after three failed cycles in a row. When sync recovers, a notice says so and the footer clears. ([#10637](https://github.com/code-yeongyu/oh-my-openagent/issues/10637), [#10958](https://github.com/code-yeongyu/oh-my-openagent/pull/10958))
|
|
91
|
+
**Agents can now reach the dori extension through the one toolkit SDK, and a retried turn never repeats a call that already ran.** With the `dori` section on, the toolkit gains its route, reply, room, send, thread, describe, file, rules, iam, call and work namespaces, and `help()` lists them with their signatures; with it off they do not exist and no dori code loads. Every call carries the caller the engine stamped and a replay key, so a replayed turn gets the recorded result back, a replay with different arguments stops with a typed error, and one older than seven days is refused instead of running again. A session waiting on a dori request also wakes when a macOS file watch drops the wake event, instead of sleeping until its timeout. ([#10458](https://github.com/code-yeongyu/oh-my-openagent/pull/10458))
|
|
92
|
+
|
|
93
|
+
**Store extensions can expose session-callable ops as real tools, and their worker no longer drags a schema library into every store.** A store extension's `sessionCallable` declaration can now carry a `toolName`, `description` and JSON Schema `parameters` per op: the session's engine registers it as `ext_<extension>_<toolName>`, the tool takes the calling session's durable id from the engine (never from arguments), and the worker stamps `caller_session_durable_id` and, when the caller thread_create'd the target, `caller_created_target` inside the op's own transaction. Args naming either caller field are refused before validation; the public `extensionCall` channel refuses every declared session op. The worker validates op arguments itself against a JSON Schema subset (`type`, `properties`, `required`, `additionalProperties` true/false, `enum`, `const`, `pattern`, length/item/number bounds, `items`, `anyOf`), and `register()` refuses any other keyword, so no constraint is skipped - which also removed typebox/value from the store worker: its bundle drops from 206 KB to 115 KB and every worker spawn (and idle-retire respawn) stops paying for its module graph. A declared `await` with `statusOp`/`expireOp`/`timeoutMs` arms a wake-file wait, and `thread_creations` records who created which session durably across host restarts. ([#9551](https://github.com/code-yeongyu/oh-my-openagent/pull/9551))
|
|
94
|
+
|
|
95
|
+
**Product copy uses clearer punctuation instead of em dashes ([#10994](https://github.com/code-yeongyu/oh-my-openagent/pull/10994)).** Agent prompts, notices, memory recall openers, and web metadata now use commas, colons, or parentheses, and an empty-value placeholder uses a hyphen. A source AST check prevents new em dashes in product strings, template elements, and JSX text while leaving comments, regexes, tests, generated code, and vendored sources alone. The same check now requires lint and TypeScript suppressions to explain why, inline or in the comment directly above, including Biome ignores.
|
|
96
|
+
|
|
97
|
+
**Inside OmO Desktop the browser skill works as in the terminal, plus the app's built-in browser.** The agent's browsers (your own through BrowserSkill, or one it starts itself) are no longer refused or changed inside the app, and `connectBuiltinBrowser()` adds the thread's tab in the app's built-in browser, which you can watch beside the chat ([#11066](https://github.com/code-yeongyu/oh-my-openagent/pull/11066)).
|
|
98
|
+
|
|
99
|
+
**A session in the OmO desktop app's Built-in browser mode can ask the app for its in-app browser.** The browser skill told the agent to use the app's in-app browser tools, but an OmO session had none, so agents reached for whatever outside browser the machine had (sisyphuslabs/omo-desktop-app#3292). A new eval-only `omo_desktop_toolkit` tool asks the app for a toolkit it hosts for the session's thread, and `connectBuiltinBrowser()` in the browser skill uses it to attach OmOWright to that thread's tab. The app's grant names a loopback endpoint and a token file, so the token never appears in a tool result; an app that offers no toolkit is answered at once, and one that does not answer within 15 s gets a timeout. Works once the desktop app ships its side. ([#10602](https://github.com/code-yeongyu/oh-my-openagent/pull/10602), fixes [#10601](https://github.com/code-yeongyu/oh-my-openagent/issues/10601))
|
|
100
|
+
|
|
101
|
+
**Parked facts leave the active queue without losing their contents.** Existing backlogs migrate once, manual retry restores them, and turn-end scans read only active entries. ([#11038](https://github.com/code-yeongyu/oh-my-openagent/pull/11038))
|
|
102
|
+
|
|
103
|
+
**The side panel's usage section reads the subscription limits the engine already keeps, so it shows the same numbers as the desktop, before the first turn, and makes no network request of its own ([#10553](https://github.com/code-yeongyu/oh-my-openagent/pull/10553)).** senpi now polls each signed-in Claude and ChatGPT subscription itself ([code-yeongyu/senpi#3687](https://github.com/code-yeongyu/senpi/pull/3687)); the panel only picks the serving account and draws its record. A rotation to an account the engine has not read yet shows nothing rather than the previous account's numbers. `side_panel.usage_poll_seconds` is retired: still accepted, now ignored.
