dsh-agy-link 0.6.12 → 0.6.14

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 CHANGED
@@ -1,5 +1,41 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.6.14 (2026-11-10)
4
+
5
+ ### English
6
+
7
+ - **Feat(stability): optional `maxTurnMinutes` wall-clock cap — a self-driving turn can no longer run unbounded (issue #92).**
8
+ - The idle watchdog is re-armed by every output line, so an agy turn that keeps emitting events (async `task-xxxx` receipts, tool chatter) never trips it: `longToolIdleMinutes` (default 240) only bounds *silence*. A new `maxTurnMinutes` bounds the whole turn regardless of activity; when it fires the runner kills the agy process tree (SIGTERM group → SIGKILL escalation, `taskkill /T /F` on Windows) and the turn fails with a resumable `maxTurnMinutes` error.
9
+ - Default `0` = **off**, so behaviour is unchanged until you opt in (e.g. `DSH_AGY_MAX_TURN_MINUTES=60` for exploratory sessions). Env + config key, README/CONFIGURATION rows, and tests for the cap and the config precedence.
10
+
11
+ ### 中文
12
+
13
+ - **新增(稳定性):可选 `maxTurnMinutes` 墙钟上限——自驱动轮次不再可能无限运行(issue #92)。**
14
+ - 空闲看门狗会被每一行输出重置,所以「持续有输出」的 agy 轮次(异步 `task-xxxx` 回执、工具碎碎念)永远不会触发;`longToolIdleMinutes`(默认 240)只约束**静默**。新增 `maxTurnMinutes` 约束整轮墙钟,触发时 runner 会杀掉 agy 进程树(组 `SIGTERM`→`SIGKILL`,Windows `taskkill /T /F`),并返回可续跑的 `maxTurnMinutes` 错误。
15
+ - 默认 `0` = **关闭**,不改变现有行为;需要时显式开启(如排障类会话设 `DSH_AGY_MAX_TURN_MINUTES=60`)。含 env + 配置键、README/CONFIGURATION 说明,以及上限与配置优先级的测试。
16
+
17
+ **Full Changelog**: https://github.com/amlyczz/dsh-agy-link/compare/v0.6.13...v0.6.14
18
+
19
+ ## 0.6.13 (2026-11-10)
20
+
21
+ ### English
22
+
23
+ - **Fix(models/warmup): follow the 5.5 migration everywhere — no shipped default points at the retired Claude 4.6 any more.**
24
+ - Found during the 5-day architecture review: after 0.6.11/0.6.12 moved the catalog to Claude 5.5, four places still pinned 4.6 — the `warmupModels` config default (`types.ts`), the warmup family fallback and its internal default list (`warmup.ts`), and the cross-family failover map (`adapter.ts`). With 4.6 retired upstream, an auto-warmup probe or a family failover could therefore target a model that no longer exists.
25
+ - All four now use `claude-sonnet-5-5`; legacy accounts still get their own 4.6 picked from `quotas.anthropic.models` first (per-account resolution from #88), so nothing regresses for older plans.
26
+ - Bare name aliases (`claude-sonnet`, `claude-opus`) now resolve to the 5.5 line, while explicit `claude-*-4.x` names keep resolving to 4.6 (legacy sessions untouched).
27
+ - Tests: warmup default + anthropic fallback + alias resolution; README/CONFIGURATION warmup rows and the `/tsl` example updated.
