omo-ai 5.1.63 → 5.1.65
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/darwin/omo-keychain +0 -0
- package/bin/lib/doctor-memory-runtime.js +167 -0
- package/bin/lib/doctor-reflection-lock.js +102 -6
- package/bin/lib/doctor.js +4 -2
- package/bin/lib/engine-boot.js +27 -0
- package/bin/lib/launcher.js +44 -3
- package/bin/lib/memory-sync-boot.js +19 -0
- package/bin/lib/provider-map.json +1 -1
- package/bin/lib/reflection-lock.generated.js +39 -1
- package/bin/lib/service-auth.generated.js +37 -18
- package/bin/lib/thread.js +28 -2
- package/package.json +2 -2
- package/plugin/CHANGELOG.md +86 -0
- package/plugin/extensions/dream-persona.md +1 -1
- package/plugin/extensions/gateway-store-worker.mjs +2 -2
- package/plugin/extensions/memory-run-supervisor.mjs +3 -3
- package/plugin/extensions/omo-claim-receipt.js +1 -1
- package/plugin/extensions/omo-computer-use.js +1 -1
- package/plugin/extensions/omo-desktop-toolkit.js +1 -1
- package/plugin/extensions/omo-device-account.js +66 -2
- package/plugin/extensions/omo-device-service.js +5 -5
- package/plugin/extensions/omo-feedback.js +2 -2
- package/plugin/extensions/omo-init-deep-advisor.js +1 -1
- package/plugin/extensions/omo-member.js +2 -2
- package/plugin/extensions/omo-memory-commands.js +3 -3
- 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 +1 -1
- package/plugin/extensions/omo-side-panel.js +2 -2
- package/plugin/extensions/omo-task.js +2 -2
- package/plugin/extensions/omo-telemetry-otlp.js +2 -2
- package/plugin/extensions/omo-telemetry-posthog.js +1 -1
- package/plugin/extensions/omo-telemetry-recovery.js +1 -1
- package/plugin/extensions/omo-thread-client.js +5 -5
- package/plugin/extensions/omo.js +3 -3
- package/plugin/package.json +2 -1
- package/plugin/runtime/agent-toolkit-sdk/sdk.js +2 -2
- package/plugin/runtime/ast-grep-mcp/manifest.json +1 -1
- package/plugin/runtime/bunshin-ts/sdk.js +2 -2
- package/plugin/runtime/category-coverage/index.js +2 -2
- package/plugin/runtime/rollback-migrate.js +1 -1
- package/plugin/runtime/task-config/index.js +3 -3
- package/plugin/runtime/thread-sdk/sdk.js +5 -5
- package/plugin/skills/browser/SKILL.md +3 -1
- package/plugin/skills/browser/runtime/omowright/index.js +232 -341
- package/plugin/skills/browser/runtime/omowright/manifest.json +5 -5
- package/plugin/skills/browser/runtime/omowright/page-bundle.js +1586 -1394
- package/plugin/skills/browser/scripts/browser-doctor-report.mjs +68 -0
- package/plugin/skills/browser/scripts/browser-doctor.mjs +8 -43
- package/plugin/skills/browser/scripts/omowright.mjs +7 -1
package/bin/lib/thread.js
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
1
|
+
import { readFileSync } from "node:fs"
|
|
1
2
|
import { userInfo } from "node:os"
|
|
2
3
|
import { join } from "node:path"
|
|
3
4
|
import { pathToFileURL } from "node:url"
|
|
@@ -186,7 +187,7 @@ export function threadExitCode(result) {
|
|
|
186
187
|
* The answer is `{ sdk }` or `{ error, code }`: `unsupported` for the runtime, `internal_error` for a
|
|
187
188
|
* plugin SDK that cannot be imported (a broken or partial install).
