@kenz1117/dsh-engram 0.9.5 → 0.9.7

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/lib/index.js CHANGED
@@ -340,6 +340,9 @@ const zh$2 = {
340
340
  "misc.exportRecordLine": "- [{status}/{kind}] {content}(id={id},importance {importance})",
341
341
  "misc.exportEdgesHeading": "## 关系边",
342
342
  "misc.importInvalidPayload": "导入文件结构有误:{error}",
343
+ "misc.bodyTooLarge": "请求体过大:超过 {limit} 上限。",
344
+ "misc.bodyMalformed": "请求体不是合法 JSON:{error}",
345
+ "misc.bodyNotObject": "请求体必须是 JSON 对象。",
343
346
  "misc.importBodyTooLarge": "导入文件过大:超过 {limit} 上限。请拆分导出文件后重试,或改用 engram_import 工具按文件路径导入。",
344
347
  "misc.importBodyMalformed": "导入请求体不是合法 JSON:{error}",
345
348
  "misc.importBodyNotObject": "导入请求体必须是 JSON 对象。",
@@ -387,6 +390,9 @@ const en$2 = {
387
390
  "misc.exportRecordLine": "- [{status}/{kind}] {content} (id={id}, importance {importance})",
388
391
  "misc.exportEdgesHeading": "## Relation edges",
389
392
  "misc.importInvalidPayload": "import file structure is invalid: {error}",
393
+ "misc.bodyTooLarge": "request body is too large: over the {limit} limit.",
394
+ "misc.bodyMalformed": "request body is not valid JSON: {error}",
395
+ "misc.bodyNotObject": "request body must be a JSON object.",
390
396
  "misc.importBodyTooLarge": "import file is too large: over the {limit} limit. Split the export and retry, or use the engram_import tool with a file path.",
391
397
  "misc.importBodyMalformed": "import request body is not valid JSON: {error}",
392
398
  "misc.importBodyNotObject": "import request body must be a JSON object.",
@@ -434,6 +440,9 @@ const ru$2 = {
434
440
  "misc.exportRecordLine": "- [{status}/{kind}] {content} (id={id}, importance {importance})",
435
441
  "misc.exportEdgesHeading": "## Связи",
436
442
  "misc.importInvalidPayload": "структура файла импорта некорректна: {error}",
443
+ "misc.bodyTooLarge": "тело запроса слишком большое: превышен предел {limit}.",
444
+ "misc.bodyMalformed": "тело запроса — не корректный JSON: {error}",
445
+ "misc.bodyNotObject": "тело запроса должно быть JSON-объектом.",
437
446
  "misc.importBodyTooLarge": "файл импорта слишком большой: превышен предел {limit}. Разделите экспорт и повторите либо используйте инструмент engram_import с путём к файлу.",
438
447
  "misc.importBodyMalformed": "тело запроса импорта — не корректный JSON: {error}",
439
448
  "misc.importBodyNotObject": "тело запроса импорта должно быть JSON-объектом.",
@@ -515,6 +524,13 @@ const zh$1 = {
515
524
  "store.migrationAlreadyMigrated": "[dsh-engram] 旧库已有迁移记录({names}):{pointer};先前迁移的库可能含有当前目录的记忆。{manual}",
516
525
  "store.mergeTargetMissing": "write-disposition: 并入目标 {id} 不存在",
517
526
  "store.embedderDownloadFailed": "嵌入模型下载/加载失败;检索已降级为纯关键词模式,可检查网络后重试",
527
+ "store.embedderDimensionMismatch": "嵌入向量维度不一致:库里是 {expected} 维,嵌入器给出 {found} 维。比较与写入直接拒绝,而不是逐维截断——请把模型改回去,或用新模型重建库中向量",
528
+ "store.embedderRemoteUnreachable": "嵌入端点不可达({baseUrl},超时 {timeoutMs} 毫秒)",
529
+ "store.embedderRemoteHttp": "嵌入端点返回 HTTP {status}{detail}",
530
+ "store.embedderRemoteBody": "嵌入端点的响应不是可解析的 JSON",
531
+ "store.embedderRemoteCount": "嵌入端点返回 {received} 个向量,而本次请求为 {expected} 个",
532
+ "store.embedderRemoteCountNa": "非数组",
533
+ "store.embedderRemoteEmpty": "嵌入端点返回了空向量",
518
534
  "store.budgetDropped": "(另有 {count} 条未展示:缩小查询范围或降低 limit 后重试)"
519
535
  };
520
536
  /** Английский набор — образец полноты. */
@@ -548,6 +564,13 @@ const en$1 = {
548
564
  "store.migrationAlreadyMigrated": "[dsh-engram] the legacy database already has a migration record ({names}): {pointer}; the database migrated earlier may hold memories of the current directory. {manual}",
549
565
  "store.mergeTargetMissing": "write-disposition: merge target {id} does not exist",
550
566
  "store.embedderDownloadFailed": "embedding model download/load failed; retrieval has degraded to keyword-only mode — check the network and retry",
567
+ "store.embedderDimensionMismatch": "embedding dimension mismatch: the store holds {expected}-dimensional vectors, the embedder returned {found}. Comparison and writes are refused instead of being silently truncated — switch the model back, or re-embed the store with the new one",
568
+ "store.embedderRemoteUnreachable": "embedding endpoint unreachable ({baseUrl}, timeout {timeoutMs} ms)",
569
+ "store.embedderRemoteHttp": "embedding endpoint returned HTTP {status}{detail}",
570
+ "store.embedderRemoteBody": "embedding endpoint returned a body that is not parseable JSON",
571
+ "store.embedderRemoteCount": "embedding endpoint returned {received} vectors instead of the requested {expected}",
572
+ "store.embedderRemoteCountNa": "not an array",
573
+ "store.embedderRemoteEmpty": "embedding endpoint returned an empty vector",
551
574
  "store.budgetDropped": "({count} more not shown: narrow the query or lower the limit and retry)"
552
575
  };
553
576