28
+
29
+ ### 中文
30
+
31
+ - **修复(models/warmup):处处跟上 5.5 迁移——不再有任何出厂默认指向已下线的 Claude 4.6。**
32
+ - 5 天架构 review 发现:0.6.11/0.6.12 把目录迁到 Claude 5.5 后,仍有四处钉着 4.6——`warmupModels` 配置默认值(`types.ts`)、预热族回退与内部默认列表(`warmup.ts`)、跨族失败转移映射(`adapter.ts`)。4.6 上游已退役,自动预热或跨族回退可能打到一个不存在的模型。
33
+ - 四处统一改为 `claude-sonnet-5-5`;旧账号仍会**优先**从 `quotas.anthropic.models` 里取自己的 4.6(#88 的按账号解析),不会回退。
34
+ - 裸名别名(`claude-sonnet`、`claude-opus`)解析到 5.5 世代;显式 `claude-*-4.x` 仍解析到 4.6(不影响历史会话)。
35
+ - 测试:预热默认值、anthropic 回退、别名解析;README/CONFIGURATION 的预热行与 `/tsl` 示例同步更新。
36
+
37
+ **Full Changelog**: https://github.com/amlyczz/dsh-agy-link/compare/v0.6.12...v0.6.13
38
+
3
39
  ## 0.6.12 (2026-11-09)
4
40
 
5
41
  ### English
package/README.md CHANGED
@@ -162,13 +162,14 @@ dsh plugin --profile web add dsh-agy-link
162
162
  | `hiddenModels` | `DSH_AGY_HIDDEN_MODELS` | `[]` | 在 `/model` 选择器中隐藏的模型 slug 列表(黑名单语义,新模型默认可见;建议直接在设置面板"模型显示"中点选管理) |
163
163
  | `timeoutMs` | `DSH_AGY_TIMEOUT_MS` | `600000` | 单轮活跃看门狗超时(毫秒) |
164
164
  | `longToolIdleMinutes` | `DSH_AGY_LONG_TOOL_IDLE_MINUTES` | `240` | 存在未完成工具步骤(ACTIVE)时的空闲预算(分钟);`0` = 不抬高预算,卡死的工具按 `timeoutMs` 快速失败 |
165
+ | `maxTurnMinutes` | `DSH_AGY_MAX_TURN_MINUTES` | `0` | 单轮**墙钟**硬上限(分钟),`0` = 不限。空闲看门狗会被每一行输出重置,因此持续自驱动的会话只能靠它兜底;超限会终止 agy 进程树并返回可续跑错误(issue #92) |
165
166
  | `contextWindowDefault` | `DSH_AGY_CONTEXT_WINDOW` | `1048576` | 上报给 DSH 的模型上下文窗口(token)。agy 的模型列表不上报窗口,所以统一用该值;若你的模型实际是 256K(如部分 Claude/Gemini 配置),设为 `262144` 即可让「上下文已用 %」正确 |
166
167
  | `maxTokensDefault` | `DSH_AGY_MAX_TOKENS` | `65536` | 单次输出 token 上限(上报给 DSH) |
167
168
  | `workspaceRoot` | `DSH_AGY_WORKSPACE_ROOT` | 会话 cwd | agy 工作区根目录(默认跟随当前会话工作区) |
168
169
  | `proxyUrl` | `DSH_AGY_PROXY_URL` | 空字符串 | 全局代理 URL;若在控制台中单独配置了代理,优先使用账号独立代理 |
169
170
  | `suppressAgyAutoUpdate` | `DSH_AGY_SUPPRESS_AUTOUPDATE` | Windows `true`,其他平台 `false` | 每次 spawn 前刷新 agy 的更新检查时间戳,阻断 `--bg-updater` 子进程链(Windows 闪窗根因)。关闭后 agy 恢复自动更新检查 |
170
171
  | `autoWarmup` | `DSH_AGY_AUTO_WARMUP` | `false` | 7 天周配额自动预热(opt-in):周额度到达重置时间或冷启动时发送极小探活请求,提前激活 7 天计时器。开启后每 5 分钟巡检一次;该路径直连 Google 内部端点,请自行评估风控风险 |
171
- | `warmupModels` | `DSH_AGY_WARMUP_MODELS` | `gemini-3.8-flash,claude-sonnet-4-6` | 自动预热使用的模型列表(逗号分隔);空数组表示不自动预热 |
172
+ | `warmupModels` | `DSH_AGY_WARMUP_MODELS` | `gemini-3.8-flash,claude-sonnet-5-5` | 自动预热使用的模型列表(逗号分隔);空数组表示不自动预热 |
172
173
  | `autoTranslate` | `DSH_AGY_AUTO_TRANSLATE` | `false` | 每轮对话完成后自动调用 `/tsl` 生成大白话翻译卡片;亦可在会话中通过 `/tsl auto on` 动态开启 |
173
174
  | `translateModel` | `DSH_AGY_TRANSLATE_MODEL` | `gemini-3.8-flash` | `/tsl` 默认使用的翻译模型,缺失时按目录回退 |
174
175
  | `translateEffort` | `DSH_AGY_TRANSLATE_EFFORT` | `high` | 翻译模型思考强度档位 |
@@ -188,7 +189,7 @@ dsh plugin --profile web add dsh-agy-link
188
189
  /tsl 3 # 翻译指定第 3 轮回复
189
190
  /tsl --full # 连同中间工具调用的叙述一并翻译
190
191
  /tsl --text 任意文本… # 纯文本翻译模式(在终端直接输出,不生成卡片)
191
- /tsl --model claude-sonnet-4-6 --effort medium # 临时指定翻译模型与思考强度
192
+ /tsl --model claude-sonnet-5-5 --effort medium # 临时指定翻译模型与思考强度
192
193
  /tsl --print # 在命令执行输出中同步回显译文
193
194
  /tsl auto on|off|status # 针对当前会话开启、关闭或查看自动翻译状态
194
195
  ```
@@ -395,13 +396,14 @@ dsh plugin --profile web add dsh-agy-link
395
396