|
|
104
|
+
|
|
105
|
+
**A finished background task no longer wakes up on its own after the task host restarts ([#11083](https://github.com/code-yeongyu/oh-my-openagent/pull/11083)).** A task that ran on the shared task host kept its session loaded there after it finished, with nothing saying it was done. When the host was replaced, the new host reopened it, and the task's own follow-ups could run again with nobody watching, spending usage. The task now marks its session as finished on the host as soon as it finishes, and again before it lets go, so the new host leaves it closed even if the parent session crashed. A finished task you continue with `task_send` is marked running again before your message is delivered; a dropped connection that reconnects does not change the mark. ([#10778](https://github.com/code-yeongyu/oh-my-openagent/issues/10778))
|
|
106
|
+
|
|
107
|
+
**The web-terminal visual QA driver can wait for text before it types.** A slow-starting TUI used to lose its first keystrokes to the fixed start delay. Put `{WaitFor:<text>}` in `--input` and the driver polls the rendered xterm buffer until `<text>` appears, then continues; if it never appears within `--wait-for-timeout-ms` (default 10000) the capture fails with an error naming the text and the timeout, and an empty `{WaitFor:}` is refused. Thanks to @cynkai. ([#9583](https://github.com/code-yeongyu/oh-my-openagent/issues/9583), [#9666](https://github.com/code-yeongyu/oh-my-openagent/pull/9666))
|
|
108
|
+
|
|
109
|
+
**git-master credits sisyphus-dev-ai on its commits again, once per commit, and you can turn it off.** Every commit git-master writes ends with `Co-authored-by: sisyphus-dev-ai <sisyphus-dev-ai@users.noreply.github.com>`. Git adds the trailer itself (`git -c trailer.ifexists=addIfDifferent commit --trailer`), so amending or rewording a commit that already carries it keeps a single copy, and after a squash git-master rewrites the joined message once. This needs git 2.32 or newer. `git_master.include_co_authored_by` in `omo.json` is a working switch again and defaults to `true`; set it to `false` to leave the trailer out. The commit footer stays opt-in, your own git author and committer are unchanged, and the `GIT_MASTER=1` prefix applies to the commit command as before. This works the same in the Native harness (reading the skill or `load_skills`) and the OpenCode plugin ([#10713](https://github.com/code-yeongyu/oh-my-openagent/issues/10713), [#10715](https://github.com/code-yeongyu/oh-my-openagent/pull/10715)).
|
|
110
|
+
|
|
111
|
+
**The count of undelivered messages matches the messages actually dropped ([#10987](https://github.com/code-yeongyu/oh-my-openagent/pull/10987)).** When a stopped parent cancelled a task, or a task that could not be resumed was closed, its queued messages were dropped with a note saying how many. That count came from a read taken just before the queue was cleared, so a message sent in between was dropped without being counted. The count now comes from the same write that clears the queue. ([#10866](https://github.com/code-yeongyu/oh-my-openagent/issues/10866))
|
|
112
|
+
|
|
113
|
+
**A relaunched task records its own session id, not the previous run's ([#10985](https://github.com/code-yeongyu/oh-my-openagent/pull/10985)).** When a task restarted on a new transcript (a model fallback or a revive) and its new session's first state read failed, the task kept the previous run's session id, so anything that looks a task up by its session id, such as the task tree or the cleanup when a parent stops, found the old session. The first id the new run learns now replaces it. ([#10892](https://github.com/code-yeongyu/oh-my-openagent/issues/10892))
|
|
114
|
+
|
|
115
|
+
**The footer shows what your subagents are spending, live, next to the session's own cost ([#10730](https://github.com/code-yeongyu/oh-my-openagent/pull/10730)).** The session cost only counted the session itself, so the number stood still while subagents worked. The footer now adds the whole subagent tree's spend as `(+$X)`, including nested children and revived runs, and updates as tasks report instead of polling ([code-yeongyu/senpi#2471](https://github.com/code-yeongyu/senpi/issues/2471)).
|
|
116
|
+
|
|
117
|
+
**Subagent rows no longer show `$0.0000` when the provider sent no usage ([#10727](https://github.com/code-yeongyu/oh-my-openagent/pull/10727)).** A turn whose provider streamed no usage block came back with an all-zero usage, and the task row counted it as a real $0, with no tok/s. Such a turn now counts as "cost not reported", so the row shows no amount, and a run where only some turns reported a cost is marked `partial` ([#10724](https://github.com/code-yeongyu/oh-my-openagent/issues/10724)).