  /** Русский набор. */
@@ -581,6 +604,13 @@ const ru$1 = {
581
604
  "store.migrationAlreadyMigrated": "[dsh-engram] У старой базы уже есть запись о переносе ({names}): {pointer}; в базе, перенесённой ранее, могут быть воспоминания текущего каталога. {manual}",
582
605
  "store.mergeTargetMissing": "write-disposition: цель слияния {id} не существует",
583
606
  "store.embedderDownloadFailed": "Не удалось скачать/загрузить модель эмбеддингов; поиск переведён в режим только по ключевым словам — проверьте сеть и повторите",
607
+ "store.embedderDimensionMismatch": "размерность векторов не совпадает: в базе {expected} измерений, эмбеддер вернул {found}. Сравнение и запись отвергаются, а не усекаются молча — верните прежнюю модель или пересчитайте векторы новой",
608
+ "store.embedderRemoteUnreachable": "эндпоинт эмбеддингов недоступен ({baseUrl}, таймаут {timeoutMs} мс)",
609
+ "store.embedderRemoteHttp": "эндпоинт эмбеддингов вернул HTTP {status}{detail}",
610
+ "store.embedderRemoteBody": "эндпоинт эмбеддингов вернул тело, которое не удалось разобрать как JSON",
611
+ "store.embedderRemoteCount": "эндпоинт эмбеддингов вернул {received} векторов вместо запрошенных {expected}",
612
+ "store.embedderRemoteCountNa": "не массив",
613
+ "store.embedderRemoteEmpty": "эндпоинт эмбеддингов вернул пустой вектор",
584
614
  "store.budgetDropped": "(ещё {count} не показано: сузьте запрос или уменьшите limit и повторите)"
585
615
  };
586
616
  //#endregion
@@ -1668,6 +1698,99 @@ function message(key, params) {
1668
1698
  return messageIn(current, key, params);
1669
1699
  }
1670
1700
  //#endregion
1701
+ //#region src/types.ts
1702
+ /** 从任意字符串铸造品牌化 id(存储层入库前调用)。 */
1703
+ function asMemoryId(raw) {
1704
+ return raw;
1705
+ }
1706
+ /** 从任意字符串铸造品牌化实体 id(存储层入库前调用)。 */
1707
+ function asEntityId(raw) {
1708
+ return raw;
1709
+ }
1710
+ /** 从任意字符串铸造品牌化事实 id(存储层入库前调用)。 */
1711
+ function asFactId(raw) {
1712
+ return raw;
1713
+ }
1714
+ /** 从任意字符串铸造品牌化信念 id(存储层入库前调用)。 */
1715
+ function asObservationId(raw) {
1716
+ return raw;
1717
+ }
1718
+ /** 归一化比较键:trim + 合并连续空白 + 小写(消解第一版只做精确归一匹配,不做相似合并)。 */
1719
+ function normalizeEntityName(raw) {
1720
+ return raw.trim().replace(/\s+/g, " ").toLowerCase();
1721
+ }
1722
+ /** episode 情景时间线的时间邻近扩展缺省窗口:锚点 createdAt ± 60 分钟。 */
1723
+ const EPISODE_PROXIMITY_MS_DEFAULT = 36e5;
1724
+ /** dsh-engram 统一错误:加载/使用期的可诊断失败都抛此类型。 */
1725
+ var EngramError = class extends Error {
1726
+ /** 机器可读原因码,如 EMBEDDER_DOWNLOAD_FAILED / SCHEMA_INCOMPATIBLE。 */
1727
+ code;
1728
+ constructor(code, message, options) {
1729
+ super(message, options);
1730
+ this.name = "EngramError";
1731
+ this.code = code;
1732
+ }
1733
+ };
1734
+ //#endregion
1735
+ //#region src/embedder/local.ts
1736
+ /**
1737
+ * transformers.js 本地嵌入实现:量化权重(q8),离线推理。模型首次使用需联网
1738
+ * 下载到 cacheDir;之后完全离线,数据不出机。
1739
+ *
1740
+ * 默认模型仍是 Xenova/bge-small-zh-v1.5(中文 bge 小模型,512 维),但模型选择是
1741
+ * 部署属性而不是代码属性——处理俄语文本要用多语言模型,因此模型标识与池化方式
1742
+ * 都来自配置。
1743
+ * @module @kenz1117/dsh-engram/embedder/local
1744
+ */
1745
+ /** 模型标识(bge 中文小模型,512 维输出);保留为默认值以兼容旧配置。 */
1746
+ const ENGRAM_MODEL_ID = "Xenova/bge-small-zh-v1.5";
1747
+ /** 进程内单例:重复调用 createLocalEmbedder 共享同一模型实例,避免重复加载权重。 */
1748
+ let cached;
1749
+ /**
1750
+ * 创建本地嵌入器。
1751
+ * @param cacheDir - 模型缓存目录;不存在会自动创建(0o700)。
1752
+ * @param remoteHost - 模型下载端点;默认 huggingface.co,网络受限环境可配镜像(如 https://hf-mirror.com)。
1753
+ * @param modelId - 模型标识;默认 ENGRAM_MODEL_ID。更换模型会让进程内单例失效并重新加载。
1754
+ * @param pooling - 池化方式;默认 mean。
1755
+ * @returns 就绪的 EngramEmbedder。
1756
+ * @throws EngramError(code=EMBEDDER_DOWNLOAD_FAILED) 模型下载/加载失败(调用方据此降级)。
1757
+ */
1758
+ async function createLocalEmbedder(cacheDir, remoteHost, modelId = ENGRAM_MODEL_ID, pooling = "mean") {
1759
+ if (cached?.dir === cacheDir && cached.model === modelId && cached.pooling === pooling) return cached.embedder;
1760
+ await mkdir(cacheDir, {
1761
+ recursive: true,
1762
+ mode: 448
1763
+ });
1764
+ try {
1765
+ const { pipeline, env } = await import("@huggingface/transformers");
1766
+ env.cacheDir = cacheDir;
1767
+ if (remoteHost !== void 0 && remoteHost !== "") env.remoteHost = remoteHost;
1768
+ const extractor = await pipeline("feature-extraction", modelId, { dtype: "q8" });
1769
+ const embedder = {
1770
+ model: modelId,
1771
+ async embed(texts) {
1772
+ if (texts.length === 0) return [];
1773
+ return (await extractor([...texts], {
1774
+ pooling,
1775
+ normalize: true
1776
+ })).tolist().map((list) => new Float32Array(list));
1777
+ },
1778
+ close: async () => {
1779
+ if (cached?.embedder === embedder) cached = void 0;
1780
+ }
1781
+ };
1782
+ cached = {
1783
+ dir: cacheDir,
1784
+ model: modelId,
1785
+ pooling,
1786
+ embedder
1787
+ };
1788
+ return embedder;
1789
+ } catch (error) {
1790
+ throw new EngramError("EMBEDDER_DOWNLOAD_FAILED", message("store.embedderDownloadFailed"), { cause: error });
