luckytri 1.0.6 → 1.0.7

package/server/knowledge/embedding-profile.js

@@ -1,114 +1,114 @@
-import { createHash } from "node:crypto";
-
-const CONFIG = "embedding-profile";
-export const EMBEDDING_DEFAULTS = {
-  enabled: false,
-  label: "知识向量",
-  baseUrl: "",
-  model: "",
-  apiKey: "",
-  dimensions: 0,
-  batchSize: 32,
-  timeoutMs: 90000,
-};
-
-export function embeddingSignature(profile) {
-  return createHash("sha256")
-    .update(
-      JSON.stringify([
-        profile.baseUrl || "",
-        profile.embeddingModel || profile.model || "",
-        profile.dimensions || 0,
-      ]),
-    )
-    .digest("hex");
-}
-
-export function publicEmbeddingProfile(repo) {
-  const saved = repo.config(CONFIG, null);
-  const { apiKey, ...profile } = { ...EMBEDDING_DEFAULTS, ...saved };
-  return {
-    ...profile,
-    configured: !!saved,
-    hasApiKey: !!(apiKey || process.env.EMBEDDING_API_KEY),
-  };
-}
-
-export function saveEmbeddingProfile(repo, input) {
-  const old = { ...EMBEDDING_DEFAULTS, ...repo.config(CONFIG, {}) };
-  const next = Object.fromEntries(
-    Object.keys(EMBEDDING_DEFAULTS).map((key) => [key, input[key] ?? old[key]]),
-  );
-  if (typeof next.enabled !== "boolean") throw Error("向量开关无效");
-  if (typeof next.apiKey !== "string") throw Error("密钥格式无效");
-  if (!next.apiKey) next.apiKey = old.apiKey;
-  next.label = String(next.label).trim().slice(0, 100);
-  next.model = String(next.model).trim().slice(0, 200);
-  next.baseUrl = String(next.baseUrl).trim().replace(/\/$/, "");
-  if (next.baseUrl) {
-    const url = new URL(next.baseUrl);
-    if (
-      !["http:", "https:"].includes(url.protocol) ||
-      url.username ||
-      url.password ||
-      url.search ||
-      url.hash
-    )
-      throw Error("向量 API 地址无效");
-  }
-  if (next.enabled && (!next.baseUrl || !next.model))
-    throw Error("启用向量需要独立地址和模型名");
-  for (const [key, min, max] of [
-    ["dimensions", 0, 65536],
-    ["batchSize", 1, 128],
-    ["timeoutMs", 1000, 300000],
-  ])
-    if (!Number.isSafeInteger(next[key]) || next[key] < min || next[key] > max)
-      throw Error(`${key} 超出范围`);
-  repo.saveConfig(CONFIG, next);
-  return publicEmbeddingProfile(repo);
-}
-
-export function resolveEmbeddingProfile(repo, models) {
-  const saved = repo.config(CONFIG, null);
-  if (saved) {
-    const profile = {
-      ...EMBEDDING_DEFAULTS,
-      ...saved,
-      id: "embedding",
-      embedding: saved.enabled,
-      embeddingModel: saved.model,
-      isolatedEmbeddingKey: true,
-    };
-    return { profile, signature: embeddingSignature(profile), legacy: false };
-  }
-  // Existing installations remain compatible until an independent profile is saved.