|
|
118
|
+
|
|
119
|
+
**Internal: a telemetry spool lock this process could not delete no longer blocks its batch until the 14-day age limit.** After a transient file error (for example on Windows), the sweep reclaims the lock once the claim lease has passed, while a lock for a move still in progress is kept ([#10951](https://github.com/code-yeongyu/oh-my-openagent/pull/10951)).
|
|
120
|
+
|
|
121
|
+
**Internal: every assistant turn opens one `omo.turn` OpenTelemetry span, and events emitted during the turn carry its trace and span ids.** On a transport with a span sink (the OTLP collector transport; the PostHog transport has none) the Senpi adapter starts the span at `turn_start` and ends it `ok`, or `error` for a failed or abandoned turn, so a `turn_completed` row, its span and an error event join on the (`trace_id`, `span_id`) pair. A process child launched by the task engine continues the delegating turn's trace. The `traceparent` header goes only to our own hosts, never to a model provider, and nothing is started or written while telemetry is off. ([#10358](https://github.com/code-yeongyu/oh-my-openagent/pull/10358))
|
|
122
|
+
|
|
123
|
+
**Internal: the telemetry spool recovers and bounds itself on every sweep, and a stalled sender never re-sends a batch it lost.** Every live collector transport sweeps the spool on start and every 60 s: it recovers expired claims and the files of dead processes, applies the 5 MB cap to everything in the directory (not only to what it sealed itself), and replays oldest first; when consent has ended the sweep deletes the spool instead. Before every send attempt the sender renames its claim to a fresh lease, so a process that stalled past its lease and lost the batch to another process stops instead of sending it again, and a batch claimed when consent is withdrawn is deleted unsent. Nothing is wired to a product yet. ([#10355](https://github.com/code-yeongyu/oh-my-openagent/pull/10355))
|
|
124
|
+
|
|
125
|
+
**The engine publishes its OpenTelemetry schema artifact.** `telemetry-schema.json` describes exactly what an OTel client of the engine may send, generated from the event schemas and kept byte-exact by a drift gate, so the telemetry collector can pin and enforce it. It is separate from the anonymous PostHog events, which are unchanged ([#10980](https://github.com/code-yeongyu/oh-my-openagent/pull/10980)).
|
|
126
|
+
|
|
127
|
+
**The frontend skill picks a layout pattern for every layout change, and visual-qa checks it.** Any change to a screen's structure (a new region, panel, split, list, sticky bar or scroll container) now fetches the matching StyleGallery pattern and writes one `Layout:` line in `DESIGN.md` naming the pattern, the element that owns the scroll and the pattern's anti-patterns, including in a project that already has a `DESIGN.md`. visual-qa's new A12 item checks the build against that line, with a probe for the element that really scrolls. Apple HIG hit targets, hierarchy, states, keyboard and motion are now rules for the first draft in the frontend skill, not only items in the QA checklist. ([#10684](https://github.com/code-yeongyu/oh-my-openagent/issues/10684), [#10685](https://github.com/code-yeongyu/oh-my-openagent/pull/10685))
|
|
128
|
+
|
|
129
|
+
**New TypeScript and JavaScript projects default to oxlint and oxfmt.** The programming skill now ships strict OXC configs and the t3code copy rules: product strings, templates, and JSX reject em dashes, and lint/type suppressions require an explanation. Existing repository toolchains are respected. ([#11030](https://github.com/code-yeongyu/oh-my-openagent/pull/11030))
|
|
130
|
+
|
|
131
|
+
**Account commands no longer repeat a failed keychain operation while reporting the error.** A native storage failure or timeout now returns exit 9 with `{"error":"store-unavailable"}` instead of attempting another store read, which could add a second 25-second wait. Existing success and locked-keychain outcomes are preserved; no new deadline or lock change is introduced ([#11040](https://github.com/code-yeongyu/oh-my-openagent/issues/11040), [#11049](https://github.com/code-yeongyu/oh-my-openagent/pull/11049)).
|
|
132
|
+
|
|
133
|
+
**A removed device ends its local account session on refresh.** When the service refuses a renewal with `device_revoked`, `omo account token` clears the saved credential and returns exit 5 with that code; `omo account show` then reports signed out. Unknown refusal codes still retain the credential as uncertain ([#11041](https://github.com/code-yeongyu/oh-my-openagent/issues/11041), [#11048](https://github.com/code-yeongyu/oh-my-openagent/pull/11048)).
