@qvac/core 0.3.3 → 0.3.5

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/CHANGELOG.md ADDED
@@ -0,0 +1,632 @@
1
+ # Changelog
2
+
3
+ All notable changes to `@qvac/core` are documented here.
4
+
5
+ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
6
+ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
7
+ While the major version is `0`, a breaking change bumps the minor version.
8
+
9
+ Route by route detail, with what a caller has to do differently, is in each
10
+ module's changelog under [`docs/api/`](docs/api/README.md). The examples below
11
+ call the routes on a `Client`, named `engine`, in the shape they had in that
12
+ release.
13
+
14
+ ## [Unreleased]
15
+
16
+ ## [0.3.5] - 2026-09-25
17
+
18
+ ### Changed
19
+
20
+ - `autobee` 2.11.4 with `strictReindex`: the boot reindex rewrites views still
21
+ on a version 1 manifest, which autobee 2.11 will no longer append to.
22
+ `corestore` 7.12.6 and `hypercore` 11.37. (#205)
23
+
24
+ ### Fixed
25
+
26
+ - A React Native app could not bundle `@qvac/core/oauth/client`: Metro rejects
27
+ this package's imports map, so `#abort-controller` did not resolve. Token
28
+ refresh now uses the runtime's own `AbortController`, and
29
+ `bare-abort-controller` only on Bare, which has none. React Native's `fetch`
30
+ gets a native signal too. (#231)
31
+
32
+ ```js
33
+ // mobile: bundles again, nothing else changes
34
+ import { createOAuthClient } from '@qvac/core/oauth/client'
35
+
36
+ const oauth = createOAuthClient({ fetch, randomBase64Url, sha256Base64Url })
37
+ const tokens = await oauth.refreshOAuthTokens(config, refreshToken)
38
+ ```
39
+
40
+ ## [0.3.4] - 2026-09-24
41
+
42
+ ### Fixed
43
+
44
+ - OAuth token refresh failed on Node 24, including Electron 42, because the
45
+ request carried a non-native `AbortSignal`. Calendar, Spotify and other
46
+ connected skills stopped working an hour after connecting. (#207)
47
+
48
+ ## [0.3.3] - 2026-09-24
49
+
50
+ ### Changed
51
+
52
+ - `@qvac/harness` 0.3.3 and `@qvac/skills` 0.1.9. The harness moves to
53
+ `@qvac/sdk` 0.20, and some addons now ship per-platform prebuild packages,
54
+ so a host that bundles Core should read the harness 0.3.3 notes. (#203)
55
+
56
+ ### Fixed
57
+
58
+ - `listChatActivity` and `chatActivityWatch` reported `generating-image` or
59
+ `generating-audio` for the whole time a media model was downloading or
60
+ loading. That wait now reads as `running`, and the media state starts with
61
+ the first sampling tick. (#177)
62
+
63
+ ## [0.3.2] - 2026-09-23
64
+
65
+ ### Added
66
+
67
+ - XLSX and PPTX uploads are indexed as Knowledge and retrieved like other
68
+ documents, including uploads that another client sends to a desktop host.
69
+ Nothing changes for the caller: upload them the way you upload a PDF.