1791
+ }
1792
+ }
1793
+ //#endregion
1671
1794
  //#region src/jev/stale.ts
1672
1795
  /**
1673
1796
  * Поля Jev прежнего формата: ключ и адрес когда-то хранились у самого плагина
@@ -1751,6 +1874,7 @@ const CONFIG_KEYS = /* @__PURE__ */ new Set([
1751
1874
  "historyBackfillIncludeSeeded",
1752
1875
  "historyBackfillIncludeNoCwd",
1753
1876
  "jev",
1877
+ "embedder",
1754
1878
  "language",
1755
1879
  "redactWords"
1756
1880
  ]);
@@ -1764,6 +1888,18 @@ const REFLECT_MODES = /* @__PURE__ */ new Set([
1764
1888
  "suggest",
1765
1889
  "auto"
1766
1890
  ]);
1891
+ /** `embedder` 块的键:未知键按错误处理,而不是静默忽略。 */
1892
+ const EMBEDDER_KEYS = /* @__PURE__ */ new Set([
1893
+ "kind",
1894
+ "localModel",
1895
+ "localPooling",
1896
+ "model",
1897
+ "baseUrl",
1898
+ "apiKey",
1899
+ "apiKeyEnv",
1900
+ "batchSize",
1901
+ "timeoutMs"
1902
+ ]);
1767
1903
  /** Schemastery 校验面(cordis.yml 读取时校验)。 */
1768
1904
  const Config = z.object({
1769
1905
  dbDir: z.string(),
@@ -1804,6 +1940,17 @@ const Config = z.object({
1804
1940
  deferAcceptBelow: z.number().min(0).max(.5),
1805
1941
  contradictMinProbability: z.number().min(.5).max(1)
1806
1942
  }),
1943
+ embedder: z.object({
1944
+ kind: z.string(),
1945
+ localModel: z.string(),
1946
+ localPooling: z.string(),
1947
+ model: z.string(),
1948
+ baseUrl: z.string(),
1949
+ apiKey: z.string(),
1950
+ apiKeyEnv: z.string(),
1951
+ batchSize: z.number().step(1).min(1).max(256),
1952
+ timeoutMs: z.number().step(1).min(1e3).max(6e5)
1953
+ }),
1807
1954
  language: z.string(),
1808
1955
  redactWords: z.array(z.string())
1809
1956
  });
@@ -1858,6 +2005,26 @@ function resolveConfig(config = {}) {
1858
2005
  if (jev.contradictMinProbability !== void 0 && (jev.contradictMinProbability < .5 || jev.contradictMinProbability > 1)) throw new Error("dsh-engram: jev.contradictMinProbability must be in [0.5, 1]");
1859
2006
  if ((jev.deferAcceptBelow ?? .15) >= (jev.deferMergeAbove ?? .85)) throw new Error("dsh-engram: jev.deferAcceptBelow must be less than jev.deferMergeAbove");
1860
2007
  }
2008
+ const embedderConfig = config.embedder;
2009
+ const embedderKind = embedderConfig?.kind ?? "local";
2010
+ if (embedderConfig !== void 0) {
2011
+ for (const key of Object.keys(embedderConfig)) if (!EMBEDDER_KEYS.has(key)) throw new Error(`dsh-engram: unknown embedder key "${key}"`);
2012
+ if (embedderConfig.kind !== void 0 && embedderConfig.kind !== "local" && embedderConfig.kind !== "remote") throw new Error("dsh-engram: embedder.kind must be one of local|remote");
2013
+ if (embedderConfig.localPooling !== void 0 && embedderConfig.localPooling !== "mean" && embedderConfig.localPooling !== "cls") throw new Error("dsh-engram: embedder.localPooling must be one of mean|cls");
2014
+ if (embedderConfig.batchSize !== void 0 && (!Number.isInteger(embedderConfig.batchSize) || embedderConfig.batchSize < 1 || embedderConfig.batchSize > 256)) throw new Error("dsh-engram: embedder.batchSize must be an integer in [1, 256]");
2015
+ if (embedderConfig.timeoutMs !== void 0 && (!Number.isInteger(embedderConfig.timeoutMs) || embedderConfig.timeoutMs < 1e3 || embedderConfig.timeoutMs > 6e5)) throw new Error("dsh-engram: embedder.timeoutMs must be an integer in [1000, 600000]");
2016
+ const baseUrl = embedderConfig.baseUrl;
2017
+ if (baseUrl !== void 0 && baseUrl !== "") try {
2018
+ const parsed = new URL(baseUrl);
2019
+ if (parsed.protocol !== "https:" && parsed.protocol !== "http:") throw new Error("unsupported protocol");
2020
+ } catch {
2021
+ throw new Error("dsh-engram: embedder.baseUrl must be an http(s) URL");
2022
+ }
2023
+ if (embedderKind === "remote") {
2024
+ if (embedderConfig.model === void 0 || embedderConfig.model === "") throw new Error("dsh-engram: embedder.kind=remote requires embedder.model");
2025
+ if (baseUrl === void 0 || baseUrl === "") throw new Error("dsh-engram: embedder.kind=remote requires embedder.baseUrl");
2026
+ }
2027
+ }
1861
2028
  const dbDir = config.dbDir ?? join(homedir(), ".dsh", "engram");
1862
2029
  return {
1863
2030
  dbDir,
@@ -1903,6 +2070,19 @@ function resolveConfig(config = {}) {
1903
2070
  deferMergeAbove: config.jev?.deferMergeAbove ?? .85,
1904
2071
  deferAcceptBelow: config.jev?.deferAcceptBelow ?? .15,
1905
2072
  contradictMinProbability: config.jev?.contradictMinProbability ?? .8
2073
+ },
2074
+ embedder: {
2075
+ kind: embedderKind,
2076
+ localModel: embedderConfig?.localModel ?? "Xenova/bge-small-zh-v1.5",
2077
+ localPooling: embedderConfig?.localPooling ?? "mean",
2078
+ remote: embedderKind === "remote" && embedderConfig?.model !== void 0 && embedderConfig.baseUrl !== void 0 ? {
2079
+ model: embedderConfig.model,
2080
+ baseUrl: embedderConfig.baseUrl,
2081
+ apiKey: embedderConfig.apiKey,
2082
+ apiKeyEnv: embedderConfig.apiKeyEnv,