|
|
188
189
|
*/
|
|
189
|
-
export async function loadThreadSdk({ pluginRoot, agentDir, env, cwd, engine, loadSqlite, importSdk, identity }) {
|
|
190
|
+
export async function loadThreadSdk({ pluginRoot, agentDir, env, cwd, engine, loadSqlite, importSdk, importThreadClient = () => import(pathToFileURL(join(pluginRoot, "extensions", "omo-thread-client.js")).href), identity }) {
|
|
190
191
|
try {
|
|
191
192
|
await (loadSqlite ?? (() => import("node:sqlite")))()
|
|
192
193
|
} catch {
|
|
@@ -200,7 +201,32 @@ export async function loadThreadSdk({ pluginRoot, agentDir, env, cwd, engine, lo
|
|
|
200
201
|
}
|
|
201
202
|
const who = identity ?? { uid: process.getuid?.() ?? 0, user: userInfo().username }
|
|
202
203
|
const engineStatusAll = async () => engine.run(["host", "status", "--json", "--all", "--include-workers"], { env: { ...env, OMO_AGENT_DIR: agentDir } }).stdout
|
|
203
|
-
|
|
204
|
+
const remote = await cliRemoteSurface({ pluginRoot, env, cwd, importThreadClient })
|
|
205
|
+
return { sdk: module.createThreadSdk({ agentDir, cwd, uid: who.uid, user: who.user, env, engineStatusAll, ...(remote === undefined ? {} : { remote }) }) }
|
|
206
|
+
}
|
|
207
|
+
|
|
208
|
+
/**
|
|
209
|
+
* The account's other devices and its alias registry for `omo thread` (#12833), from the plugin's own thread client and
|
|
210
|
+
* device-account sidecars: `read` then resolves an alias or another device's thread as the agent tools do. A plugin
|
|
211
|
+
* whose thread client does not offer it keeps the local-only behavior. The SDK closes it on dispose.
|
|
212
|
+
*/
|
|
213
|
+
async function cliRemoteSurface({ pluginRoot, env, cwd, importThreadClient }) {
|
|
214
|
+
let client
|
|
215
|
+
try {
|
|
216
|
+
client = await importThreadClient()
|
|
217
|
+
} catch {
|
|
218
|
+
return undefined
|
|
219
|
+
}
|
|
220
|
+
if (typeof client.createCliRemoteSurface !== "function") return undefined
|
|
221
|
+
let version = "0.0.0"
|
|
222
|
+
try {
|
|
223
|
+
const declared = JSON.parse(readFileSync(join(pluginRoot, "package.json"), "utf8")).version
|
|
224
|
+
if (typeof declared === "string") version = declared
|
|
225
|
+
} catch {
|
|
226
|
+
// A plugin manifest that cannot be read only loses the version the device service is told; reads still work.
|
|
227
|
+
}
|
|
228
|
+
const loadAccess = async () => (await import(pathToFileURL(join(pluginRoot, "extensions", "omo-device-account.js")).href)).cliServiceAccess({ env, cwd })
|
|
229
|
+
return client.createCliRemoteSurface({ version, loadAccess })
|
|
204
230
|
}
|
|
205
231
|
|
|
206
232
|
export async function runThreadCommand(args, options) {
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "omo-ai",
|
|
3
|
-
"version": "5.1.
|
|
3
|
+
"version": "5.1.65",
|
|
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": "2026.11.
|
|
20
|
+
"@code-yeongyu/senpi": "2026.11.6-3"
|
|
21
21
|
},
|
|
22
22
|
"repository": {
|
|
23
23
|
"type": "git",
|
package/plugin/CHANGELOG.md
CHANGED
|
@@ -7,6 +7,92 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
7
7
|
|
|
8
8
|
## [Unreleased]
|
|
9
9
|
|
|
10
|
+
## [5.1.65] - 2026-11-08
|
|
11
|
+
|
|
12
|
+
**A message an agent sends to a session on another device now says who sent it.** It used to arrive there as plain text, indistinguishable from that session's own user. It now carries the same sender line and authority notice as a message between sessions on one device (the sending session, its workspace and project, and the device it came from), and a leading `/` in it no longer runs a command or expands a prompt template on the receiving side. `omo remote send` from a terminal is unchanged. ([#12858](https://github.com/code-yeongyu/oh-my-openagent/pull/12858), part of [#12852](https://github.com/code-yeongyu/oh-my-openagent/issues/12852))
|
|
13
|
+
|
|
14
|
+
**Memory retention no longer evaluates its completion-record guard twice per record.** The completion sweep now reads the guard once and keeps or removes the record on that single answer, and the retention tests check their own preconditions instead of asserting them away. No change to what retention keeps or removes. ([#12783](https://github.com/code-yeongyu/oh-my-openagent/pull/12783), follow-up to [#12532](https://github.com/code-yeongyu/oh-my-openagent/pull/12532))
|
|
15
|
+
|
|
16
|
+
**The standalone `omo` binary's `omo doctor` reports memory runtime size again, like the npm install.** Each memory identity's runtime size against its cap and the total across identities are printed by both distributions, so binary/npm doctor parity holds ([#12785](https://github.com/code-yeongyu/oh-my-openagent/issues/12785)). ([#12783](https://github.com/code-yeongyu/oh-my-openagent/pull/12783))
|
|
17
|
+
|
|
18
|
+