70
+ Requires `rag-extract` 0.1.1. (#161)
71
+ - In a chat where more than one person speaks, each user turn reaches the model
72
+ with its sender's name, so an agent can tell who asked for what. A chat with
73
+ one speaker produces the same prompt as before, so its KV cache still
74
+ matches. (#102)
75
+
76
+ ### Changed
77
+
78
+ - **Breaking:** OAuth provider config moves out of Core and into each skill's
79
+ manifest. Core keeps a generic engine, and the provider-named functions are
80
+ gone: read the skill catalog, look up the spec for a credential key, and run
81
+ the flow from it. `cancelOAuthFlow` takes `{ credentialKey }`. A skill
82
+ manifest is replicated content, so `specFromRoute` drops any route it cannot
83
+ run safely. (#169)
84
+
85
+ ```js
86
+ // 0.3.1
87
+ import { connectGoogle } from '@qvac/core/oauth'
88
+
89
+ const tokens = await connectGoogle({
90
+ clientId,
91
+ clientSecret,
92
+ credentialKey: 'gmail_access_token',
93
+ openUrl
94
+ })
95
+
96
+ // 0.3.2
97
+ import { buildProviderConfig, oauthSpecFor, runOAuthFlow } from '@qvac/core/oauth'
98
+
99
+ const { skills } = await engine.skills({ agentId })
100
+ const spec = oauthSpecFor(skills, 'gmail_access_token')
101
+ const tokens = await runOAuthFlow(buildProviderConfig(spec, { clientId, clientSecret }), {
102
+ openUrl
103
+ })
104
+ ```
105
+
106
+ - `@qvac/harness` 0.3.1 and `@qvac/skills` 0.1.6. (#191)
107
+
108
+ ### Removed
109
+
110
+ - The provider-named OAuth exports (`connectGoogle`, `exchangeAsanaCode`,
111
+ `refreshSpotifyToken`, every `*_CREDENTIAL_KEY` constant and the rest),
112
+ `providerRequiresSecret`, `providerSupportsDcr`,
113
+ `providerSupportsSharedIosSession`, `OAUTH_PROVIDERS` and
114
+ `verifyOAuthTokens`. Read the same facts off the spec instead. See
115
+ [`docs/api/runtime.md`](docs/api/runtime.md) for the full list. (#169)
116
+
117
+ ```js
118
+ // instead of providerRequiresSecret(provider)
119
+ const needsSecret = spec.tokenAuth !== 'pkce-only'
120
+ // instead of providerSupportsDcr(provider)
121
+ const registers = Boolean(spec.registrationEndpoint)
122
+ // instead of providerSupportsSharedIosSession(provider)
123
+ const shared = spec.sharedIosSession === true
124
+ ```
125
+
126
+ ### Fixed
127
+
128
+ - A device that booted straight into a room showed every mesh unnamed, and
129
+ `shared`/`sharedAt` were wrong after a restart. A boot now always opens the
130
+ account next to the mesh it lands on. (#150)
131
+ - A `closeMesh` that landed while the same mesh was still opening closed
132
+ nothing, and left the mesh open and replicating. (#150)
133
+ - `duck`'s `/save` dropped the extension of every attachment that was not a
134
+ JPEG or PNG, so a saved `.mp4` or `.pdf` would not open. (#182)
135
+
136
+ ## [0.3.1] - 2026-09-22
137
+
138
+ ### Added
139
+
140
+ - The prompt inspector: `inferenceCalls` returns the model calls recorded for a
141
+ chat, merging the rounds this device ran with those the agent's host ran.
142
+ Nothing is recorded until the device-local `inspector.captureInferenceCalls`
143
+ key is set to the string `'on'`. Records stay on the device, are dropped with
144
+ their chat, and are pruned by count and age. (#58)
145
+
146
+ ```js
147
+ await engine.configSet({ key: 'inspector.captureInferenceCalls', value: 'on' })
148
+
149
+ const { calls, truncated } = await engine.inferenceCalls({ chatId })
150
+ for (const { runId, at, call } of calls ?? []) inspect(runId, at, call)
151
+ ```
152
+
153
+ - `senderMemberId` on a user message's chunk, stamped by Core, so a message
154
+ records which member sent it even after the device that wrote it is removed.