2083
+ batchSize: embedderConfig.batchSize ?? 32,
2084
+ timeoutMs: embedderConfig.timeoutMs ?? 3e4
2085
+ } : void 0
1906
2086
  }
1907
2087
  };
1908
2088
  }
@@ -2103,89 +2283,146 @@ function resolveRoomInput(input) {
2103
2283
  }
2104
2284
  }
2105
2285
  //#endregion
2106
- //#region src/types.ts
2107
- /** 从任意字符串铸造品牌化 id(存储层入库前调用)。 */
2108
- function asMemoryId(raw) {
2109
- return raw;
2110
- }
2111
- /** 从任意字符串铸造品牌化实体 id(存储层入库前调用)。 */
2112
- function asEntityId(raw) {
2113
- return raw;
2114
- }
2115
- /** 从任意字符串铸造品牌化事实 id(存储层入库前调用)。 */
2116
- function asFactId(raw) {
2117
- return raw;
2118
- }
2119
- /** 从任意字符串铸造品牌化信念 id(存储层入库前调用)。 */
2120
- function asObservationId(raw) {
2121
- return raw;
2122
- }
2123
- /** 归一化比较键:trim + 合并连续空白 + 小写(消解第一版只做精确归一匹配,不做相似合并)。 */
2124
- function normalizeEntityName(raw) {
2125
- return raw.trim().replace(/\s+/g, " ").toLowerCase();
2126
- }
2127
- /** episode 情景时间线的时间邻近扩展缺省窗口:锚点 createdAt ± 60 分钟。 */
2128
- const EPISODE_PROXIMITY_MS_DEFAULT = 36e5;
2129
- /** dsh-engram 统一错误:加载/使用期的可诊断失败都抛此类型。 */
2130
- var EngramError = class extends Error {
2131
- /** 机器可读原因码,如 EMBEDDER_DOWNLOAD_FAILED / SCHEMA_INCOMPATIBLE。 */
2132
- code;
2133
- constructor(code, message, options) {
2134
- super(message, options);
2135
- this.name = "EngramError";
2136
- this.code = code;
2137
- }
2138
- };
2139
- //#endregion
2140
- //#region src/embedder/local.ts
2286
+ //#region src/embedder/remote.ts
2141
2287
  /**
2142
- * transformers.js 本地嵌入实现:Xenova/bge-small-zh-v1.5,量化权重(q8),离线推理。
2143
- * 模型首次使用需联网下载到 cacheDir;之后完全离线,数据不出机。
2144
- * @module @kenz1117/dsh-engram/embedder/local
2288
+ * OpenAI 兼容端点嵌入器:`POST {baseUrl}/embeddings`。
2289
+ *
2290
+ * 端点既可以是公开的嵌入服务,也可以是局域网内的推理服务器(LM Studio、Ollama、vLLM)。
2291
+ * 对插件而言区别只在 baseUrl 的取值:「本地/远程」的边界是部署决定,不是代码决定,
2292
+ * 因此实现只有一份。
2293
+ *
2294
+ * 密钥只从配置或环境变量读取:不进入日志、不进入错误文本、不进入诊断字段 model。
2295
+ * @module @kenz1117/dsh-engram/embedder/remote
2145
2296
  */
2146
- /** 模型标识(bge 中文小模型,512 维输出)。 */
2147
- const ENGRAM_MODEL_ID = "Xenova/bge-small-zh-v1.5";
2148
- /** 进程内单例:重复调用 createLocalEmbedder 共享同一模型实例,避免重复加载权重。 */
2149
- let cached;
2297
+ /** 单次请求的文本条数上限:过大的批通常被服务端整体拒绝。 */
2298
+ const DEFAULT_BATCH_SIZE = 32;
2299
+ /** 单次请求的超时毫秒数。 */
2300
+ const DEFAULT_TIMEOUT_MS = 3e4;
2150
2301
  /**
2151
- * 创建本地嵌入器。
2152
- * @param cacheDir - 模型缓存目录;不存在会自动创建(0o700)。
2153
- * @param remoteHost - 模型下载端点;默认 huggingface.co,网络受限环境可配镜像(如 https://hf-mirror.com)。
2302
+ * 加载期试探请求的正文:只用来验证端点可达且模型存在,不参与任何检索。
2303
+ * 用 ASCII 常量而不是 message():这段文字发往外部服务,不是给人看的界面文本。
2304
+ */
2305
+ const PROBE_TEXT = "dsh-engram: embedder probe";
2306
+ /**
2307
+ * 把向量归一到单位长度。
2308
+ *
2309
+ * 接口承诺归一化向量,这同时消除了余弦与点积的差别:范数非零时两者相等。
2310
+ * 零向量原样返回——调用方把它当作「没有信号」。
2311
+ */
2312
+ function toUnitVector(values) {
2313
+ const out = new Float32Array(values.length);
2314
+ let sumSquares = 0;
2315
+ for (let i = 0; i < values.length; i++) sumSquares += values[i] * values[i];
2316
+ if (sumSquares === 0) return out;
2317
+ const scale = 1 / Math.sqrt(sumSquares);
2318
+ for (let i = 0; i < values.length; i++) out[i] = values[i] * scale;
2319
+ return out;
2320
+ }
2321
+ /**
2322
+ * 创建远程嵌入器。
2323
+ *
2324
+ * 创建时先发一次试探请求:端点不可达、密钥错误、模型缺失都在加载期暴露,
2325
+ * 调用方据此把检索降级为纯关键词,而不是等到第一次写入才失败。
2326
+ *
2327
+ * @param options - 端点参数。
2154
2328
  * @returns 就绪的 EngramEmbedder。
2155
- * @throws EngramError(code=EMBEDDER_DOWNLOAD_FAILED) 模型下载/加载失败(调用方据此降级)。
2329
+ * @throws EngramError(code=EMBEDDER_REMOTE_FAILED) 试探请求失败。
2156
2330
  */
2157
- async function createLocalEmbedder(cacheDir, remoteHost) {
2158
- if (cached?.dir === cacheDir) return cached.embedder;
2159
- await mkdir(cacheDir, {
2160
- recursive: true,
2161
- mode: 448
2162
- });
2163
- try {
2164
- const { pipeline, env } = await import("@huggingface/transformers");
2165
- env.cacheDir = cacheDir;
2166
- if (remoteHost !== void 0 && remoteHost !== "") env.remoteHost = remoteHost;
2167
- const extractor = await pipeline("feature-extraction", ENGRAM_MODEL_ID, { dtype: "q8" });
2168
- const embedder = {
2169
- model: ENGRAM_MODEL_ID,
2170
- async embed(texts) {
2171
- if (texts.length === 0) return [];
2172
- return (await extractor([...texts], {
2173
- pooling: "mean",
2174
- normalize: true
2175
- })).tolist().map((list) => new Float32Array(list));
2176
- },
2177
- close: async () => {
2178
- if (cached?.embedder === embedder) cached = void 0;
2179
- }
2180
- };
2181
- cached = {
2182
- dir: cacheDir,
2183
- embedder
2331
+ async function createRemoteEmbedder(options) {
2332
+ const endpoint = `${options.baseUrl.replace(/\/+$/, "")}/embeddings`;
2333
+ const batchSize = options.batchSize ?? DEFAULT_BATCH_SIZE;
2334
+ const timeoutMs = options.timeoutMs ?? DEFAULT_TIMEOUT_MS;
2335
+ const headers = { "Content-Type": "application/json" };
2336
+ /**
2337