-  const profile = models.profile("default");
-  return { profile, signature: embeddingSignature(profile), legacy: true };
-}
-
-export async function embedBatches(models, profile, texts) {
-  if (!profile.embedding) return [];
-  const vectors = [];
-  const size = profile.batchSize || 32;
-  let dimension = profile.dimensions || 0;
-  for (let start = 0; start < texts.length; start += size) {
-    const batch = texts.slice(start, start + size);
-    const result = await models.embed(profile, batch);
-    if (!Array.isArray(result) || result.length !== batch.length)
-      throw Error("向量数量与输入不一致");
-    for (const vector of result) {
-      if (
-        !Array.isArray(vector) ||
-        !vector.length ||
-        !vector.every(Number.isFinite) ||
-        !vector.some((value) => value !== 0)
-      )
-        throw Error("向量内容无效");
-      dimension ||= vector.length;
-      if (vector.length !== dimension) throw Error("向量维度不一致");
-      vectors.push(vector);
-    }
-  }
-  return vectors;
-}
+import { createHash } from "node:crypto";
+
+const CONFIG = "embedding-profile";
+export const EMBEDDING_DEFAULTS = {
+  enabled: false,
+  label: "知识向量",
+  baseUrl: "",
+  model: "",
+  apiKey: "",
+  dimensions: 0,
+  batchSize: 32,
+  timeoutMs: 90000,
+};
+
+export function embeddingSignature(profile) {
+  return createHash("sha256")
+    .update(
+      JSON.stringify([
+        profile.baseUrl || "",
+        profile.embeddingModel || profile.model || "",
+        profile.dimensions || 0,
+      ]),
+    )
+    .digest("hex");
+}
+
+export function publicEmbeddingProfile(repo) {
+  const saved = repo.config(CONFIG, null);
+  const { apiKey, ...profile } = { ...EMBEDDING_DEFAULTS, ...saved };
+  return {
+    ...profile,
+    configured: !!saved,
+    hasApiKey: !!(apiKey || process.env.EMBEDDING_API_KEY),
+  };
+}
+
+export function saveEmbeddingProfile(repo, input) {
+  const old = { ...EMBEDDING_DEFAULTS, ...repo.config(CONFIG, {}) };
+  const next = Object.fromEntries(
+    Object.keys(EMBEDDING_DEFAULTS).map((key) => [key, input[key] ?? old[key]]),
+  );
+  if (typeof next.enabled !== "boolean") throw Error("向量开关无效");
+  if (typeof next.apiKey !== "string") throw Error("密钥格式无效");
+  if (!next.apiKey) next.apiKey = old.apiKey;
+  next.label = String(next.label).trim().slice(0, 100);
+  next.model = String(next.model).trim().slice(0, 200);
+  next.baseUrl = String(next.baseUrl).trim().replace(/\/$/, "");
+  if (next.baseUrl) {
+    const url = new URL(next.baseUrl);
+    if (
+      !["http:", "https:"].includes(url.protocol) ||
+      url.username ||
+      url.password ||
+      url.search ||
+      url.hash
+    )
+      throw Error("向量 API 地址无效");
+  }
+  if (next.enabled && (!next.baseUrl || !next.model))
+    throw Error("启用向量需要独立地址和模型名");
+  for (const [key, min, max] of [
+    ["dimensions", 0, 65536],
+    ["batchSize", 1, 128],
+    ["timeoutMs", 1000, 300000],
+  ])
+    if (!Number.isSafeInteger(next[key]) || next[key] < min || next[key] > max)
+      throw Error(`${key} 超出范围`);
+  repo.saveConfig(CONFIG, next);
+  return publicEmbeddingProfile(repo);
+}
+
+export function resolveEmbeddingProfile(repo, models) {
+  const saved = repo.config(CONFIG, null);
+  if (saved) {
+    const profile = {
+      ...EMBEDDING_DEFAULTS,
+      ...saved,
+      id: "embedding",
+      embedding: saved.enabled,
+      embeddingModel: saved.model,
+      isolatedEmbeddingKey: true,
+    };
+    return { profile, signature: embeddingSignature(profile), legacy: false };
+  }
+  // Existing installations remain compatible until an independent profile is saved.