155
+ Messages from before this carry none. (#101)
156
+
157
+ ```js
158
+ const { chunks } = await engine.listChunks({ chatId, type: 'user-message' })
159
+ const { members } = await engine.listMembers({ meshId })
160
+ const sender = (chunk) =>
161
+ members.find((member) => b4a.equals(member.id, chunk.senderMemberId))?.name
162
+ ```
163
+
164
+ ### Changed
165
+
166
+ - The extraction, chunking and retrieval pipeline comes from `rag-extract`, and
167
+ behaviour is unchanged. A host that gates installs on release age has to
168
+ exempt `rag-extract`. (#112)
169
+
170
+ ```toml
171
+ # bunfig.toml
172
+ [install]
173
+ minimumReleaseAgeExcludes = ["rag-extract"]
174
+ ```
175
+
176
+ - `autobee` 2.10. The on-disk record layout is unchanged, so released peers
177
+ still read it. (#154)
178
+ - `@qvac/harness` 0.2.9, which sends `stopReason`, so `duck` now marks an
179
+ answer the context window cut short. (#163)
180
+ - `deleteChat` also drops the chat's prompt-inspector records, and fails if
181
+ that drop fails. (#58)
182
+
183
+ ### Fixed
184
+
185
+ - A new room could be created with the device's name as the member name, when
186
+ the user had already given their own name. Co-members saw the device name
187
+ until a rename caught up. (#159)
188
+
189
+ ## [0.3.0] - 2026-09-21
190
+
191
+ ### Added
192
+
193
+ - `verifyOAuthTokens` checks a fresh token before a host saves it: it refuses an
194
+ Asana token from a non-MCP app, and a Spotify account without Premium. A
195
+ network error or timeout still lets the connect go through. (#136)
196
+
197
+ ```js
198
+ import { verifyOAuthTokens } from '@qvac/core/oauth'
199
+
200
+ const result = await verifyOAuthTokens(credentialKey, tokens.accessToken, fetch)
201
+ if (!result.ok) throw new Error(result.error) // e.g. OAuth error: spotify_premium_required
202
+ ```
203
+
204
+ - `duck`: `/usage` shows how full the chat's context window is and what fills
205
+ it, the palette completes each command's verbs (tab or enter), and a new
206
+ session reopens the agent and chat the last one left active. (#132, #133,
207
+ #134)
208
+
209
+ ```
210
+ /usage the window, what fills it, and the chat's totals
211
+ /chat n completes to /chat new and runs it
212
+ ```
213
+
214
+ ### Changed
215
+
216
+ - A watch opened without a `meshId` stayed on the mesh selected when it
217
+ opened. It now ends when the selection moves, and the caller reopens it on
218
+ the mesh it wants. Omitting `meshId` is deprecated. To keep one watch across
219
+ a selection change, name its mesh. (#139)
220
+
221
+ ```js
222
+ // ends when the user selects another mesh
223
+ for await (const frame of engine.chatsWatch({})) render(frame)
224
+
225
+ // stays on the mesh it names
226
+ for await (const frame of engine.chatsWatch({ meshId })) render(frame)
227
+ ```
228
+
229
+ - `runRequestsWatch` only serves the requests whose agent names the calling
230
+ device. (#144)
231
+ - A chat is compacted only before a run starts, never after an answer. The
232
+ turn that crosses the threshold finishes untouched, and the next message
233
+ folds first. (#130)
234
+
235
+ ### Fixed
236
+
237
+ - After switching mesh, the assistant stopped claiming runs, so no model
238
+ loaded for a request on the newly selected mesh. (#139)
239
+ - `listKnowledge`, `listIndexing`, `knowledgeWatch`, `indexingWatch` and
240
+ `indexingProgressWatch` ignored the `meshId` they were given and answered
241
+ for the selected mesh instead. (#142)
242
+ - A device name had no length limit, and a very long one broke the room UI.