+ * 请求头在每次请求时组装,而不是在创建时确定:变量名由配置给出,值却可能
2338
+ * 在运行中变化(轮换密钥、注入到测试环境),换密钥不该要求重启插件。
2339
+ */
2340
+ const requestHeaders = () => {
2341
+ const fromEnv = options.apiKeyEnv === void 0 || options.apiKeyEnv === "" ? void 0 : process.env[options.apiKeyEnv];
2342
+ const apiKey = fromEnv !== void 0 && fromEnv !== "" ? fromEnv : options.apiKey;
2343
+ if (apiKey === void 0 || apiKey === "") return headers;
2344
+ return {
2345
+ ...headers,
2346
+ Authorization: `Bearer ${apiKey}`
2184
2347
  };
2185
- return embedder;
2186
- } catch (error) {
2187
- throw new EngramError("EMBEDDER_DOWNLOAD_FAILED", message("store.embedderDownloadFailed"), { cause: error });
2188
- }
2348
+ };
2349
+ /** 单次请求;服务端给了 index 就按 index 还原顺序。 */
2350
+ const request = async (texts) => {
2351
+ let response;
2352
+ try {
2353
+ response = await fetch(endpoint, {
2354
+ method: "POST",
2355
+ headers: requestHeaders(),
2356
+ body: JSON.stringify({
2357
+ model: options.model,
2358
+ input: [...texts],
2359
+ encoding_format: "float"
2360
+ }),
2361
+ signal: AbortSignal.timeout(timeoutMs)
2362
+ });
2363
+ } catch (error) {
2364
+ throw new EngramError("EMBEDDER_REMOTE_FAILED", message("store.embedderRemoteUnreachable", {
2365
+ baseUrl: options.baseUrl,
2366
+ timeoutMs
2367
+ }), { cause: error });
2368
+ }
2369
+ if (!response.ok) {
2370
+ const detail = await response.text().catch(() => "");
2371
+ throw new EngramError("EMBEDDER_REMOTE_FAILED", message("store.embedderRemoteHttp", {
2372
+ status: response.status,
2373
+ detail: detail === "" ? "" : `: ${detail.slice(0, 200)}`
2374
+ }));
2375
+ }
2376
+ let payload;
2377
+ try {
2378
+ payload = await response.json();
2379
+ } catch (error) {
2380
+ throw new EngramError("EMBEDDER_REMOTE_FAILED", message("store.embedderRemoteBody"), { cause: error });
2381
+ }
2382
+ const data = payload?.data;
2383
+ if (!Array.isArray(data) || data.length !== texts.length) throw new EngramError("EMBEDDER_REMOTE_FAILED", message("store.embedderRemoteCount", {
2384
+ received: Array.isArray(data) ? data.length : message("store.embedderRemoteCountNa"),
2385
+ expected: texts.length
2386
+ }));
2387
+ return (data.every((item) => typeof item.index === "number") ? [...data].sort((left, right) => (left.index ?? 0) - (right.index ?? 0)) : data).map((item) => {
2388
+ const values = item.embedding;
2389
+ if (!Array.isArray(values) || values.length === 0) throw new EngramError("EMBEDDER_REMOTE_FAILED", message("store.embedderRemoteEmpty"));
2390
+ return toUnitVector(values);
2391
+ });
2392
+ };
2393
+ const embed = async (texts) => {
2394
+ if (texts.length === 0) return [];
2395
+ const vectors = [];
2396
+ for (let start = 0; start < texts.length; start += batchSize) vectors.push(...await request(texts.slice(start, start + batchSize)));
2397
+ return vectors;
2398
+ };
2399
+ await embed([PROBE_TEXT]);
2400
+ return {
2401
+ model: options.model,
2402
+ embed,
2403
+ close: async () => {}
2404
+ };
2405
+ }
2406
+ //#endregion
2407
+ //#region src/embedder/select.ts
2408
+ /**
2409
+ * 按配置创建嵌入器。
2410
+ * @param config - 插件解析后的配置。
2411
+ * @returns 就绪的 EngramEmbedder。
2412
+ * @throws EngramError 本地模型加载失败,或远程端点没有响应。
2413
+ */
2414
+ function createEmbedder(config) {
2415
+ const embedder = config.embedder;
2416
+ const remote = embedder.remote;
2417
+ if (remote !== void 0) return createRemoteEmbedder({
2418
+ model: remote.model,
2419
+ baseUrl: remote.baseUrl,
2420
+ apiKey: remote.apiKey,
2421
+ apiKeyEnv: remote.apiKeyEnv,
2422
+ batchSize: remote.batchSize,
2423
+ timeoutMs: remote.timeoutMs
2424
+ });
2425
+ return createLocalEmbedder(config.modelCacheDir, config.hfEndpoint, embedder.localModel, embedder.localPooling);
2189
2426
  }
2190
2427
  //#endregion
2191
2428
  //#region src/llm/client.ts
@@ -3585,27 +3822,27 @@ async function runHistoryBackfill(deps, defaults, overrides, onProgress, signal)
3585
3822
  * 向量相似度共享实现:store 检索与巩固聚类共用一份,替代历史上的两处复制。
3586
3823
  * @module @kenz1117/dsh-engram/vec
3587
3824
  */
3588
- /** 维度不匹配告警只发一次(模块级标记);上游单一嵌入模型下维度恒等,永不触发。 */
3589
- let warnedDimensionMismatch = false;
3590
3825
  /**
3591
3826
  * 余弦相似度;任一向量零范数时返回 0。
3592
- * 嵌入器当前为单一硬编码模型,维度恒等;一旦嵌入器可配置且两向量维度不同,
3593
- * 逐维截断比较会产出「看起来像相似度的数」——这里响亮告警一次(不抛错,
3594
- * 检索照常降级),供换模型时第一时间发现存量向量与新模型维度不一致。
3827
+ *
3828
+ * 维度不同的两个向量不比较,而是抛错。逐维截断会得到一个「看起来像相似度」的数,
3829
+ * 而嵌入器现在是可配置的部署属性:换模型之后存量向量与新向量维度不同是常规情形,
3830
+ * 不是意外。与其静默降级、让人按错误的相似度做合并决定,不如在第一次比较时就说清
3831
+ * 两边的维度——错误原样交给上层,人就有机会改回模型,或用新模型重建向量。
3595
3832
  * @param a - 向量 A。
3596
3833
  * @param b - 向量 B。
3597
3834
  * @returns 余弦相似度,范围约 [-1, 1];零范数输入返回 0。
3835
+ * @throws EngramError(code=EMBEDDER_DIMENSION_MISMATCH) 两向量维度不同。
3598
3836
  */
3599
3837
  function cosine(a, b) {
3600
- if (a.length !== b.length && !warnedDimensionMismatch) {
3601
- warnedDimensionMismatch = true;
3602