|
|
134
|
+
|
|
135
|
+
**Project memory reads work on Windows when a path uses a short user-directory name.** Read grants now compare native canonical paths on both sides, accounting for Windows case and separator differences without granting sibling memory repositories. A granted repository that is later replaced by a link elsewhere is no longer covered, and a note that does not exist yet is still allowed so the read can report it missing. ([#10982](https://github.com/code-yeongyu/oh-my-openagent/pull/10982))
|
|
136
|
+
|
|
137
|
+
**Each model family's current default is named once.** Every builtin default, fallback rung, category gate, Claude Code alias and migration target that means "the newest Opus / Sonnet / Haiku / Fable / GPT / GLM / Kimi / Gemini" now reads its id from one table, so the next model bump is one edit. Resolved models stay the same, with one exception: the Claude Code `sonnet` alias now means the current Sonnet (`claude-sonnet-5-5`) instead of `claude-sonnet-4-6`. ([#10483](https://github.com/code-yeongyu/oh-my-openagent/pull/10483), [#10141](https://github.com/code-yeongyu/oh-my-openagent/issues/10141))
|
|
138
|
+
|
|
139
|
+
**A builtin model profile can run on your own provider.** When the only provider serving a profile's model is one you declared in `models.json` (for example an API-key proxy serving `claude-opus-5-5`), Daily · Normal and the other builtin lanes now pick it instead of reporting that no model is available. Only the models you declared for that provider count, the lane's own providers still come first, and gateways you never declared still never serve a builtin lane. Reported by 채완. ([#10493](https://github.com/code-yeongyu/oh-my-openagent/pull/10493), [#10485](https://github.com/code-yeongyu/oh-my-openagent/issues/10485))
|
|
140
|
+
|
|
141
|
+
**A config with escaped non-ASCII text no longer loads the gateway -> dori migration at every start.** The startup check now looks only for `\u` escapes of the letters in `gateway`, so a prompt written with escaped characters no longer imports the migration sidecar on each launch; a legacy `gateway` section, however it is spelled, is still migrated. ([#11051](https://github.com/code-yeongyu/oh-my-openagent/pull/11051), fixes [#10910](https://github.com/code-yeongyu/oh-my-openagent/issues/10910))
|
|
142
|
+
|
|
143
|
+
**`omo dori` loads a separately installed `dori-extension` package.** The loader verifies its package identity, host contract and resolved entry path before importing it. Existing `gateway` configuration migrates to `dori`, and `omo gateway` remains a deprecated alias for one release. Without a `dori` section, doctor adds no extension rows. ([#10315](https://github.com/code-yeongyu/oh-my-openagent/issues/10315), [#10421](https://github.com/code-yeongyu/oh-my-openagent/pull/10421))
|
|
144
|
+
|
|
145
|
+
**`/mass-ulw` and `run mass ulw ...` now enter the mode instead of offering a pointer the model could skip.** The slash command (`/mass-ulw`, `/mulw`, `/ulw-research`) or a prompt that starts with `run`, `start`, `use`, `do` or `execute` followed by the keyword (`run mass ulw research on <link>`; in Korean, `mass ulw 돌려줘` or `실행해줘`) now carries a binding instruction: read the skill before any other tool call, and run mass-ulw as workflow runs. Before, the hint asked the model to decide whether you meant it, and in a real run it skipped mass-ulw and started no workflow. Any other mention, including a prompt that only ends with `mass ulw research` or asks a question, still gets the hint, and quoted, relayed, negated and delegated mentions never arm it. A single-quoted name (`'omo-senpi ultrawork ...'`) and a relay header naming several recipients (`[a -> b + c]`) now count as quotes and relays for `ulw` too, so they no longer arm ultrawork ([#10810](https://github.com/code-yeongyu/oh-my-openagent/issues/10810), [#10822](https://github.com/code-yeongyu/oh-my-openagent/pull/10822)).
|
|
146
|
+
|
|
147
|
+
**A session can consult the memory of a project it touches.** The first successful read, edit or write under a child project with existing memory lists its note paths and descriptions, without injecting note bodies. Linked projects in a non-git workspace and linked worktrees follow the session's `auto_identity` mode. Kibitzer ranks up to three recently touched projects separately from the session's own notes, reserves up to half its seats for them, and uses the same persisted search-term cache for each repository. Project notes receive read-only grants; their `system/` tree, git metadata, writes and enumeration remain denied. Both `memory.project_scope.nudge` and `.recall` default to `true` and can be disabled separately. ([#10044](https://github.com/code-yeongyu/oh-my-openagent/issues/10044), [#10921](https://github.com/code-yeongyu/oh-my-openagent/pull/10921))
|
|
148
|
+
|
|
149
|
+
**omo.json can say which workspace a session belongs to ([#10852](https://github.com/code-yeongyu/oh-my-openagent/pull/10852)).** `provenance.workspaces` lists workspaces by name with optional `git_owners` and `paths`; messages between sessions use it to show where the sender works (#10794).
|
|
150
|
+
|
|
151
|
+
**Messages between sessions start with the sender's workspace and project ([#10853](https://github.com/code-yeongyu/oh-my-openagent/pull/10853), [#10794](https://github.com/code-yeongyu/oh-my-openagent/issues/10794)).** The sender's environment, git owner or configured paths decide it, never the message text; older stored messages say provenance was not recorded (#10794).