+  const profile = models.profile("default");
+  return { profile, signature: embeddingSignature(profile), legacy: true };
+}
+
+export async function embedBatches(models, profile, texts) {
+  if (!profile.embedding) return [];
+  const vectors = [];
+  const size = profile.batchSize || 32;
+  let dimension = profile.dimensions || 0;
+  for (let start = 0; start < texts.length; start += size) {
+    const batch = texts.slice(start, start + size);
+    const result = await models.embed(profile, batch);
+    if (!Array.isArray(result) || result.length !== batch.length)
+      throw Error("向量数量与输入不一致");
+    for (const vector of result) {
+      if (
+        !Array.isArray(vector) ||
+        !vector.length ||
+        !vector.every(Number.isFinite) ||
+        !vector.some((value) => value !== 0)
+      )
+        throw Error("向量内容无效");
+      dimension ||= vector.length;
+      if (vector.length !== dimension) throw Error("向量维度不一致");
+      vectors.push(vector);
+    }
+  }
+  return vectors;
+}

package/server/knowledge/index.js

@@ -1,13 +1,13 @@
-export { KnowledgeManager } from "./manager.js";
-export { indexChunk, indexMemory, migrateKnowledge } from "./schema.js";
-export {
-  asPlainDocument,
-  chunkText,
-  cosine,
-  ftsMatchQuery,
-  ftsTokens,
-  lexicalTerms,
-  overlapScore,
-  packVector,
-  unpackVector,
-} from "./retrieval.js";
+export { KnowledgeManager } from "./manager.js";
+export { indexChunk, indexMemory, migrateKnowledge } from "./schema.js";
+export {
+  asPlainDocument,
+  chunkText,
+  cosine,
+  ftsMatchQuery,
+  ftsTokens,
+  lexicalTerms,
+  overlapScore,
+  packVector,
+  unpackVector,
+} from "./retrieval.js";

package/server/knowledge/retrieval.js

@@ -1,93 +1,93 @@
-export function lexicalTerms(s) {
-  return new Set(
-    String(s)
-      .toLowerCase()
-      .match(/[a-z0-9]+|[\u4e00-\u9fff]{1,2}/g) || [],
-  );
-}
-
-export function ftsTokens(s) {
-  return [...lexicalTerms(s)].join(" ");
-}
-
-export function ftsMatchQuery(s) {
-  return [...lexicalTerms(s)]
-    .slice(0, 24)
-    .map((t) => `"${t.replace(/"/g, "")}"`)
-    .filter((t) => t.length > 2)
-    .join(" OR ");
-}
-
-export function overlapScore(text, queryTerms) {
-  if (!queryTerms?.size) return 0;
-  return [...lexicalTerms(text)].filter((t) => queryTerms.has(t)).length;
-}
-
-export function cosine(a, b) {
-  if (!a?.length || !b?.length || a.length !== b.length) return 0;
-  let dot = 0,
-    na = 0,
-    nb = 0;
-  for (let i = 0; i < a.length; i++) {
-    dot += a[i] * b[i];
-    na += a[i] * a[i];
-    nb += b[i] * b[i];
-  }
-  return dot / (Math.sqrt(na) * Math.sqrt(nb) || 1);
-}
-
-export function packVector(values) {
-  return Buffer.from(new Float32Array(values).buffer);
-}
-
-export function unpackVector(blob) {
-  if (!blob) return null;
-  const buf = Buffer.isBuffer(blob) ? blob : Buffer.from(blob);
-  return Array.from(
-    new Float32Array(buf.buffer, buf.byteOffset, buf.byteLength / 4),
-  );
-}
-
-export function asPlainDocument(text) {
-  let source = String(text || "").replace(/\r\n/g, "\n");
-  if (/<\/?[a-z][\s\S]*>/i.test(source))
-    source = source
-      .replace(/<script[\s\S]*?<\/script>/gi, "\n")
-      .replace(/<style[\s\S]*?<\/style>/gi, "\n")
-      .replace(/<h([1-3])[^>]*>([\s\S]*?)<\/h\1>/gi, (_, level, title) => {
-        const hashes = "#".repeat(Number(level) || 1);
-        return `\n${hashes} ${title.replace(/<[^>]+>/g, "").trim()}\n`;
-      })
-      .replace(/<[^>]+>/g, " ");
-  return source
-    .replace(/[ \t]+\n/g, "\n")
-    .replace(/\n{3,}/g, "\n\n")