243
+ Names are now capped at 100 characters, and a longer one is refused. (#84)
244
+
245
+ ## [0.2.2] - 2026-09-18
246
+
247
+ ### Added
248
+
249
+ - Compute meshes: `createRoom({ compute: true })` makes a mesh that holds only
250
+ devices and members, so a device can lend its model to another device
251
+ without receiving its chats, knowledge or agents. The borrower keeps the run,
252
+ and only the completion call crosses the wire. (#67)
253
+
254
+ ```js
255
+ // the device with the model
256
+ const { meshId } = await desktop.createRoom({ compute: true })
257
+ const { invite } = await desktop.meshInvite({ member: true, meshId })
258
+
259
+ // the device that borrows it
260
+ await phone.joinMesh({ invite })
261
+ const { id, computeLenderId } = await phone.deviceInfo({})
262
+ await phone.createAgent({
263
+ name: 'Borrower',
264
+ bio: 'runs here, thinks there',
265
+ providerDeviceId: id,
266
+ modelName: 'Qwen3.5-9B',
267
+ lenderDeviceId: computeLenderId
268
+ })
269
+ ```
270
+
271
+ - Runs are served from every open mesh, not only the selected one, so a
272
+ co-member's message in a room is answered while the host is looking at
273
+ another mesh. (#93)
274
+ - `listDevices`, `devicesWatch` and `listMembers` take a `meshId`, and
275
+ `membersWatch` is new. A background run can read the rosters of its own
276
+ mesh. (#87, #122)
277
+
278
+ ```js
279
+ const { devices } = await engine.listDevices({ meshId: room.meshId })
280
+ for await (const { members } of engine.membersWatch({ meshId: room.meshId })) {
281
+ render(members)
282
+ }
283
+ ```
284
+
285
+ - `updateMember` and `removeMember`. Any member may remove another, and doing
286
+ so revokes every device of that member at once. A removed member no longer
287
+ appears in `listMembers`. A device cannot remove its own member. (#89, #92)
288
+
289
+ ```js
290
+ await engine.updateMember({ id: memberId, name: 'Ada at work', meshId })
291
+ await engine.removeMember({ id: otherMemberId, meshId })
292
+ ```
293
+
294
+ - `shared` and `sharedAt` on each mesh entry, marking the room a member invite
295
+ reuses. (#91)
296
+
297
+ ```js
298
+ const { meshes } = await engine.listMeshes({})
299
+ const sharedRoom = meshes.find((mesh) => mesh.shared)
300
+ ```
301
+
302
+ - `listArchivedMeshes`, `archivedMeshContents` and `migrateArchivedMesh`: after
303
+ a device invite moves a device onto another account, the account it left can
304
+ be counted and its chats, agents and knowledge copied across. A second
305
+ migrate answers `migrated: false`. (#105)
306
+
307
+ ```js
308
+ await phone.joinMesh({ invite: (await laptop.meshInvite({})).invite })
309
+
310
+ const [left] = (await phone.listArchivedMeshes({})).meshes ?? []
311
+ const { chats, agents, knowledge } = await phone.archivedMeshContents({ meshId: left.meshId })
312
+ if (await askUser(`Bring over ${chats} chats, ${agents} agents, ${knowledge} documents?`)) {
313
+ await phone.migrateArchivedMesh({ meshId: left.meshId })
314
+ }
315
+ ```
316
+
317
+ - Account recovery from the identity phrase: `accountRecoverable`,
318
+ `recoveryPublish` and `recoverAccount`. A copy of the account seed that only
319
+ the phrase can open is mirrored by the blind peers, so a wiped device can get
320
+ its account back. An account made before this has to be published once.
321
+ (#107)
322
+
323
+ ```js
324
+ // once, on a device that has the account
325
+ const { recoverable } = await engine.accountRecoverable({})
326
+ if (!recoverable) await engine.recoveryPublish({ phrase })
327
+
328
+ // on a wiped device
329
+ await engine.recoverAccount({ phrase, timeout: 30_000 })
330
+ ```
331
+
332
+ - `displayName` and `setDisplayName`: a member is named after its user rather
333
+ than after the device that founded it. The account is minted before
334
+ onboarding runs, so set the name once the user has entered it. (#108, #109)
335
+
336
+ ```js
337
+ await engine.setDisplayName({ name: 'Ada' })
338
+ const { name } = await engine.displayName({})
339
+ ```
340
+
341
+ - `sortChatGroupsByName`. Groups nobody has arranged sort by name, and this
342
+ drops every arrangement to go back to that order. (#110)
343
+
344
+ ```js
345
+ const { groups } = await engine.sortChatGroupsByName({})
346
+ ```
347
+
348
+ - Transcription resolves the newest audio attachment when no id is given.