|
|
152
|
+
|
|
153
|
+
**A memory git command that hangs past its timeout can now leave evidence behind.** With `OMO_GIT_TIMEOUT_DIAGNOSTICS_DIR` set, a timed-out git records the running process table (and on Windows the stuck git's thread wait states) in that directory before it is stopped; unset, nothing changes. Windows CI uses it to diagnose a rare stalled `git add`. ([#10821](https://github.com/code-yeongyu/oh-my-openagent/issues/10821), [#10914](https://github.com/code-yeongyu/oh-my-openagent/pull/10914))
|
|
154
|
+
|
|
155
|
+
**Stopping a task no longer risks matching a task with the same id from another task store ([#10903](https://github.com/code-yeongyu/oh-my-openagent/pull/10903)).** When a task's session shuts down, it finds its own task by the task id in its session context. Task ids are unique only within one task store, so that id is now trusted only when the session's state folder is exactly this store's folder for that task; otherwise the session falls back to matching by its session id. ([#10881](https://github.com/code-yeongyu/oh-my-openagent/issues/10881))
|
|
156
|
+
|
|
157
|
+
**Anonymous usage stats now count each person once.** The anonymous id is kept once per user instead of per agent directory, so runs that use their own agent directory, and the CLI next to OmO Desktop, no longer show up as separate people; runs in CI, under a test runner, or with a temporary home directory send nothing by default (`OMO_TELEMETRY_ALLOW_AUTOMATION=1` opts back in). This also holds on Windows: a temporary home folder there is recognised too, a home folder that cannot be found turns usage stats off rather than on, and a real profile under the users folder is never treated as temporary. On some machines the anonymous id changes once with this update ([#10935](https://github.com/code-yeongyu/oh-my-openagent/pull/10935)).
|
|
158
|
+
|
|
159
|
+
**Refused child tasks report startup failure without waiting for slow teardown.** Shutdown still waits for pending cleanup ([#10988](https://github.com/code-yeongyu/oh-my-openagent/pull/10988), fixes [#10402](https://github.com/code-yeongyu/oh-my-openagent/issues/10402)).
|
|
160
|
+
|
|
161
|
+
### Fixed
|
|
162
|
+
|
|
163
|
+
- Memory facts extraction no longer loops until its deadline when a model fills the optional `person` field on a project fact with an empty placeholder: the placeholder is dropped and the fact is recorded, while a project fact naming a real person is still refused. A model that resends a refused `record_fact` call unchanged now ends the run with the facts it already recorded, and a parked queue entry no longer holds back that conversation's later entries ([#10918](https://github.com/code-yeongyu/oh-my-openagent/issues/10918)). ([#10924](https://github.com/code-yeongyu/oh-my-openagent/pull/10924))
|
|
164
|
+
|
|
10
165
|
## [5.1.44] - 2026-10-23
|
|
11
166
|
|
|
12
167
|
**The engine moves to senpi 2026.10.21-3: long 1M-context conversations stop re-writing their cached prefix, Anthropic subscription context is counted once, and "No credential slots available" no longer locks every account out ([#10965](https://github.com/code-yeongyu/oh-my-openagent/pull/10965)).** When context reduction trims a long conversation on a 1M window, the reduced prefix now stays identical and keeps its cache checkpoint, so each step writes only what changed. Anthropic subscription turns with internal SDK re-requests no longer double the context percentage or start compaction early, and a configured compaction model now summarizes while your session provider is healthy, overflow recovery included. The subscription lane is the only layer that picks its accounts, so one revoked or briefly rate-limited account no longer blocks the rest until you log in again. Every turn end says what happens next, a turn-end report names the hold your queued message released, the trailing "Anything else? (optional)" card is gone from clients without native questions, and moving a session between machines runs through one implementation. Forced recovery after a context overflow no longer picks the same compaction cut again when an invisible entry leads the session; an overflow still never moves a task to its next fallback model.
|
|
@@ -18,8 +173,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
18
173
|
**Prompts stop freezing a shared host after every memory commit ([#10731](https://github.com/code-yeongyu/oh-my-openagent/pull/10731)).** Recall skips memory notes the conversation already mentions, and it rechecked the last 200 messages against every note again after each memory commit, one note at a time. On a 7,500-note memory that took 1 to 1.7 s per session per prompt, with every session on the host waiting. Earlier answers now carry over and only notes the commit added are checked, so a recheck takes about 6 ms and the first check about 0.2 s; with seven sessions and a commit every 20 s, the host's blocked time fell from 185 s to 28 s over four and a half minutes. The notes skipped are unchanged. ([#10729](https://github.com/code-yeongyu/oh-my-openagent/issues/10729))
|
|
19
174
|
|
|
20
175
|
**A relaunch's first prompt no longer recompiles your memory.** The compiled memory block and its size estimate are now saved by commit inside the memory repository's `.git` folder, never committed or synced, and read back on the next launch. On a 6,862-note memory this cut the memory step of a prompt from a median of about 178 ms to about 52 ms in the desktop app. A block built after a git read failed is used once and never saved. ([#10674](https://github.com/code-yeongyu/oh-my-openagent/pull/10674))
|
|
21
|
-
|
|
22
|
-
|
|
23
176
|
**Anonymous usage stats now count how many accounts you have linked per provider.** `session_started` reports, per provider kind, how many accounts the engine lists for it (the same list the app's account list shows: signed-in accounts and environment-variable slots), used or not, as capped counts only, never an account name, email, token or value ([#10919](https://github.com/code-yeongyu/oh-my-openagent/issues/10919)).