**Old binary runtimes are pruned safely when two `omo` launches start at once.** A launch no longer deletes a runtime another launch is still pruning, and a runtime that launch already removed is not reported as a failure. `omo --version`, `--help` and `omo update` never prune. ([#12770](https://github.com/code-yeongyu/oh-my-openagent/pull/12770), follow-up to [#12627](https://github.com/code-yeongyu/oh-my-openagent/pull/12627))
|
|
19
|
+
|
|
20
|
+
**A task child whose engine cannot hold its designated model chain is no longer started.** Before, when a process child's engine could not take its chain (an older engine, a refused `set_retry_fallback`, or no answer in time) or a task host did not advertise `retry_fallback_profile`, the child was still prompted and ran on the engine's own fallback settings, which can reach models outside its chain. The start now fails before the first prompt and names every designated model and the reason, and such a host is refused as incompatible, naming the same models. ([#12759](https://github.com/code-yeongyu/oh-my-openagent/issues/12759)) ([#12761](https://github.com/code-yeongyu/oh-my-openagent/pull/12761))
|
|
21
|
+
|
|
22
|
+
**A command on a closed session on another device now reports the same error as on this device.** It answered `session_not_found` remotely and `unknown_session` locally; both now answer `unknown_session`. ([#12755](https://github.com/code-yeongyu/oh-my-openagent/issues/12755)) ([#12756](https://github.com/code-yeongyu/oh-my-openagent/pull/12756))
|
|
23
|
+
|
|
24
|
+
**omo hands its device keys to the device service from memory, with no credential store read for relay.** A fresh sign-in (`omo login`, `omo account adopt`) deletes the signed-out marker and then hands its in-memory keys and account to a running device service. The memory sync session re-adopts from the credential it already holds when the service reports no adopted device. Every sign-out path (logout, forget, a session the service ended) tells the service to forget the device. The start-up bootstrap no longer reads the stored credential. `device.adopt` carries `account` only to a device service at 0.1.3 or newer, and the pinned device service is raised to 0.1.3 (design D PR 4). ([#12735](https://github.com/code-yeongyu/oh-my-openagent/pull/12735))
|
|
25
|
+
|
|
26
|
+
**`omo remote` can open, send, switch model, interrupt and close sessions on your other devices.** `omo remote open <device> [--project P] [--model M]` starts a session there and prints `<device>/<session>`; `omo remote send <device>/<session> <text>` prints the answer as it streams in and returns when the turn settles (`--steer` steers the running turn, `--follow-up` queues behind it, `--no-wait` returns once the session took the message); `model`, `interrupt` and `close` do what they say. Each verb is the same remote op the agent's thread tools use, and `--json` prints its typed result. A steer to a session on another device now reaches it when that device's engine reports its running turn, instead of always being refused. ([#12709](https://github.com/code-yeongyu/oh-my-openagent/pull/12709), part of [#12320](https://github.com/code-yeongyu/oh-my-openagent/issues/12320))
|
|
27
|
+
|
|
28
|
+
Creating or forking a thread on another device no longer waits for a separate `session.locate` round trip: the create asks the relay for the new session's id on the `open_session` reply and uses it. A relay daemon without that answer (an older host, or an older service on this device) is handled as before, with `session.locate`. ([#12686](https://github.com/code-yeongyu/oh-my-openagent/pull/12686))
|
|
29
|
+
|
|
30
|
+
**`omo update` on the standalone binary now replaces the installed `omo`.** It used to print a replace command for the copy under `~/.omo/binary-runtime/<version>/`, which the next launch put back, so the update never took. A launch also removes old runtime versions there, keeping the previous one for a rollback and any version still running. ([#12627](https://github.com/code-yeongyu/oh-my-openagent/pull/12627), fixes [#12626](https://github.com/code-yeongyu/oh-my-openagent/issues/12626))
|
|
31
|
+
|
|
32
|
+
**Each memory identity's runtime directory is capped at 2 GiB, and `omo doctor` and `/doctor` show how much it holds.** After retention, the 12-hour memory maintenance pass evicts oldest-first until the identity fits `memory.runtime.max_bytes` (default 2 GiB, `0` for unbounded, per-agent overridable under `memory.agents.<id>.runtime`). It removes folded reflection runs first, then the transcripts of fully reflected sessions, idle sidecars, and finally old failed runs beyond the newest ten. Journals that were never or only partly reflected are kept by default; `memory.runtime.discard_unreflected_journals: true` lets the cap evict them too, and only while reflection is succeeding on the identity. Live sessions, in-flight runs and the run-id high water are never touched. `omo doctor` prints each identity's runtime size and the total, and warns when an identity reaches 80 % of its cap. `/doctor` adds a `runtime-size` check with the largest stores and what retention would still free. ([#12538](https://github.com/code-yeongyu/oh-my-openagent/pull/12538))
|