  | `hiddenModels` | `DSH_AGY_HIDDEN_MODELS` | `[]` | Model slugs hidden from the `/model` picker (deny-list, newly discovered models stay visible; manage them via the "Model visibility" row in the settings panel) |
396
397
  | `timeoutMs` | `DSH_AGY_TIMEOUT_MS` | `600000` | Activity watchdog timeout in milliseconds |
397
398
  | `longToolIdleMinutes` | `DSH_AGY_LONG_TOOL_IDLE_MINUTES` | `240` | Idle budget while a tool step is ACTIVE (minutes); `0` keeps the normal `timeoutMs` budget so wedged tools fail fast |
399
+ | `maxTurnMinutes` | `DSH_AGY_MAX_TURN_MINUTES` | `0` | Hard **wall-clock** cap per turn (minutes); `0` = unlimited. The idle watchdog is re-armed on every output line, so a self-driving session can only be bounded by this; exceeding it kills the agy process tree and returns a resumable error (issue #92) |
398
400
  | `contextWindowDefault` | `DSH_AGY_CONTEXT_WINDOW` | `1048576` | Model context window (tokens) reported to DSH. agy's model list carries no window, so this single value is used for every model; if yours is really 256K, set `262144` and the "context used %" becomes correct |
399
401
  | `maxTokensDefault` | `DSH_AGY_MAX_TOKENS` | `65536` | Output token cap reported to DSH |
400
402
  | `workspaceRoot` | `DSH_AGY_WORKSPACE_ROOT` | session cwd | Working directory root |
401
403
  | `proxyUrl` | `DSH_AGY_PROXY_URL` | empty string | Global proxy URL; per-account proxies configured in console take precedence |
402
404
  | `suppressAgyAutoUpdate` | `DSH_AGY_SUPPRESS_AUTOUPDATE` | `true` on Windows, `false` elsewhere | Refreshes agy's update-check timestamp before every spawn so its `--bg-updater` child chain (the Windows console-flash root cause) never starts. `false` restores agy's own auto-update checks |
403
405
  | `autoWarmup` | `DSH_AGY_AUTO_WARMUP` | `false` | Opt-in 7-day weekly quota warmup: sends a minimal probe when the weekly window has reset or is cold, activating the 7-day timer. Scans every 5 minutes; this path talks directly to Google's internal endpoints, so assess the risk-control tradeoff yourself |
404
- | `warmupModels` | `DSH_AGY_WARMUP_MODELS` | `gemini-3.8-flash,claude-sonnet-4-6` | Model list used for warmup probes (comma-separated); an empty list disables automatic warmup |
406
+ | `warmupModels` | `DSH_AGY_WARMUP_MODELS` | `gemini-3.8-flash,claude-sonnet-5-5` | Model list used for warmup probes (comma-separated); an empty list disables automatic warmup |
405
407
  | `autoTranslate` | `DSH_AGY_AUTO_TRANSLATE` | `false` | Auto plain-language translation after every completed turn (card under the reply). Per-session override: `/tsl auto on\|off` |
406
408
  | `translateModel` | `DSH_AGY_TRANSLATE_MODEL` | `gemini-3.8-flash` | Model used by `/tsl` and auto mode; falls back through the live catalog when absent |
407
409
  | `translateEffort` | `DSH_AGY_TRANSLATE_EFFORT` | `high` | Reasoning effort for effort-capable translate models |
package/dist/index.js CHANGED
@@ -78,6 +78,7 @@ function defaultConfig() {
78
78
  defaultEffort: "high",
79
79
  timeoutMs: 6e5,
80
80
  longToolIdleMinutes: 240,
81
+ maxTurnMinutes: 0,
81
82
  /**
82
83
  * Hard cap (minutes) on waiting for the Windows credential slot before the
83
84
  * dispatch releases, backs off and retries (dispatchRetries). Each attempt