|
|
24
177
|
|
|
25
178
|
**The desktop app and `omo login` can share one device on a machine.** A new hidden `omo account` command (`show`, `adopt`, `token`, `export`, `forget`) lets the desktop use the same OS-keychain sign-in the CLI writes, so a machine counts as one device for memory sync and only one process ever refreshes its tokens. Secrets pass only on stdin and stdout, a refused renewal (exit 5) names the service's refusal code so the desktop can say why the sign-in ended, and a locked OS keychain is reported as its own outcome (exit 7) instead of a generic failure. On Linux, where a locked Secret Service collection reads as empty, the default collection's lock state is read over D-Bus before any secret call (libsecret would otherwise ask the keyring to show an unlock prompt), a locked collection is reported (exit 7) instead of read, and when that check cannot answer (no `gdbus` on a minimal install, no Secret Service on the bus, a slow bus) the CLI prints one notice and uses libsecret directly as before, so login never depends on `gdbus` ([#10658](https://github.com/code-yeongyu/oh-my-openagent/pull/10658)).
|
|
@@ -54,16 +207,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
54
207
|
|
|
55
208
|
**OmO sessions prewarm the prompt cache again, and anything that blocks it is named.** Since v5.1.17 every OmO session skipped senpi's prompt-cache prewarm, because the gateway, onboarding and task-host `before_agent_start` handlers were not registered as preview-safe; they now are, and a preview changes no state (no onboarding claim, no gateway store, no host launch). An unsafe `before_agent_start` registration warns immediately, including one added after startup; a recorded skipped prewarm is reported once per session at the next lifecycle event, with its reason and the blocking components and source locations. Gateway preview lookup errors now pass the prompt through and warn once instead of throwing. ([#10356](https://github.com/code-yeongyu/oh-my-openagent/issues/10356), [#10372](https://github.com/code-yeongyu/oh-my-openagent/pull/10372))
|
|
56
209
|
|
|
57
|
-
|
|
58
210
|
### Fixed
|
|
59
211
|
|
|
60
212
|
- While a delegated task hands off to its next fallback model, its row, `task_output` and completion notice no longer claim it ran on a model nothing has run on yet: the record keeps the last model a child was seen running until the new one reports, and the parent-model note reads "switching to" during the handoff and "was never confirmed running on" if the task ends before a child is seen on it ([#10797](https://github.com/code-yeongyu/oh-my-openagent/issues/10797)). ([#10807](https://github.com/code-yeongyu/oh-my-openagent/pull/10807))
|
|
61
213
|
|
|
62
214
|
## [5.1.42] - 2026-10-22
|
|
63
215
|
|
|
64
|
-
|
|
65
216
|
**Memory sync checks each new keyring entry without re-reading the chain before it.** The client now uses the shared receiver's incremental entry point (omo-service #600, re-vendored), and the receiver moves along with the verified chain state. Appending one entry to a 1,000-entry chain used to hand the receiver all 1,006 entries again; it now hands it one, at any chain length. What the receiver refuses is unchanged, including a replayed or lower-seq entry. ([#10694](https://github.com/code-yeongyu/oh-my-openagent/pull/10694), fixes [#10655](https://github.com/code-yeongyu/oh-my-openagent/issues/10655))
|
|
66
|
-
|
|
67
217
|
**The engine moves to senpi 2026.10.21-2: idle task hosts exit again, reopened sessions stop spending quota, and sessions share their prompt cache ([#10893](https://github.com/code-yeongyu/oh-my-openagent/pull/10893)).** A monitor watch no longer keeps every task shard host on the machine running, so idle hosts exit as designed. Reopening an unattended session no longer spends quota on goals or background events until you send input or choose Resume. Responses prompt-cache keys follow the model, instructions and tool schemas instead of the session, so new sessions and task children with the same prefix reuse one cache. A goal waiting on a subscription limit retries at the provider's reset time instead of hammering the spent account. On Windows, input sent while a turn is running no longer gets stuck or races through the whole fallback chain. Switching a long conversation to a smaller-context model while the current one is at its limit now applies at the next message, saving an API key over an OAuth login (or the reverse) replaces the credential instead of storing one that is never used, a busy command monitor is no longer muted after 200 lines a day, and every assistant reply records its timing (time to first token and generation rate).