349
+ (#66)
350
+ - `duck`: `/leave`, `/compact`, a marker where a chat was compacted, `/ctx`
351
+ and a menu entry to set the agent's context size, and a note when the
352
+ context window cut an answer short. (#86, #127, #129)
353
+
354
+ ```
355
+ /leave leave the mesh in use
356
+ /compact compact the chat now
357
+ /ctx 8k run the agent at an 8,192-token window; /ctx auto clears it
358
+ ```
359
+
360
+ ### Changed
361
+
362
+ - An approval rule records the member who wrote it, and only applies to that
363
+ member. In a shared room, one member's standing rule no longer approves
364
+ another member's tool calls. A rule written before this has no author and
365
+ applies to nobody, so its call is asked again. (#123)
366
+
367
+ ```js
368
+ const { rules, ourMemberId } = await engine.listApprovalRules({ meshId })
369
+ const mine = rules.filter((rule) => rule.memberId && b4a.equals(rule.memberId, ourMemberId))
370
+ ```
371
+
372
+ - A room's default agent is named after the member whose device runs it,
373
+ rather than after the built-in agent. (#96)
374
+ - When a member leaves a room on one device, the room is also retired on that
375
+ member's other devices. (#95)
376
+
377
+ ### Fixed
378
+
379
+ - The turn that triggered compaction lost its own answer from later prompts,
380
+ and on 4k and 8k windows compaction ran a turn too late. The threshold now
381
+ depends on the window size. (#126)
382
+ - When two devices of one member redeemed the same member invite, the room
383
+ ended up with two members instead of one member with two devices. (#111)
384
+ - With a room selected, a device invite targeted the room and was refused. It
385
+ now targets the account. (#81)
386
+ - The inviter's `meshesWatch` never saw a member join a room it was not
387
+ looking at. (#100)
388
+ - A member invite warned that a room with only its own member in it was not
389
+ empty. (#97)
390
+ - With Knowledge off, a document over 20 MiB was not read at all. Its opening
391
+ is now read whatever the file's size. (#83)
392
+
393
+ ### Security
394
+
395
+ - Once a mesh holds a verified device signature, it refuses unsigned device
396
+ rows. Before this, any admitted writer could create or rewrite a device row
397
+ for a key it did not hold. Meshes that predate this are sealed by the first
398
+ signing build to boot on them. (#88)
399
+
400
+ ## [0.2.1] - 2026-09-16
401
+
402
+ ### Fixed
403
+
404
+ - When a user attached many documents and asked a topic-specific question,
405
+ retrieval could leave some of them out. Every attached file now gets a share
406
+ of the token budget. (#64)
407
+
408
+ ## [0.2.0] - 2026-09-15
409
+
410
+ ### Added
411
+
412
+ - Member invites: `meshInvite({ member: true })` opens a shared room and
413
+ invites a second person into it as a member of their own, keeping their
414
+ identity. The account is never invitable this way. `listMembers` returns the
415
+ members, and the response's `notice` says when the room already holds
416
+ people. (#49, #57)
417
+
418
+ ```js
419
+ // Ada
420
+ const { invite, meshId, notice } = await ada.meshInvite({ member: true })
421
+
422
+ // Ben, on his own device
423
+ await ben.joinMesh({ invite })
424
+ ```
425
+
426
+ - Multiple meshes per device: `listMeshes`, `meshesWatch`, `openMesh`,
427
+ `selectMesh`, `closeMesh` and `setMeshLabel`, and a `meshId` on the agent
428
+ and chat calls. Joining a room keeps the mesh already in use, the member's
429