|
33
|
+
|
|
34
|
+
**Memory no longer grows without bound on disk.** The background memory maintenance pass now removes finished reflection runs, fully reflected transcripts, recall sidecars and legacy stores past configurable windows (`memory.runtime.retention_days` 7, `completions_days` 30), in time-boxed slices that resume where they stopped. Transcript input reflection never folded into memory is kept unless you turn on `discard_unreflected_journals` (then removed after `abandoned_journal_days`, 14, only while reflection works for that agent). Nothing a live session owns, no locked or unfinished work and no run id reflection still needs is touched. ([#12532](https://github.com/code-yeongyu/oh-my-openagent/pull/12532))
|
|
35
|
+
|
|
36
|
+
A reflection lock left behind by a run whose processes died (launcher killed before the run started, supervisor and child gone with no outcome, an unreadable run ledger, or a torn `active.lock`) is now released at session start, on the next turn, and by the new `omo doctor --fix` (`--dry-run` to preview, `/doctor --fix` in a session), instead of keeping every later reflection queued; `omo doctor --fix` also reclaims runtime locks whose owner is gone. ([#12517](https://github.com/code-yeongyu/oh-my-openagent/pull/12517))
|
|
37
|
+
|
|
38
|
+
### Fixed
|
|
39
|
+
|
|
40
|
+
- Senpi: omo now names any senpi feature it relies on that the running senpi lacks in one startup warning, and opening a monitor on another device's session on a senpi without remote wake fails with `remote_wake_unavailable` instead of starting a watch that never wakes. A contract test checks every such feature against the pinned senpi. ([#12862](https://github.com/code-yeongyu/oh-my-openagent/pull/12862), fixes [#12861](https://github.com/code-yeongyu/oh-my-openagent/issues/12861))
|
|
41
|
+
- The browser skill now depends on the published `omowright` package (`^0.3.0`, the first release that ships a built `dist/` that loads under Node as well as Bun) instead of a git commit. Its runtime stager and loader accept omowright's TypeScript sources, the older JavaScript sources, or a compiled `dist/`, in that order, so a future layout change cannot silently break the staged browser runtime (#12842, #12851, #12847). ([#12847](https://github.com/code-yeongyu/oh-my-openagent/pull/12847))
|
|
42
|
+
- Memory facts extraction records facts again with senpi 2026.11.6-3: the facts child (also used by onboarding's understand step) keeps the approval mode its own settings give it and may run only its own record tools without asking, instead of being refused as a child whose parent is gone. ([#12839](https://github.com/code-yeongyu/oh-my-openagent/pull/12839))
|
|
43
|
+
- Concurrent memory account resolves on a desktop host share one open, so the app sees one `descriptor` request instead of one per caller; a failed open is retried by the next resolve (carried from #12717, which this PR supersedes). ([#12839](https://github.com/code-yeongyu/oh-my-openagent/pull/12839))
|
|
44
|
+
- Thread tools: an agent can now reach a session on another device of the same account by its name. A name that matches the title of one live session on another device sends there, and a name shared across devices (or with a thread on this device) is refused as ambiguous with every candidate and its device listed, never picked silently. Claiming or renaming a thread alias on a signed-in CLI device now reaches the account's alias registry instead of always answering that the registry cannot be reached; an alias claimed on one device now reaches that thread from every other device instead of reporting its device offline, and a registry outage never changes how a plain name resolves. `omo thread read` resolves aliases and other devices' threads the way the agent's `thread_read` does. ([#12837](https://github.com/code-yeongyu/oh-my-openagent/pull/12837), fixes [#12833](https://github.com/code-yeongyu/oh-my-openagent/issues/12833))
|
|
45
|
+
- Release: the service-auth release step now refuses a bundle that carries more than one release key set, so it always checks the set the bundle ships. ([#12812](https://github.com/code-yeongyu/oh-my-openagent/pull/12812), fixes [#12617](https://github.com/code-yeongyu/oh-my-openagent/issues/12617))
|
|
46
|
+