- console.warn(`[dsh-engram] cosine dimension mismatch (${String(a.length)} vs ${String(b.length)}): mixed embedder models? retrieval quality is degraded`);
3603
- }
3838
+ if (a.length !== b.length) throw new EngramError("EMBEDDER_DIMENSION_MISMATCH", message("store.embedderDimensionMismatch", {
3839
+ expected: b.length,
3840
+ found: a.length
3841
+ }));
3604
3842
  let dot = 0;
3605
3843
  let na = 0;
3606
3844
  let nb = 0;
3607
- const len = Math.min(a.length, b.length);
3608
- for (let i = 0; i < len; i++) {
3845
+ for (let i = 0; i < a.length; i++) {
3609
3846
  dot += a[i] * b[i];
3610
3847
  na += a[i] * a[i];
3611
3848
  nb += b[i] * b[i];
@@ -3613,6 +3850,17 @@ function cosine(a, b) {
3613
3850
  if (na === 0 || nb === 0) return 0;
3614
3851
  return dot / (Math.sqrt(na) * Math.sqrt(nb));
3615
3852
  }
3853
+ /**
3854
+ * 是不是「维度不一致」这类错误。
3855
+ *
3856
+ * 这类错误在比较向量时产生,要由调用方区分:能承受的调用方(检索)降级继续,
3857
+ * 不能承受的(写入)照原样抛出。区分的依据是代码而不是文本——文本会随语言变。
3858
+ * @param error - 接住的值。
3859
+ * @returns true,若是带 EMBEDDER_DIMENSION_MISMATCH 代码的 EngramError。
3860
+ */
3861
+ function isDimensionMismatch(error) {
3862
+ return error instanceof EngramError && error.code === "EMBEDDER_DIMENSION_MISMATCH";
3863
+ }
3616
3864
  //#endregion
3617
3865
  //#region src/consolidation/run.ts
3618
3866
  /** 单条整理动作(写入 op_log 的统一标记)。 */
@@ -4348,17 +4596,6 @@ async function readJsonBody(req, limit = BODY_LIMIT_BYTES) {
4348
4596
  body: parsed
4349
4597
  };
4350
4598
  }
4351
- /**
4352
- * 读取 JSON body 并宽容降级:坏 body 视为空对象。
4353
- *
4354
- * 给「body 整体可选」的普通写接口用(补做、配置、复习等)——它们的 body 都是
4355
- * 小结构,坏 body 落到字段校验上照样 400,不需要区分原因。要区分原因的路由
4356
- * (导入)自己用 {@link readJsonBody}。
4357
- */
4358
- async function readJsonBodyOrEmpty(req, limit = BODY_LIMIT_BYTES) {
4359
- const result = await readJsonBody(req, limit);
4360
- return result.ok ? result.body : {};
4361
- }
4362
4599
  //#endregion
4363
4600
  //#region src/jev/client.ts
4364
4601
  /**
@@ -5643,6 +5880,29 @@ function json(res, status, body) {
5643
5880
  res.writeHead(status, { "content-type": "application/json; charset=utf-8" });
5644
5881
  res.end(JSON.stringify(body));
5645
5882
  }
5883
+ /**
5884
+ * 读 JSON body,并把失败原因说清楚:超限 413、坏 JSON 400、不是对象 400。
5885
+ *
5886
+ * 为什么不给「宽容降级成空对象」:那样超限/坏 body 会掉进字段校验,报成
5887
+ * 「id required」「archivePath required」——把服务端的锅甩给调用方(#34 的同款毛病)。
5888
+ * 机器已经能分辨三种原因,就别再合成一句话。
5889
+ *
5890
+ * @returns body;返回 null = 已写好错误响应,调用方直接 return。
5891
+ */
5892
+ async function readBody(req, res, language) {
5893
+ const result = await readJsonBody(req);
5894
+ if (result.ok) return result.body;
5895
+ if (result.reason === "too-large") {
5896
+ json(res, 413, { error: messageIn(language, "misc.bodyTooLarge", { limit: formatBytes(result.limit) }) });
5897
+ return null;
5898
+ }
5899
+ if (result.reason === "malformed") {
5900
+ json(res, 400, { error: messageIn(language, "misc.bodyMalformed", { error: result.error }) });
5901
+ return null;
5902
+ }
5903
+ json(res, 400, { error: messageIn(language, "misc.bodyNotObject") });
5904
+ return null;
5905
+ }
5646
5906
  /** 从 URL searchParams 收敛 scope(缺省 user)。 */
5647
5907
  function scopeOf$1(raw, fallback) {
5648
5908
  if (raw === "user" || raw === "project" || raw === "shared") return raw;
@@ -5987,7 +6247,8 @@ function registerEngramRoutes(ctx, deps) {
5987
6247
  }
5988
6248
  if (req.method === "POST" && route === "review-answer") {
5989
6249
  if (!guardWrite(req, res)) return;
5990
- const body = await readJsonBodyOrEmpty(req);
6250
+ const body = await readBody(req, res, deps.language);
6251
+ if (body === null) return;
5991
6252
  if (typeof body.id !== "string" || body.id === "") {
5992
6253
  json(res, 400, { error: "id required" });
5993
6254
  return;
@@ -6050,27 +6311,6 @@ function registerEngramRoutes(ctx, deps) {
6050
6311
  });
6051
6312
  return;
6052
6313
  }
6053
- if (req.method === "POST" && route === "review-answer") {
6054
- if (!guardWrite(req, res)) return;
6055
- const body = await readJsonBodyOrEmpty(req);
6056
- if (typeof body.id !== "string" || body.id === "") {
6057
- json(res, 400, { error: "id required" });
6058
- return;
6059
- }
6060
- const grade = Number(body.grade);
6061
- if (!Number.isInteger(grade) || grade < 0 || grade > 5) {
6062
- json(res, 400, { error: "grade must be an integer 0-5" });
6063
- return;
6064
- }
6065
- adoptBodySelector(body);
6066
- const record = await (await storeFor(scopeOf$1(typeof body.scope === "string" ? body.scope : null, "user"))).scheduleReview(body.id, grade);
6067
- if (record === void 0) {
6068
- json(res, 404, { error: messageIn(deps.language, "misc.recordNotFound", { id: body.id }) });
6069
- return;
6070
- }
6071
- json(res, 200, { record });
6072
- return;
6073
- }
6074
6314
  if (req.method === "GET" && route === "export") {
6075
6315
  const scope = scopeOf$1(url.searchParams.get("scope"), "user");
6076
6316