+ other devices pick the room up, and each member gets their own default agent
430
+ in it, on their own device. (#50, #51, #52, #53, #54, #55)
431
+
432
+ ```js
433
+ const { meshes } = await engine.listMeshes({})
434
+ // [{ meshId, state, writable, open, active, home, label, members }, …]
435
+
436
+ const room = meshes.find((mesh) => !mesh.home)
437
+ await engine.setMeshLabel({ meshId: room.meshId, label: 'Family' })
438
+ await engine.selectMesh({ meshId: room.meshId })
439
+ const { chats } = await engine.listChats({ meshId: room.meshId })
440
+ ```
441
+
442
+ - `moveChat` and `moveChatGroup`: chats and groups can be dragged into an
443
+ order. A dropped row lands directly below `afterId`, or at the top when it is
444
+ left out, and a chat can change group in the same call. (#44)
445
+
446
+ ```js
447
+ await engine.moveChat({ id: chatId, groupId, afterId: chatAboveId })
448
+ await engine.moveChat({ id: chatId }) // ungrouped, at the top
449
+ await engine.moveChatGroup({ id: groupId, afterId: groupAboveId })
450
+ ```
451
+
452
+ ### Changed
453
+
454
+ - **Breaking:** a device is adopted into a member only if it presents a claim,
455
+ so pairing needs an invite from an upgraded device. A device still on an
456
+ older build connects, but is not listed as yours until it is updated and
457
+ paired again. (#70)
458
+
459
+ ```js
460
+ // on the upgraded device
461
+ const { invite } = await laptop.meshInvite({})
462
+ // on the device being paired, once it runs 0.2.0
463
+ await phone.joinMesh({ invite })
464
+ ```
465
+
466
+ - **Breaking:** the member join is a new mesh route, and
467
+ `VENDORED_MIN_SUPPORTED_VERSION` moves to `>=0.18.0`. A peer on an older
468
+ build withdraws a mesh as needing an upgrade once a joined room is recorded
469
+ in it. (#49, #55)
470
+ - `home` on a mesh entry now marks the account mesh, not the mesh this device
471
+ minted. Leaving or being kicked archives a mesh instead of deleting it, so
472
+ its seed survives. (#56)
473
+
474
+ ## [0.1.7] - 2026-09-17
475
+
476
+ Published under the `release-0.1` dist-tag.
477
+
478
+ ### Added
479
+
480
+ - `moveChat` and `moveChatGroup`, and groups nobody has arranged sort by name,
481
+ backported from 0.2.0 and 0.2.2. See 0.2.0 for an example. (#44, #110)
482
+
483
+ ## [0.1.6] - 2026-09-16
484
+
485
+ Published under the `release-0.1` dist-tag.
486
+
487
+ ### Changed
488
+
489
+ - `@qvac/harness` 0.2.7.
490
+
491
+ ### Fixed
492
+
493
+ - Every attached file gets a share of the retrieval budget, backported from
494
+ 0.2.1. (#64)
495
+
496
+ ## [0.1.5] - 2026-09-15
497
+
498
+ ### Added
499
+
500
+ - Joining through a fast-forward loads the device, agent, member and chat lists
501
+ the app needs to start, so a cold join no longer stalls on "Requesting host
502
+ setup". Nothing to call: it runs on every fast-forward. (#43)
503
+
504
+ ### Fixed
505
+
506
+ - Mesh keys are derived from the `qvac-assistant/mesh` namespace again. Since
507
+ 0.1.0 a seed stored by the `qvac-assistant` package opened as a fresh, empty
508
+ mesh with no error, and its chats, agents and groups were left behind on
509
+ disk. They come back with this release. A mesh first created by 0.1.0 to
510
+ 0.1.4 derives differently here. (#59)
511
+ - After a few skill turns, the prompt could end on a tool turn, so the model
512
+ narrated loading a skill or gave no answer. (#48)
513
+
514
+ ## [0.1.4] - 2026-09-12
515
+
516
+ ### Added
517
+
518
+ - `duck --help` and `duck --version`. Both answer without loading the engine.