- Runtime fallback: when a session times out and no fallback model can be prepared, the user is now warned even if nothing was awaiting a fallback, the warning names the real reason ("no available fallback models" while attempts remain, instead of "max fallback attempts reached"), and a compaction that outlives its watchdog budget returns control as a real cancellation without staying marked as recovering. ([#12790](https://github.com/code-yeongyu/oh-my-openagent/pull/12790), fixes [#12762](https://github.com/code-yeongyu/oh-my-openagent/issues/12762))
|
|
47
|
+
- Standalone binary on Windows: running the provisioned runtime copy directly (`~/.omo/binary-runtime/<version>/omo.exe`) no longer hands off to itself and reports itself as the installed launcher when the home path is spelled with an 8.3 short name or different letter case. ([#12772](https://github.com/code-yeongyu/oh-my-openagent/pull/12772), fixes [#12769](https://github.com/code-yeongyu/oh-my-openagent/issues/12769))
|
|
48
|
+
- While service accounts are closed, the compiled `omo` binary recognises the device-code `omo login` with the same option parse as the command itself, so `omo login -- --device` and similar argv get the immediate refusal instead of being handed on to the command. ([#12723](https://github.com/code-yeongyu/oh-my-openagent/pull/12723), fixes [#12646](https://github.com/code-yeongyu/oh-my-openagent/issues/12646))
|
|
49
|
+
- The browser skill's doctor no longer calls a browser ready when its only BrowserSkill extension is 0.4.0. That release left a "BrowserSkill" tab group in the user's browser after every task. The doctor now reports `outdated-extension` and tells the user to update the extension and delete the saved groups. ([#12663](https://github.com/code-yeongyu/oh-my-openagent/pull/12663))
|
|
50
|
+
- `omo` no longer warns "Could not download the OmO device service. Check the connection and try again." on every interactive start while the device service release it pins has not been published. A 404 for that release is now its own `not_published` state: the start stays quiet, and `omo remote bootstrap` says remote devices are not available in this build yet. A connection failure, a server error or a half-published release still warns. ([#12659](https://github.com/code-yeongyu/oh-my-openagent/pull/12659))
|
|
51
|
+
- When a dream or reflection run deletes, archives or moves a memory note, `[[path]]` and relative markdown links to it in other notes are now updated in the same run, after the completion check has decided which of the run's changes land. A moved note's links follow it, and the moved note's own relative links keep pointing at their targets. A note the run archived, meaning its `ARCHIVE.md` entry names the file (its path, or its file name when no other note shares it), is linked as `[[ARCHIVE.md|<old path> (archived)]]`. Links to any other removed note, such as one the user asked to forget, become plain text. A note the completion check refuses, or one whose rewrite it would refuse (for example a note that already holds secret-like text), or one that is not valid UTF-8, keeps its links as they were, so `/doctor` can still report those. A targeted `/dream --to <doc>` that removes its own target is no longer refused because of these link updates. Facts extraction also no longer writes into a month ledger path that was archived or deleted: that month's new facts go to `notes/facts/<month>-2.md` (then `-3`, ...), so an old link can never start pointing at unrelated new content. ([#12642](https://github.com/code-yeongyu/oh-my-openagent/issues/12642)) ([#12645](https://github.com/code-yeongyu/oh-my-openagent/pull/12645))
|
|
52
|
+
- Runtime fallback keeps compaction retries separate from the primary model chain and preserves substantive retry prompts. Compaction watchdogs defer within a bounded budget, exhausted native retry signals do not trigger an abort, and stuck-session watchdogs still return control when no replacement can be dispatched. Thanks @mradwankhalil. ([#11907](https://github.com/code-yeongyu/oh-my-openagent/pull/11907), fixes [#9786](https://github.com/code-yeongyu/oh-my-openagent/issues/9786))
|
|
53
|
+
- Background tasks let runtime fallback decide who recovers a session error before starting a replacement or finalizing the task. Exhausted and refused retries release recovery ownership, model-less retry errors can finish the task, and a slow decision never starts a competing replacement. ([#9683](https://github.com/code-yeongyu/oh-my-openagent/pull/9683), [#11907](https://github.com/code-yeongyu/oh-my-openagent/pull/11907))
|
|
54
|
+