|
|
68
218
|
|
|
69
219
|
**OmO records when an answer says "tests pass" or "done" with nothing in the session behind it.** When a turn's final answer says tests pass, a build or typecheck is clean, a feature already exists, or the work is done, OmO checks the session for a check that passed after the last code edit, and records in the session whether the claim is backed. It starts in shadow mode: nothing is shown to you or the agent and no turn is blocked or continued, because on a week of our own sessions most flags were still false positives. Setting the `omo-senpi-claim-receipt-nudge` flag turns on a short note under a flagged answer (once per message you send) and passes the flagged claims to a gate reviewer spawned afterwards. ([#10785](https://github.com/code-yeongyu/oh-my-openagent/pull/10785))
|
|
@@ -82,7 +232,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
82
232
|
|
|
83
233
|
**Git worktrees of one repository can share the main checkout's memory.** Set `memory.auto_identity: "git"` (for example in `~/.omo/omo.jsonc`, so worktrees created outside the main checkout see it) and every linked worktree, including the ones Orca creates per agent, reads and writes the main checkout's memory instead of starting empty. The main checkout keeps its existing memory, sessions that are already bound keep theirs, and the default `"path"` leaves behavior unchanged. ([#9659](https://github.com/code-yeongyu/oh-my-openagent/issues/9659), [#10047](https://github.com/code-yeongyu/oh-my-openagent/pull/10047))
|
|
84
234
|
|
|
85
|
-
|
|
86
235
|
## [5.1.41] - 2026-10-22
|
|
87
236
|
|
|
88
237
|
**The engine moves to senpi 2026.10.21-1: engine hosts start faster and idle far more quietly, Claude subscriptions use the native 1M context again, and desktop threads are lighter ([#10740](https://github.com/code-yeongyu/oh-my-openagent/pull/10740)).** A new engine host starts about 0.1 s faster, and an idle engine host wakes 405 times per 90 s instead of 1,106. Current Claude subscription models (Fable 5, Sonnet 5, Haiku 5.5, Opus 4.7 and later) get their native 1M window back without asking for extra usage, and the Claude subscription lane runs Claude Code 2.1.306. Each desktop thread's engine supervisor starts from its own small program instead of the whole CLI, and messages and tool cards no longer keep duplicate copies (about 640 bytes per message and 2 KB per tool card). Models that models.dev marks deprecated stay available, labeled deprecated, instead of disappearing (GitHub Copilot Claude Opus 4.7, GPT-5.5 and GPT-5.4, Baseten Kimi K2 and others), and the model catalog is refreshed; some models get updated thinking levels (for example Nemotron 3 Ultra: off is now none). A message that was already delivered no longer stays listed under Steering, and with several accounts for one provider a usage limit stops failing turns once its reset has passed; a quota failure now says when it resets.
|
|
@@ -105,7 +254,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
105
254
|
|
|
106
255
|
**`ulw-research` shows its report inline and checks a visual reproduction against the source ([#10812](https://github.com/code-yeongyu/oh-my-openagent/pull/10812)).** When the host can show an HTML page in the thread (`show_html_page`, `html_render`), the run ends with the report there, before the file paths, on every tier and even when the research was shortened; a built app or 3D scene shown inline no longer stands in for it. When the ask is to recreate something visual, such as a clip rebuilt in 3D, the run puts real frames of the source next to its own output at the same moments, writes down what differs, fixes the biggest gaps, and shows the side-by-side pairs in the report.
|
|
107
256
|
|
|
108
|
-
|
|
109
257
|
### Changed
|
|
110
258
|
|
|
111
259
|
- Combine macOS root-test shards and Codex/Senpi compatibility into one runner job, preserving their existing check names through Ubuntu status jobs. Full-matrix CI now requests two macOS runners per head instead of five, including unchanged native binary parity; fast paths skip the combined allocation ([#10804](https://github.com/code-yeongyu/oh-my-openagent/pull/10804)).
|
|
@@ -75,7 +75,7 @@ Read `$SYSTEM_TOKENS_PATH` before consolidating. When its `totalTokens` is great
|
|
|
75
75
|
|
|
76
76
|
**Contradiction handling**: when files disagree, never silently pick a winner. Keep both versions, mark the disagreement on each entry (a contradiction comment naming the other file and the evidence dates), and surface the conflict in your final report so a human decides. Identical duplicates may still be deduped; genuinely conflicting content may not.
|
|
77
77
|
|
|
78
|
-
Throughout, edit surgically. Persona and behavioral files are load-bearing: append, modify specific entries, adjust wording. Never rewrite them wholesale or silently overwrite established identity. Keep
|
|
78
|
+
Throughout, edit surgically. Persona and behavioral files are load-bearing: append, modify specific entries, adjust wording. Never rewrite them wholesale or silently overwrite established identity. Keep `[[path]]` cross-references accurate as content moves. A `description` is one line of at most 1024 characters saying what the note is and when to read it. Keep history, changelogs, and topic lists in the body, never append them to the description.