519
+ (#36)
520
+
521
+ ```sh
522
+ $ duck --version
523
+ @qvac/core 0.1.4
524
+ @qvac/harness 0.2.4
525
+ @qvac/sdk 0.19.0
526
+ ```
527
+
528
+ ### Changed
529
+
530
+ - `duck` draws its first frame before loading the engine, and
531
+ `QVAC_INPROCESS` is gone: `duck` runs the same sidecar topology as the
532
+ desktop app. (#37)
533
+
534
+ ### Fixed
535
+
536
+ - The README image on npm was broken, so `ducks.png` ships in the package
537
+ again. (#36)
538
+
539
+ ## [0.1.3] - 2026-09-12
540
+
541
+ ### Removed
542
+
543
+ - **Breaking:** the `./dist/assistant.cjs` export. Nothing relocated the bundle
544
+ this package published. A host that spawns Core from its own bundle stows
545
+ `worker/assistant-sidecar.mjs` in its own build, as the desktop and mobile
546
+ apps do, and passes the result as `entry`. (#33)
547
+
548
+ ```js
549
+ const engine = new AssistantClient({
550
+ storagePath,
551
+ device,
552
+ entry: '/path/to/your/assistant.bundle'
553
+ })
554
+ ```
555
+
556
+ ### Fixed
557
+
558
+ - 0.1.2 installed and then hung with no output: 14 of the 47 native addons
559
+ outside the published bundle were at a different version from the ones the
560
+ bundle was built with. `duck` now runs from the installed files, and the
561
+ tarball shrinks from 11.9 MB to 517 KB. Requires `@qvac/harness` 0.2.4. (#33)
562
+
563
+ ## [0.1.2] - 2026-09-11
564
+
565
+ ### Added
566
+
567
+ - `duck`, the terminal app, ships in the package. It runs the same assistant
568
+ sidecar that the desktop app ships. Joining a mesh is `/join` inside the app,
569
+ not an argument at startup. `bare` has to be on your PATH. (#30)
570
+
571
+ ```sh
572
+ npm i -g @qvac/core
573
+ duck
574
+ ```
575
+
576
+ ## [0.1.1] - 2026-09-11
577
+
578
+ ### Added
579
+
580
+ - A `./wire-codec` export for the generated hyperschema messages, so a host can
581
+ decode what the real decoder produces. (#28)
582
+
583
+ ```js
584
+ import { decode, encode } from '@qvac/core/wire-codec'
585
+
586
+ const chunk = decode('@assistant/chunk', encode('@assistant/chunk', value))
587
+ ```
588
+
589
+ ### Fixed
590
+
591
+ - The compiled library could not load: eight modules in `dist-lib` imported
592
+ `schema/helpers.js`, which was missing from the package. (#28)
593
+
594
+ ## [0.1.0] - 2026-09-11
595
+
596
+ ### Added
597
+
598
+ - First release as `@qvac/core`. It contains `Core`, `Client` and
599
+ `QvacAssistant`, moved out of `qvac-app` with their history. The hosting
600
+ contract and the schema codegen come from `@qvac/harness`. (#1, #4, #7, #11)
601
+
602
+ ```js
603
+ import { AssistantClient, QvacAssistant } from '@qvac/core'
604
+
605
+ const engine = new AssistantClient({
606
+ storagePath: '/path/to/state',
607
+ device: { name: 'laptop', capabilityProfile: 'high', recommendedModelName: 'Qwen3.5-9B' }
608
+ })
609
+ await engine.ready()
610
+
611
+ const assistant = new QvacAssistant(engine, harness)
612
+ await assistant.ready()
613
+ ```
614
+
615
+ [Unreleased]: https://github.com/tetherto/qvac-core/compare/v0.3.5...HEAD
616
+ [0.3.5]: https://github.com/tetherto/qvac-core/compare/v0.3.4...v0.3.5
617
+ [0.3.4]: https://github.com/tetherto/qvac-core/compare/v0.3.3...v0.3.4
618