- Runtime fallback keeps its place in the fallback chain when its own abort echoes back as a stop or idle event, and no longer mistakes its own retry for a manual model change. A failing fallback now moves on to the next model instead of going back to the model that just failed. ([#11901](https://github.com/code-yeongyu/oh-my-openagent/pull/11901), fixes [#6751](https://github.com/code-yeongyu/oh-my-openagent/issues/6751))
|
|
55
|
+
- Runtime fallback: a session timeout whose fallback dispatch is rejected now uses up the attempt, stops at `max_fallback_attempts`, and ends with an abort the caller sees instead of re-arming the timer forever. ([#11898](https://github.com/code-yeongyu/oh-my-openagent/pull/11898), fixes [#6637](https://github.com/code-yeongyu/oh-my-openagent/issues/6637))
|
|
56
|
+
- First-prompt refusal tests now await launch settlement when the bounded start report arrives first, preserving both untyped and non-credential refusal diagnostics without depending on subprocess startup speed (#11732). ([#11852](https://github.com/code-yeongyu/oh-my-openagent/pull/11852))
|
|
57
|
+
|
|
58
|
+
### Changed
|
|
59
|
+
|
|
60
|
+
- Task children ask through the parent session instead of being refused: a child's permission prompt (and its other select, confirm, input and editor dialogs) is raised on the parent's UI, so Allow runs the child's call and Deny does not. A child the parent turn is waiting on asks on the connection that started that turn; a background child asks every attached client that may answer. Secret inputs and permanent grant selections are never forwarded. Children in their own process or on the task host open in the parent's approval mode and follow every later change of it, one change at a time and always ending on the latest; a child that could not follow is reported to the parent session. ([#12839](https://github.com/code-yeongyu/oh-my-openagent/pull/12839))
|
|
61
|
+
- `omo remote open` and the thread tools' create and fork on another device name the approval mode the new session runs in: the creating session's own mode, or for `omo remote open` the new `--permission-preset` flag, else this device's own setting; a setting that cannot be read is a usage error rather than the loosest mode. ([#12839](https://github.com/code-yeongyu/oh-my-openagent/pull/12839))
|
|
62
|
+
- The compiled engine no longer has the unused `installEngineCaTrust` startup routine. That routine set `NODE_USE_SYSTEM_CA=1` but had no caller, and engine launches set up certificate trust in the launcher trust stage. A new test checks that an engine started without `NODE_USE_SYSTEM_CA` in its environment does not get it from the launcher. Validation of `OMO_EXTRA_CA_CERTIFICATES` is unchanged. ([#12628](https://github.com/code-yeongyu/oh-my-openagent/pull/12628), part of #10762)
|
|
63
|
+
|
|
64
|
+
## [5.1.64] - 2026-11-07
|
|
65
|
+
|
|
66
|
+
**Desktop memory sync now starts as soon as the engine boots, without waiting for a session.** An engine started by `omo host ensure` loads extensions only when a session opens, while the desktop opens none until the engine has sent the memory-sync bootstrap handshake, so the two waited on each other. The compiled `omo` and the npm launcher now start that one claim before the engine loads whenever an RPC engine's launch declares a `memory-sync` channel; the first session reuses it. A claim that fails still reports memory sync as unavailable and never falls back to CLI credentials. ([#12702](https://github.com/code-yeongyu/oh-my-openagent/issues/12702)) ([#12704](https://github.com/code-yeongyu/oh-my-openagent/pull/12704))
|
|
67
|
+
|
|
68
|
+
Internal (tests only): macOS trust acquisition reads `/etc/ssl/cert.pem` through the already-injectable bundle reader, and a test now proves that a malformed default file yields the structured `openssl-default-file-malformed` drop entry (file, cause, block index, narrowing). ([#12633](https://github.com/code-yeongyu/oh-my-openagent/pull/12633))
|
|
69
|
+
|
|
70
|
+
Internal (tests only): the OpenSSL default-file importer tests now pin that any high byte on a BEGIN line, not only a BOM, drops the whole file. ([#12632](https://github.com/code-yeongyu/oh-my-openagent/pull/12632))
|
|
71
|
+
|
|
72
|
+
Internal (release tooling): the omo-native trust mutation harness counts a mutant as killed only when a named test fails by an assertion; a mutant that only throws a runtime error or crashes on load is now reported as `crashed` and fails the run. ([#12631](https://github.com/code-yeongyu/oh-my-openagent/pull/12631))
|
|
73
|
+
|
|
74
|
+
### Fixed
|
|
75
|
+
|
|
76
|
+
- Memory sync no longer loses an edit when two devices change the same line of a memory file at the same time. Devices on this version keep both edits as whole lines, and every device converges. Before, the edits were merged character by character into text that matched neither (#12279).