@@ -136,7 +137,7 @@ function defaultConfig() {
136
137
  proxyUrl: "",
137
138
  suppressAgyAutoUpdate: process.platform === "win32",
138
139
  autoWarmup: false,
139
- warmupModels: ["gemini-3.8-flash", "claude-sonnet-4-6"],
140
+ warmupModels: ["gemini-3.8-flash", "claude-sonnet-5-5"],
140
141
  autoTranslate: false,
141
142
  translateModel: "gemini-3.8-flash",
142
143
  translateEffort: "high",
@@ -530,6 +531,7 @@ function resolveConfig(entry, env = process.env, overrides = readOverrides()) {
530
531
  defaultEffort: asString(get("defaultEffort")) ?? base.defaultEffort,
531
532
  timeoutMs: asNum(get("timeoutMs")) ?? base.timeoutMs,
532
533
  longToolIdleMinutes: asNum(get("longToolIdleMinutes")) ?? base.longToolIdleMinutes,
534
+ maxTurnMinutes: asNum(get("maxTurnMinutes")) ?? base.maxTurnMinutes,
533
535
  dispatchTimeoutMinutes: asNum(get("dispatchTimeoutMinutes")) ?? base.dispatchTimeoutMinutes,
534
536
  dispatchRetries: asNum(get("dispatchRetries")) ?? base.dispatchRetries,
535
537
  parallelSameAccount: asBool(get("parallelSameAccount")) ?? base.parallelSameAccount,
@@ -596,6 +598,10 @@ function resolveConfig(entry, env = process.env, overrides = readOverrides()) {
596
598
  const t = asNum(env.DSH_AGY_LONG_TOOL_IDLE_MINUTES);
597
599
  if (t !== void 0 && t >= 0) cfg.longToolIdleMinutes = t;
598
600
  }
601
+ if (env.DSH_AGY_MAX_TURN_MINUTES) {
602
+ const mt = asNum(env.DSH_AGY_MAX_TURN_MINUTES);
603
+ if (mt !== void 0 && mt >= 0) cfg.maxTurnMinutes = mt;
604
+ }
599
605
  if (env.DSH_AGY_CONTEXT_WINDOW) {
600
606
  const w = asNum(env.DSH_AGY_CONTEXT_WINDOW);
601
607
  if (w !== void 0 && w > 0) cfg.contextWindowDefault = Math.floor(w);
@@ -2739,8 +2745,10 @@ var ModelCatalog = class ModelCatalog {
2739
2745
  };
2740
2746
  function resolveModelSlug(id) {
2741
2747
  const s = id.trim().toLowerCase();
2742
- if (s === "claude-opus-4-6" || s === "claude-opus-4-8" || s === "claude-opus" || s === "claude-opus-4.6" || s === "claude-opus-4-5") return "claude-opus-4-6-thinking";
2743
- if (s === "claude-sonnet" || s === "claude-sonnet-4.6" || s === "claude-sonnet-4-5") return "claude-sonnet-4-6";
2748
+ if (s === "claude-opus") return "claude-opus-5-5";
2749
+ if (s === "claude-sonnet") return "claude-sonnet-5-5";
2750
+ if (s === "claude-opus-4-6" || s === "claude-opus-4-8" || s === "claude-opus-4.6" || s === "claude-opus-4-5") return "claude-opus-4-6-thinking";
2751
+ if (s === "claude-sonnet-4.6" || s === "claude-sonnet-4-5") return "claude-sonnet-4-6";
2744
2752
  if (s === "gpt-oss-120b" || s === "gpt-oss-20b" || s === "gpt-oss") return "gpt-oss-120b-medium";
2745
2753
  const claude55 = s.match(/^(claude-(?:sonnet|opus)-5-5)-(?:low|medium|high)$/);
2746
2754
  if (claude55 !== null && claude55[1] !== void 0) return claude55[1];
@@ -3963,6 +3971,7 @@ function startAgyProcess(opts) {
3963
3971
  let stderr = "";
3964
3972
  let timedOut = false;
3965
3973
  let aborted = false;
3974
+ let capped = false;
3966
3975
  let settled = false;
3967
3976
  child.stdin?.on("error", () => {});
3968
3977
  if (opts.stdinPayload !== void 0) try {
@@ -3983,6 +3992,14 @@ function startAgyProcess(opts) {
3983
3992
  }, idleBudgetMs);
3984
3993
  };
3985
3994
  refreshWatchdog();