+ [0.3.3]: https://github.com/tetherto/qvac-core/compare/v0.3.2...v0.3.3
619
+ [0.3.2]: https://github.com/tetherto/qvac-core/compare/v0.3.1...v0.3.2
620
+ [0.3.1]: https://github.com/tetherto/qvac-core/compare/v0.3.0...v0.3.1
621
+ [0.3.0]: https://github.com/tetherto/qvac-core/compare/v0.2.2...v0.3.0
622
+ [0.2.2]: https://github.com/tetherto/qvac-core/compare/v0.2.1...v0.2.2
623
+ [0.2.1]: https://github.com/tetherto/qvac-core/compare/v0.2.0...v0.2.1
624
+ [0.2.0]: https://github.com/tetherto/qvac-core/compare/v0.1.5...v0.2.0
625
+ [0.1.7]: https://github.com/tetherto/qvac-core/compare/v0.1.6...v0.1.7
626
+ [0.1.6]: https://github.com/tetherto/qvac-core/compare/v0.1.5...v0.1.6
627
+ [0.1.5]: https://github.com/tetherto/qvac-core/compare/v0.1.4...v0.1.5
628
+ [0.1.4]: https://github.com/tetherto/qvac-core/compare/v0.1.3...v0.1.4
629
+ [0.1.3]: https://github.com/tetherto/qvac-core/compare/v0.1.2...v0.1.3
630
+ [0.1.2]: https://github.com/tetherto/qvac-core/compare/v0.1.1...v0.1.2
631
+ [0.1.1]: https://github.com/tetherto/qvac-core/compare/v0.1.0...v0.1.1
632
+ [0.1.0]: https://github.com/tetherto/qvac-core/releases/tag/v0.1.0
@@ -211,7 +211,8 @@ export class Mesh extends ReadyResource {
211
211
  encryptionKey: this._encryptionKey,
212
212
  wakeupCapability: { key: this._keyPair.publicKey, discoveryKey: this.discoveryKey },
213
213
  wakeup: this._wakeup,
214
- fastForward: { boot: this._bootFrom ? { head: this._bootFrom } : null }
214
+ fastForward: { boot: this._bootFrom ? { head: this._bootFrom } : null },
215
+ strictReindex: true
215
216
  });
216
217
  this.autobee.on('error', (err) => this._log?.warn(this._scope, 'autobee error', err));
217
218
  this.autobee.on('move-to', (to, from) => this._log?.info(this._scope, 'fast-forwarded onto a head', {
@@ -1,6 +1,7 @@
1
- import { AbortController } from 'bare-abort-controller';
1
+ import { AbortController as BareAbortController } from 'bare-abort-controller';
2
2
  import * as z from 'zod';
3
3
  const TOKEN_REQUEST_TIMEOUT_MS = 30_000;
4
+ const AbortController = globalThis.AbortController ?? BareAbortController;
4
5
  const TokenResponseSchema = z.object({
5
6
  access_token: z.string(),
6
7
  token_type: z.string(),
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@qvac/core",
3
- "version": "0.3.3",
3
+ "version": "0.3.5",
4
4
  "type": "module",
5
5
  "description": "Peer-to-peer AI across your devices",
6
6
  "exports": {
@@ -224,7 +224,7 @@
224
224
  "@qvac/harness": "^0.3.3",
225
225
  "@qvac/skills": "^0.1.9",
226
226
  "asenion": "^0.1.0",
227
- "autobee": "^2.10.0",
227
+ "autobee": "^2.11.4",
228
228
  "b4a": "^1.8.1",
229
229
  "bare-abort-controller": "^1.1.2",
230
230
  "bare-buffer": "^3.6.1",
@@ -255,7 +255,7 @@
255
255
  "blind-peer": "^3.10.2",
256
256
  "blind-peering": "^2.7.0",
257
257
  "compact-encoding": "^3.4.0",
258
- "corestore": "^7.11.0",
258
+ "corestore": "^7.12.6",
259
259
  "hrpc": "^4.3.0",
260
260
  "hyperblobs": "^2.12.1",
261
261
  "hyperconf": "^1.0.3",
@@ -335,7 +335,7 @@
335
335
  "lodash": "4.18.1",
336
336
  "hyperschema": "1.26.2",
337
337
  "hyperdb": "^6.8.0",
338
- "hypercore": "^11.36.1",
338
+ "hypercore": "^11.37.0",
339
339
  "tar": "^7.5.22"
340
340
  },
341
341
  "publishConfig": {