|
|
77
|
+
- Between devices on this version, an edit does not disturb a line nobody edited. The exception is a very large rewrite: when one device adds and removes more than about 2,000 lines of a note in one save, that save is merged character by character, as before this fix, so a concurrent edit to the same line can still splice.
|
|
78
|
+
- When the note's body or the new text does not end with a newline, the edit is still merged character by character, as before, so two concurrent edits to that last line can still splice.
|
|
79
|
+
- With a device that still runs an older version, its old character merge can still go wrong in these ways:
|
|
80
|
+
- **Adjacent lines can fuse.** The older device rewrites the line right after a line this version edited. Depending on the two devices' internal ids, the two edits can fuse into one line followed by an empty line. Measured: from `head`, `one`, `two`, `tail`, this version writes `ONE` and the older device writes `TWO`, and the result can be `head`, `TWOONE`, an empty line, `tail`. Two older devices give `head`, `ONE`, `TWO`, `tail`.
|
|
81
|
+
- **Leftover text moves.** The older device edits a line this version deletes. Its leftover text joins a neighbouring line.
|
|
82
|
+
- **Its own diff touches other lines.** The older device's character merge crosses into a line its user did not edit. This also happens between two older devices.
|
|
83
|
+
- Two older devices keep their old character merge until they update. ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
84
|
+
- Memory sync: the push phase of a large memory set no longer needs memory in proportion to the whole set. Every push step used to re-read and re-write one journal file holding every pending item's data, several times per batch. A device pushing a few thousand memory files therefore peaked at 2 GiB or more (#12286). The journal now keeps a small index of pending items with their data in separate files, and a push reads only the batch it is sending. Capturing the set before the push still holds each captured file's text until the capture ends. A journal written by an older build is converted the first time it is read, including one where an older build added items next to the new index. ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
85
|
+
- Memory sync: a device receiving a large memory set no longer slows down with every file it lands. The render step used to re-read every render record of the cycle for each file, so the first pull of a few thousand memory files took tens of minutes (#12285). Each file now costs the same, and the crash-resume behavior is unchanged. ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
86
|
+
- Memory git no longer takes `core.hooksPath`, `core.fsmonitor`, `include.path` or `includeIf.*.path` from an inherited `GIT_CONFIG_COUNT` pair or `GIT_CONFIG_PARAMETERS` entry, so an exported pair such as `core.hooksPath=/dev/null` can no longer silently skip the memory repository's hooks and mirror push. Every other injected pair (`url.*.insteadOf`, `http.extraHeader`, credentials, `safe.directory`, ...) still applies, renumbered; an injected list git itself cannot parse is dropped whole. Memory git also no longer inherits repository redirection (`GIT_DIR`, `GIT_WORK_TREE`, `GIT_COMMON_DIR`, `GIT_OBJECT_DIRECTORY`, `GIT_NAMESPACE`, `GIT_CONFIG`, ...), so it can no longer be pointed at another repository; `GIT_INDEX_FILE`, `GIT_CONFIG_GLOBAL` and `GIT_CONFIG_NOSYSTEM` are kept. ([#12360](https://github.com/code-yeongyu/oh-my-openagent/pull/12360), fixes [#12354](https://github.com/code-yeongyu/oh-my-openagent/issues/12354)) ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
87
|
+
- Memory sync no longer stops every push from a device when one memory file's change is larger than the 8 MiB batch limit: that file is held back and named in the sync status with what to do, every other memory keeps syncing, and the file syncs on its own once it fits. A change another device makes to a held file is kept and lands in this device's copy once the file fits, and a change an older build had already queued for such a file is recovered the same way instead of staying stuck. The TUI footer shows ` sync held` and the notice says the file is being held back, and a refused entry that pauses sync is never hidden behind a held file. ([#12284](https://github.com/code-yeongyu/oh-my-openagent/issues/12284)) ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
88
|
+