-    .trim();
-}
-
-export function chunkText(text, { size = 600, overlap = 80 } = {}) {
-  const source = String(text || "")
-    .replace(/\r\n/g, "\n")
-    .trim();
-  if (!source) return [];
-  const parts = source.split(/(?=^#{1,3} )/m);
-  const chunks = [];
-  for (const part of parts) {
-    const heading = part.match(/^#{1,3} ([^\n]+)/)?.[1] || "";
-    const chars = Array.from(part.trim());
-    if (!chars.length) continue;
-    if (chars.length <= size) {
-      chunks.push({ heading, text: part.trim() });
-      continue;
-    }
-    for (let i = 0; i < chars.length; i += Math.max(1, size - overlap)) {
-      const slice = chars
-        .slice(i, i + size)
-        .join("")
-        .trim();
-      if (slice) chunks.push({ heading, text: slice });
-      if (i + size >= chars.length) break;
-    }
-  }
-  return chunks;
-}
+export function lexicalTerms(s) {
+  return new Set(
+    String(s)
+      .toLowerCase()
+      .match(/[a-z0-9]+|[\u4e00-\u9fff]{1,2}/g) || [],
+  );
+}
+
+export function ftsTokens(s) {
+  return [...lexicalTerms(s)].join(" ");
+}
+
+export function ftsMatchQuery(s) {
+  return [...lexicalTerms(s)]
+    .slice(0, 24)
+    .map((t) => `"${t.replace(/"/g, "")}"`)
+    .filter((t) => t.length > 2)
+    .join(" OR ");
+}
+
+export function overlapScore(text, queryTerms) {
+  if (!queryTerms?.size) return 0;
+  return [...lexicalTerms(text)].filter((t) => queryTerms.has(t)).length;
+}
+
+export function cosine(a, b) {
+  if (!a?.length || !b?.length || a.length !== b.length) return 0;
+  let dot = 0,
+    na = 0,
+    nb = 0;
+  for (let i = 0; i < a.length; i++) {
+    dot += a[i] * b[i];
+    na += a[i] * a[i];
+    nb += b[i] * b[i];
+  }
+  return dot / (Math.sqrt(na) * Math.sqrt(nb) || 1);
+}
+
+export function packVector(values) {
+  return Buffer.from(new Float32Array(values).buffer);
+}
+
+export function unpackVector(blob) {
+  if (!blob) return null;
+  const buf = Buffer.isBuffer(blob) ? blob : Buffer.from(blob);
+  return Array.from(
+    new Float32Array(buf.buffer, buf.byteOffset, buf.byteLength / 4),
+  );
+}
+
+export function asPlainDocument(text) {
+  let source = String(text || "").replace(/\r\n/g, "\n");
+  if (/<\/?[a-z][\s\S]*>/i.test(source))
+    source = source
+      .replace(/<script[\s\S]*?<\/script>/gi, "\n")
+      .replace(/<style[\s\S]*?<\/style>/gi, "\n")
+      .replace(/<h([1-3])[^>]*>([\s\S]*?)<\/h\1>/gi, (_, level, title) => {
+        const hashes = "#".repeat(Number(level) || 1);
+        return `\n${hashes} ${title.replace(/<[^>]+>/g, "").trim()}\n`;
+      })
+      .replace(/<[^>]+>/g, " ");
+  return source
+    .replace(/[ \t]+\n/g, "\n")
+    .replace(/\n{3,}/g, "\n\n")
+    .trim();
+}
+
+export function chunkText(text, { size = 600, overlap = 80 } = {}) {
+  const source = String(text || "")
+    .replace(/\r\n/g, "\n")
+    .trim();
+  if (!source) return [];
+  const parts = source.split(/(?=^#{1,3} )/m);
+  const chunks = [];
+  for (const part of parts) {
+    const heading = part.match(/^#{1,3} ([^\n]+)/)?.[1] || "";
+    const chars = Array.from(part.trim());
+    if (!chars.length) continue;
+    if (chars.length <= size) {
+      chunks.push({ heading, text: part.trim() });
+      continue;
+    }
+    for (let i = 0; i < chars.length; i += Math.max(1, size - overlap)) {
+      const slice = chars
+        .slice(i, i + size)
+        .join("")
+        .trim();
+      if (slice) chunks.push({ heading, text: slice });
+      if (i + size >= chars.length) break;
+    }
+  }
+  return chunks;
+}