  const format = url.searchParams.get("format") === "json" ? "json" : "markdown";
@@ -6345,7 +6585,8 @@ function registerEngramRoutes(ctx, deps) {
6345
6585
  }
6346
6586
  if (req.method === "POST" && route === "history-backfill/start") {
6347
6587
  if (!guardWrite(req, res)) return;
6348
- const body = await readJsonBodyOrEmpty(req);
6588
+ const body = await readBody(req, res, deps.language);
6589
+ if (body === null) return;
6349
6590
  const started = deps.history.start(historyRulesOf(body));
6350
6591
  if (!started.ok) {
6351
6592
  json(res, 409, {
@@ -6379,7 +6620,8 @@ function registerEngramRoutes(ctx, deps) {
6379
6620
  }
6380
6621
  if (req.method === "POST" && route === "jev-config") {
6381
6622
  if (!guardWrite(req, res)) return;
6382
- const body = await readJsonBodyOrEmpty(req);
6623
+ const body = await readBody(req, res, deps.language);
6624
+ if (body === null) return;
6383
6625
  const parsed = parseJevPatch(loadJevOverride(deps.jev.dbDir), body);
6384
6626
  if (typeof parsed === "string") {
6385
6627
  json(res, 400, { error: parsed });
@@ -6398,7 +6640,8 @@ function registerEngramRoutes(ctx, deps) {
6398
6640
  }
6399
6641
  if (req.method === "POST" && route === "jev-test") {
6400
6642
  if (!guardWrite(req, res)) return;
6401
- const body = await readJsonBodyOrEmpty(req);
6643
+ const body = await readBody(req, res, deps.language);
6644
+ if (body === null) return;
6402
6645
  const base = effectiveJevConfig(deps.jev.base, loadJevOverride(deps.jev.dbDir));
6403
6646
  const provider = typeof body.provider === "string" && body.provider.trim() !== "" ? body.provider.trim() : base.provider;
6404
6647
  const model = typeof body.model === "string" && body.model.trim() !== "" ? body.model.trim() : base.model;
@@ -6427,7 +6670,8 @@ function registerEngramRoutes(ctx, deps) {
6427
6670
  }
6428
6671
  if (req.method === "POST" && route === "jev-models") {
6429
6672
  if (!guardWrite(req, res)) return;
6430
- const body = await readJsonBodyOrEmpty(req);
6673
+ const body = await readBody(req, res, deps.language);
6674
+ if (body === null) return;
6431
6675
  const provider = typeof body.provider === "string" ? body.provider.trim() : "";
6432
6676
  if (provider === "") {
6433
6677
  json(res, 200, {
@@ -6446,7 +6690,8 @@ function registerEngramRoutes(ctx, deps) {
6446
6690
  }
6447
6691
  if (req.method === "POST" && route === "consolidate") {
6448
6692
  if (!guardWrite(req, res)) return;
6449
- const body = await readJsonBodyOrEmpty(req);
6693
+ const body = await readBody(req, res, deps.language);
6694
+ if (body === null) return;
6450
6695
  adoptBodySelector(body);
6451
6696
  const rawScope = typeof body.scope === "string" ? body.scope : "user";
6452
6697
  const scope = rawScope === "project" ? "project" : rawScope === "shared" ? "shared" : "user";
@@ -6470,7 +6715,8 @@ function registerEngramRoutes(ctx, deps) {
6470
6715
  }
6471
6716
  if (req.method === "POST" && route === "restore-backup") {
6472
6717
  if (!guardWrite(req, res)) return;
6473
- const body = await readJsonBodyOrEmpty(req);
6718
+ const body = await readBody(req, res, deps.language);
6719
+ if (body === null) return;
6474
6720
  if (typeof body.archivePath !== "string" || body.archivePath === "") {
6475
6721
  json(res, 400, { error: "archivePath required" });
6476
6722
  return;
@@ -6496,7 +6742,8 @@ function registerEngramRoutes(ctx, deps) {
6496
6742
  }
6497
6743
  if (req.method === "POST" && (route === "update" || route === "forget" || route === "restore")) {
6498
6744
  if (!guardWrite(req, res)) return;
6499
- const body = await readJsonBodyOrEmpty(req);
6745
+ const body = await readBody(req, res, deps.language);
6746
+ if (body === null) return;
6500
6747
  if (typeof body.id !== "string" || body.id === "") {
6501
6748
  json(res, 400, { error: "id required" });
6502
6749
  return;
@@ -7364,6 +7611,34 @@ async function openEngramStore(path, rankBoost = NO_BOOST, automation = DEFAULT_
7364
7611
  });
7365
7612
  }
7366
7613
  }
7614
+ /**
7615
+ * 库中已有向量的维度(按浮点数个数计,不按字节)。
7616
+ *
7617
+ * 打开时读一次,且两张带向量的表一起读:记忆与信念彼此比较,任何一边对不上
7618
+ * 都会让语义检索失真。库里还没有向量时维度未知——由第一次写入定下来,在那之前
7619
+ * 没有可拒绝的东西。
7620
+ */
7621
+ const storedDimRow = db.prepare(`SELECT LENGTH(embedding) AS bytes FROM nodes WHERE embedding IS NOT NULL
7622
+ UNION ALL SELECT LENGTH(embedding) AS bytes FROM observations WHERE embedding IS NOT NULL LIMIT 1`).get();
7623
+ let storedDim = storedDimRow === void 0 ? void 0 : storedDimRow.bytes / Float32Array.BYTES_PER_ELEMENT;
7624
+ /** 降级为关键词检索的告警只打一次:同一个进程里重复刷同一句没有信息量。 */
7625
+ let warnedDegradedRead = false;
7626
+ /**
7627
+ * 放行向量入库,或拒绝并说出两边的维度。
7628
+ *
7629
+ * 严格而不是「能比多少算多少」:截断后的相似度仍是个像样的数字,合并决策却会
7630
+ * 据此做出。空库由第一次写入定维;此后必须一致,换模型就要重建向量。
7631
+ * @throws EngramError(code=EMBEDDER_DIMENSION_MISMATCH) 维度与库中不符。
7632
+ */
7633