3995
+ let turnCap = null;
3996
+ if (opts.maxTurnMs !== void 0 && opts.maxTurnMs > 0) {
3997
+ turnCap = setTimeout(() => {
3998
+ capped = true;
3999
+ killTree(child);
4000
+ }, opts.maxTurnMs);
4001
+ turnCap.unref?.();
4002
+ }
3986
4003
  const onAbort = () => {
3987
4004
  aborted = true;
3988
4005
  killTree(child);
@@ -4017,6 +4034,7 @@ function startAgyProcess(opts) {
4017
4034
  if (settled) return;
4018
4035
  settled = true;
4019
4036
  if (watchdog) clearTimeout(watchdog);
4037
+ if (turnCap) clearTimeout(turnCap);
4020
4038
  opts.signal?.removeEventListener("abort", onAbort);
4021
4039
  const stdoutTail = stdoutDecoder.flush();
4022
4040
  if (stdoutTail !== "") {
@@ -4034,6 +4052,7 @@ function startAgyProcess(opts) {
4034
4052
  signal,
4035
4053
  timedOut,
4036
4054
  aborted,
4055
+ capped,
4037
4056
  stdout,
4038
4057
  stderrTail: stderr,
4039
4058
  durationMs: Date.now() - started
@@ -4053,6 +4072,7 @@ function startAgyProcess(opts) {
4053
4072
  if (reason === "timeout") timedOut = true;
4054
4073
  else aborted = true;
4055
4074
  if (watchdog) clearTimeout(watchdog);
4075
+ if (turnCap) clearTimeout(turnCap);
4056
4076
  killTree(child);
4057
4077
  }
4058
4078
  };
@@ -31544,7 +31564,7 @@ var AgyAdapter = class AgyAdapter extends LlmAdapter {
31544
31564
  if (cfg.autoFallbackModel) {
31545
31565
  const familyModel = {
31546
31566
  google: "gemini-3.8-flash",
31547
- anthropic: "claude-sonnet-4-6",
31567
+ anthropic: "claude-sonnet-5-5",
31548
31568
  openai: "gpt-oss-120b-medium"
31549
31569
  };
31550
31570
  for (const fbFam of [
@@ -32062,6 +32082,7 @@ var AgyAdapter = class AgyAdapter extends LlmAdapter {
32062
32082
  args,
32063
32083
  cwd: workspaceRoot,
32064
32084
  timeoutMs: cfg.timeoutMs,
32085
+ maxTurnMs: cfg.maxTurnMinutes > 0 ? cfg.maxTurnMinutes * 6e4 : void 0,
32065
32086
  signal: options.signal,
32066
32087
  env,
32067
32088
  stdinPayload,
@@ -32146,6 +32167,11 @@ var AgyAdapter = class AgyAdapter extends LlmAdapter {
32146
32167
  code: "ABORTED",
32147
32168
  message: "agy run aborted by caller"
32148
32169
  };
32170
+ else if (outcome.capped) failure = {
32171
+ kind: "error",
32172
+ code: Err.TIMEOUT,
32173
+ message: "agy turn exceeded maxTurnMinutes (" + cfg.maxTurnMinutes + " min) — the agy process tree was terminated. Re-send to continue. 本轮超过 maxTurnMinutes 上限,已终止 agy 进程树"
32174
+ };
32149
32175
  else if (outcome.timedOut) failure = {
32150
32176
  kind: "error",
32151
32177
  code: Err.TIMEOUT,
@@ -32268,7 +32294,7 @@ var AgyAdapter = class AgyAdapter extends LlmAdapter {
32268
32294
  }
32269
32295
  }
32270
32296
  this.deps.onRun?.({
32271
- processOk: !outcome.aborted && !outcome.timedOut && outcome.code === 0,
32297
+ processOk: !outcome.aborted && !outcome.timedOut && !outcome.capped && outcome.code === 0,
32272
32298
  processCode: outcome.aborted ? "ABORTED" : outcome.timedOut ? Err.TIMEOUT : outcome.code === 0 ? "OK" : "EXIT_" + String(outcome.code),
32273
32299
  toolErrors: rec.toolErrors(),
32274
32300
  durationMs: outcome.durationMs,
@@ -35892,7 +35918,7 @@ function pickWarmupModelForFamily(family, monitoredModels = [], account) {
35892
35918