|
|
79
79
|
Never write or copy a `provenance:` frontmatter line; only the memory tool writes one, and the commit hook refuses a hand-written one. When you move or merge a note, leave that line out of the new file. When you edit a note whose `provenance:` line starts with `user said`, delete that line (it is a legacy header that no longer means anything) and add the trailer `Omo-Provenance-Stripped: <path>` to your commit message, one per note.
|
|
80
80
|
|
|
81
81
|
|
|
@@ -1,2 +1,2 @@
|
|
|
1
|
-
// omo:
|
|
1
|
+
// omo:EDhT2m1VSMsgQ5uh8_IIvcRgZJA6OOMbSLCmLpqmqH8:5LLt7853QZ0R275mM__hYa8UeDUj-ySee_PqmOY8R7U
|
|
2
2
|
function e(e){return e.replaceAll("&","&").replaceAll("<","<").replaceAll(">",">")}function r(r,s){let i=[`<operating-rules version="${e(r).replaceAll('"',""")}">`];for(let r of s)i.push(`- ${e(r)}`);return i.push("</operating-rules>"),i.join("\n")}function s(e,r){if(null===e||"object"!=typeof e||Array.isArray(e))throw Error(`${r} must be an object`);return e}function i(e,r){let s=e[r];if("string"!=typeof s||0===s.length)throw Error(`${r} must be a non-empty string`);return s}function o(e){let r=s(e,"a rules target"),o=i(r,"session_durable_id"),n=r.binding_id;if(null!=n&&"string"!=typeof n)throw Error("binding_id must be a string");let t=r.behavioral;if(null===t)return{session_durable_id:o,binding_id:n??void 0,behavioral:null};if(!Array.isArray(t)||!t.every(e=>"string"==typeof e))throw Error("behavioral must be an array of rule texts, or null to clear the session's rules");return{session_durable_id:o,binding_id:n??void 0,behavioral:t}}function n(e,r){let o=i(s(r,"blockForSession args"),"session_durable_id"),n=e.one(["version","block"],"SELECT version, block FROM gateway_rules_blocks WHERE session_durable_id = ?",[o]);return void 0===n?null:{version:String(n.version),block:String(n.block)}}function t(e,r){let o=i(s(r,"sessionsWithRules args"),"scope");return{sessions:e.all(["session_durable_id","version"],"SELECT session_durable_id, version FROM gateway_rules_blocks WHERE scope = ?",[o],"session_durable_id").map(e=>({session_durable_id:String(e.session_durable_id),version:String(e.version)}))}}async function l(n,t){let l=s(t,"rulesCommitted args"),a=i(l,"scope"),d=i(l,"version"),u=l.now??Date.now();if("number"!=typeof u||!Number.isFinite(u))throw Error("now must be a finite timestamp");if(!Array.isArray(l.targets))throw Error("targets must be an array");let _=[];for(let s of l.targets){let i=o(s);if(null===i.behavioral){let e=n.exec("DELETE FROM gateway_rules_blocks WHERE session_durable_id = ?",[i.session_durable_id]);_.push({session_durable_id:i.session_durable_id,outcome:e>0?"removed":"absent"});continue}let t=r(d,i.behavioral),l=n.exec("INSERT INTO gateway_rules_blocks (session_durable_id, scope, version, block, updated_at) VALUES (?, ?, ?, ?, ?)\n ON CONFLICT(session_durable_id) DO UPDATE SET scope = excluded.scope, version = excluded.version, block = excluded.block, updated_at = excluded.updated_at\n WHERE gateway_rules_blocks.version != excluded.version",[i.session_durable_id,a,d,t,u]);if(void 0===i.binding_id){_.push({session_durable_id:i.session_durable_id,outcome:l>0?"updated":"unchanged"});continue}let c=await n.enqueue({binding_id:i.binding_id,event_id:`rules_changed:${a}:${d}:${i.session_durable_id}`,text:[`<operating-rules-changed scope="${e(a).replaceAll('"',""")}" version="${e(d).replaceAll('"',""")}">`,"The operating rules for this gateway scope changed; this session's system prompt carries the new block from its next turn.","</operating-rules-changed>"].join("\n"),mode:"follow_up"});_.push({session_durable_id:i.session_durable_id,outcome:l>0?"updated":"unchanged",delivery:"ok"===c.kind?{kind:"ok",delivery_id:c.delivery_id,deduplicated:!0===c.deduplicated}:{kind:"error",error:c.error}})}return{scope:a,version:d,outcomes:_}}export{n as blockForSession,l as rulesCommitted,t as sessionsWithRules};
|