+ const guardDimension = (vector) => {
7634
+ const length = vector instanceof Float32Array ? vector.length : vector.byteLength / Float32Array.BYTES_PER_ELEMENT;
7635
+ const expected = storedDim ?? length;
7636
+ if (length !== expected || !Number.isInteger(length)) throw new EngramError("EMBEDDER_DIMENSION_MISMATCH", message("store.embedderDimensionMismatch", {
7637
+ expected,
7638
+ found: length
7639
+ }));
7640
+ storedDim = length;
7641
+ };
7367
7642
  const sqlGet = db.prepare("SELECT * FROM nodes WHERE id = ?");
7368
7643
  const sqlInsert = db.prepare(`INSERT INTO nodes
7369
7644
  (id, scope, kind, content, importance, confidence, status, created_at, last_accessed_at, access_count,
@@ -7507,6 +7782,7 @@ async function openEngramStore(path, rankBoost = NO_BOOST, automation = DEFAULT_
7507
7782
  * 事务由调用方持有(withTransaction)。
7508
7783
  */
7509
7784
  const insertRecord = (id, input, content, importance, confidence, at, sourceSessionId, embedding, imagery, op) => {
7785
+ if (embedding !== null) guardDimension(embedding);
7510
7786
  const stored = embedding === null ? null : embedding instanceof Float32Array ? vecToBlob(embedding) : embedding;
7511
7787
  const initialReview = input.initialReviewAt ?? null;
7512
7788
  sqlInsert.run(id, input.scope, input.kind, content, importance, confidence, at, at, sourceSessionId, input.sourceRound ?? null, input.sourceSeq ?? null, stored, imageryToJson(imagery), input.slot?.room ?? null, input.slot?.index ?? null, input.imageryScore ?? null, initialReview, initialReview === null ? null : 2.5, initialReview === null ? null : 0);
@@ -7660,8 +7936,7 @@ async function openEngramStore(path, rankBoost = NO_BOOST, automation = DEFAULT_
7660
7936
  });
7661
7937
  });
7662
7938
  let degraded = true;
7663
- if (queryVector !== void 0) {
7664
- degraded = false;
7939
+ if (queryVector !== void 0) try {
7665
7940
  vectorPool(query.scopes, query.rooms).map((row) => ({
7666
7941
  row,
7667
7942
  sim: cosine(queryVector, blobToVec(row.embedding))
@@ -7677,6 +7952,16 @@ async function openEngramStore(path, rankBoost = NO_BOOST, automation = DEFAULT_
7677
7952
  via: "both"
7678
7953
  });
7679
7954
  });
7955
+ degraded = false;
7956
+ } catch (error) {
7957
+ if (!isDimensionMismatch(error)) throw error;
7958
+ if (!warnedDegradedRead) {
7959
+ warnedDegradedRead = true;
7960
+ console.warn(message("store.embedderDimensionMismatch", {
7961
+ expected: storedDim ?? queryVector.length,
7962
+ found: queryVector.length
7963
+ }));
7964
+ }
7680
7965
  }
7681
7966
  if (rankBoost.recencyWeight !== 0 || rankBoost.proofWeight !== 0) {
7682
7967
  const now = Date.now();
@@ -8441,6 +8726,7 @@ async function openEngramStore(path, rankBoost = NO_BOOST, automation = DEFAULT_
8441
8726
  const now = Date.now();
8442
8727
  const sourceIds = [...new Set(input.sourceIds.map(String))].map(asMemoryId);
8443
8728
  if (sourceIds.length === 0) throw new EngramError("BAD_INPUT", message("store.observationEvidenceRequired"));
8729
+ if (input.embedding !== void 0) guardDimension(input.embedding);
8444
8730
  const stored = input.embedding === void 0 ? null : vecToBlob(input.embedding);
8445
8731
  withTransaction(() => {
8446
8732
  sqlObservationInsert.run(id, input.scope, input.belief, JSON.stringify(sourceIds.map(String)), sourceIds.length, stored, now, now, now);
@@ -8462,6 +8748,7 @@ async function openEngramStore(path, rankBoost = NO_BOOST, automation = DEFAULT_
8462
8748
  if (existing === void 0) return void 0;
8463
8749
  const merged = [.../* @__PURE__ */ new Set([...observationRowToRecord(existing).sourceIds.map(String), ...addSourceIds.map(String)])].map(asMemoryId);
8464
8750
  const now = Date.now();
8751
+ if (embedding !== void 0) guardDimension(embedding);
8465
8752
  const stored = embedding === void 0 ? null : vecToBlob(embedding);
8466
8753
  withTransaction(() => {
8467
8754
  sqlObservationUpdate.run(belief, "active", JSON.stringify(merged.map(String)), merged.length, stored, now, now, String(id));
@@ -11826,7 +12113,7 @@ function estimateTokens(text) {
11826
12113
  }
11827
12114
  //#endregion
11828
12115
  //#region package.json
11829
- var version = "0.9.5";
12116
+ var version = "0.9.7";
11830
12117
  //#endregion
11831
12118
  //#region src/index.ts
11832
12119
  /**
@@ -12275,7 +12562,7 @@ function apply(ctx, config = {}) {
12275
12562
  * 面板「项目」scope、工具读写与自动摄取的 project 落点因此同一口径。
12276
12563
  */
12277
12564
  const resolveProjectStore = (cwd) => cwd === void 0 ? openStore("project") : openStoreForProjectCwd(cwd);
12278
- const embedder = createLocalEmbedder(resolved.modelCacheDir, resolved.hfEndpoint).catch((error) => {
12565
+ const embedder = createEmbedder(resolved).catch((error) => {
12279
12566
  console.warn(message("core.warnEmbedderUnavailable"), error);
12280
12567
  });
12281
12568
  const logIngestRequest = (data) => {