  const anyClaude = accountModels.find((m) => m.toLowerCase().includes("claude"));
35893
35919
  if (anyClaude) return anyClaude;
35894
35920
  }
35895
- return monitoredModels.find((m) => m.toLowerCase().includes("claude")) ?? "claude-sonnet-4-6";
35921
+ return monitoredModels.find((m) => m.toLowerCase().includes("claude")) ?? "claude-sonnet-5-5";
35896
35922
  }
35897
35923
  if (family === "openai") return monitoredModels.find((m) => m.toLowerCase().includes("gpt")) ?? "gpt-oss-120b-medium";
35898
35924
  return monitoredModels.find((m) => m.toLowerCase().includes("gemini") && !m.toLowerCase().includes("argon")) ?? "gemini-3.8-flash";
@@ -36214,7 +36240,7 @@ var WarmupService = class {
36214
36240
  skipped: true,
36215
36241
  skipReason: "No warmup models configured"
36216
36242
  }];
36217
- const monitored = configuredModels ?? ["gemini-3.8-flash", "claude-sonnet-4-6"];
36243
+ const monitored = configuredModels ?? ["gemini-3.8-flash", "claude-sonnet-5-5"];
36218
36244
  const quotas = account.quotas ?? {};
36219
36245
  const families = Object.keys(quotas);
36220
36246
  const modelsToPing = [];
@@ -32,6 +32,7 @@ Resolution order (highest first): `DSH_AGY_*` environment variable → runtime o
32
32
  | Key | Environment | Default | Description |
33
33
  | --- | --- | --- | --- |
34
34
  | `timeoutMs` | `DSH_AGY_TIMEOUT_MS` | `600000` | Activity watchdog timeout per turn |
35
+ | `maxTurnMinutes` | `DSH_AGY_MAX_TURN_MINUTES` | `0` | Hard wall-clock cap per turn (0 = unlimited); bounds a self-driving turn that never goes idle |
35
36
  | `longToolIdleMinutes` | `DSH_AGY_LONG_TOOL_IDLE_MINUTES` | `240` | Idle budget while a tool step is open; `0` keeps `timeoutMs` so wedged tools fail fast |
36
37
  | `dispatchTimeoutMinutes` | `—` | `15` | Cap on waiting for the Windows credential slot before backing off and retrying; `0` waits indefinitely |
37
38
  | `dispatchRetries` | `—` | `1` | Extra dispatch attempts after a credential-slot timeout |
@@ -81,7 +82,7 @@ Resolution order (highest first): `DSH_AGY_*` environment variable → runtime o
81
82
  | `quotaPollIntervalMs` | `DSH_AGY_QUOTA_POLL_INTERVAL_MS` | `3600000` | Background quota poll interval (minimum 60 s) |
82
83
  | `quotaViaCli` | `DSH_AGY_QUOTA_VIA_CLI` | `true` | Prefer the agy CLI path for quota queries |
83
84
  | `autoWarmup` | `DSH_AGY_AUTO_WARMUP` | `false` | Opt-in weekly-quota warmup |
84
- | `warmupModels` | `DSH_AGY_WARMUP_MODELS` | `gemini-3.8-flash,claude-sonnet-4-6` | Models used for warmup probes |
85
+ | `warmupModels` | `DSH_AGY_WARMUP_MODELS` | `gemini-3.8-flash,claude-sonnet-5-5` | Models used for warmup probes |
85
86
  | `suppressAgyAutoUpdate` | `DSH_AGY_SUPPRESS_AUTOUPDATE` | `true` on Windows, `false` elsewhere | Refresh agy's update-check timestamp before every spawn to stop the `--bg-updater` console flash chain |
86
87
 
87
88
  ## Translation (`/tsl`)
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "dsh-agy-link",
3
- "version": "0.6.12",
3
+ "version": "0.6.14",
4
4
  "description": "Google Antigravity (agy CLI) models for DeepSeek Harness — stream Gemini/Claude/GPT-OSS subscriptions into DSH with thinking, tool activity, token usage and in-GUI Google OAuth login.",
5
5
  "type": "module",
6
6
  "license": "MIT",