- Opening a memory repository starts up to six fewer git processes: the local config defaults are read with one `git config --list` instead of one `git config --get` per key, and the line-ending repair is skipped while the repository tracks nothing. Every session that opens memory pays this, and on a loaded machine each git process costs hundreds of milliseconds. memory-core's own tests now run under the canonical 20 s test budget instead of per-file 5-15 s limits, which made them fail only on busy hosts. ([#12104](https://github.com/code-yeongyu/oh-my-openagent/pull/12104), fixes [#12103](https://github.com/code-yeongyu/oh-my-openagent/issues/12103)) ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
89
|
+
- `thread_send` no longer reports a message as `queued` when nothing will ever pick it up. When the target thread's session has no running drain, the send now answers `queued_offline`: the message is kept and the session takes it the next time it runs, so a caller knows no turn started and can act instead of waiting. The drain also reports a message it is holding behind the user's unsent draft, so a live drain is never mistaken for a missing one. ([#12488](https://github.com/code-yeongyu/oh-my-openagent/pull/12488), fixes [#12486](https://github.com/code-yeongyu/oh-my-openagent/issues/12486))
|
|
90
|
+
- Verify recursive removal (sync and async) actually removed its path, retry bounded silent leftovers, and join sweep-test removers before deleting their fixture parents. ([#11834](https://github.com/code-yeongyu/oh-my-openagent/pull/11834))
|
|
91
|
+
|
|
92
|
+
### Added
|
|
93
|
+
|
|
94
|
+
- Windows releases now carry `omo-git-job-helper.exe`, a small native helper that runs each memory git command inside its own kill-on-close job object, so a later writer can prove that a killed writer's git is gone before it reclaims a stranded lock. Nothing uses the helper yet: memory-core keeps today's Windows behavior until a follow-up wires it in, and no release claims the new protocol until its in-flight-creation stress row has passed on Windows CI. ([#12105](https://github.com/code-yeongyu/oh-my-openagent/pull/12105), part of [#10391](https://github.com/code-yeongyu/oh-my-openagent/issues/10391)) ([#12597](https://github.com/code-yeongyu/oh-my-openagent/pull/12597))
|
|
95
|
+
|
|
10
96
|
## [5.1.63] - 2026-11-07
|
|
11
97
|
|
|
12
98
|
**The macOS keychain helper's lookup paths and tests are correct on Windows CI.** The macOS install locations (the OmO.app runtime and helper copies) are built with POSIX separators on every host, and the #11262 keychain-helper tests use platform-correct paths, so dev's Windows test legs pass again. ([#12690](https://github.com/code-yeongyu/oh-my-openagent/pull/12690), part of [#11262](https://github.com/code-yeongyu/oh-my-openagent/issues/11262))
|
|
@@ -66,7 +66,7 @@ This is the dream's core duty: make memory smaller, better placed, and less redu
|
|
|
66
66
|
- Demotion is reversible: it MOVES the file to `reference/` with a `[[path]]` cross-reference at the former point of use. It never deletes content.
|
|
67
67
|
- A `system/` file whose frontmatter `expires` instant has passed is already out of context (the compiler skips it): demote it to `reference/` with its dates, and move any durable line inside it to the right `system/` file first. A temporary section inside a durable `system/` file has no machine expiry: split it into its own file with `expires: <ISO 8601>` (a UTC date or an instant with `Z` or an offset), or demote it when its stated end has passed.
|
|
68
68
|
|
|
69
|
-
**Fact archiving**: entries in `notes/facts/` older than 6 months get summarized into `ARCHIVE.md`, the single non-system root archive file. Compress them into concise dated summary entries, append those to `ARCHIVE.md
|
|
69
|
+
**Fact archiving**: entries in `notes/facts/` older than 6 months get summarized into `ARCHIVE.md`, the single non-system root archive file. Compress them into concise dated summary entries, append those to `ARCHIVE.md` with each entry naming the file it came from (for example `- 2026-03 (notes/facts/2026-03.md): ...`), and remove the summarized originals. Links to an archived file are pointed at `ARCHIVE.md` only when its entry names the file; links to anything you delete instead become plain text. Keep anything younger, and keep anything old that's still clearly load-bearing. Delete (don't archive) content the user asked to forget, sensitive or wrong content, and junk with no future-reference value.
|
|
70
70
|
|
|
71
71
|
### System Token Budget Contract
|
|
72
72
|
|