@opengeni/documents 0.2.5 → 0.2.8

package/LICENSE

@@ -0,0 +1,190 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   Copyright 2026 Cloudgeni-ai
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

package/dist/index.d.ts

@@ -1,5 +1,5 @@
 import { Settings } from '@opengeni/config';
-import { FileAsset, AddDocumentRequest, Document, CreateDocumentBaseRequest, DocumentBase, DocumentSearchResult } from '@opengeni/contracts';
+import { FileAsset, DocumentSearchMode, KnowledgeSourceKind, AddDocumentRequest, Document, CreateDocumentBaseRequest, DocumentBase, DocumentSearchResult } from '@opengeni/contracts';
 import { Database } from '@opengeni/db';
 import { ObjectStorage } from '@opengeni/storage';
 
@@ -34,6 +34,15 @@ type DocumentServices = {
     chunker: DocumentChunker;
     embedder: DocumentEmbedder;
 };
+type DocumentSearchInput = {
+    workspaceId: string;
+    query: string;
+    baseIds?: string[] | undefined;
+    limit?: number | undefined;
+    mode?: DocumentSearchMode | undefined;
+    sourceKinds?: KnowledgeSourceKind[] | undefined;
+    aclTags?: string[] | undefined;
+};
 type DocumentIndexHooks = {
     beforeEmbed?: (input: {
         accountId: string;
@@ -110,15 +119,10 @@ declare function listDocuments(db: Database, workspaceId: string, baseId: string
 declare function getDocument(db: Database, workspaceId: string, documentId: string): Promise<Document | null>;
 declare function queueDocumentForReindex(db: Database, workspaceId: string, documentId: string): Promise<Document>;
 declare function indexDocumentNow(db: Database, objectStorage: ObjectStorage, workspaceId: string, documentId: string, services?: DocumentServices, hooks?: DocumentIndexHooks): Promise<Document>;
-declare function searchDocuments(db: Database, input: {
-    workspaceId: string;
-    query: string;
-    baseIds?: string[];
-    limit?: number;
-}, services?: Pick<DocumentServices, "embedder">): Promise<DocumentSearchResult[]>;
+declare function searchDocuments(db: Database, input: DocumentSearchInput, services?: Pick<DocumentServices, "embedder">): Promise<DocumentSearchResult[]>;
 declare function getDocumentChunk(db: Database, workspaceId: string, chunkId: string): Promise<DocumentSearchResult | null>;
 declare function parseDocumentBytes(bytes: Uint8Array, file: FileAsset, parser?: DocumentParser): Promise<ParsedDocument>;
 declare function chunkText(text: string, maxChars?: number, overlapChars?: number): string[];
 declare function deterministicEmbedding(text: string, dimensions?: number): number[];
 
-export { DEFAULT_DOCUMENT_CHUNK_OVERLAP, DEFAULT_DOCUMENT_CHUNK_SIZE, DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS, DEFAULT_DOCUMENT_EMBEDDING_MODEL, DEFAULT_DOCUMENT_PARSER, DeterministicEmbeddingProvider, type DocumentChunk, type DocumentChunker, type DocumentEmbedder, type DocumentIndexHooks, type DocumentParser, type DocumentServices, LiteParseDocumentParser, OpenAIEmbeddingProvider, type ParsedDocument, RecursiveTextChunker, addDocumentToBase, chunkText, createDocumentBase, createDocumentServices, deleteDocumentFromBase, deterministicEmbedding, documentOpenAIEmbeddingConfig, getDocument, getDocumentBase, getDocumentChunk, indexDocumentNow, listDocumentBases, listDocuments, parseDocumentBytes, queueDocumentForReindex, searchDocuments };
+export { DEFAULT_DOCUMENT_CHUNK_OVERLAP, DEFAULT_DOCUMENT_CHUNK_SIZE, DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS, DEFAULT_DOCUMENT_EMBEDDING_MODEL, DEFAULT_DOCUMENT_PARSER, DeterministicEmbeddingProvider, type DocumentChunk, type DocumentChunker, type DocumentEmbedder, type DocumentIndexHooks, type DocumentParser, type DocumentSearchInput, type DocumentServices, LiteParseDocumentParser, OpenAIEmbeddingProvider, type ParsedDocument, RecursiveTextChunker, addDocumentToBase, chunkText, createDocumentBase, createDocumentServices, deleteDocumentFromBase, deterministicEmbedding, documentOpenAIEmbeddingConfig, getDocument, getDocumentBase, getDocumentChunk, indexDocumentNow, listDocumentBases, listDocuments, parseDocumentBytes, queueDocumentForReindex, searchDocuments };

package/dist/index.js

@@ -221,7 +221,20 @@ async function addDocumentToBase(db, input) {
     const now = /* @__PURE__ */ new Date();
     const [existing] = await scopedDb.select().from(schema.documents).where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.baseId, input.baseId), eq(schema.documents.fileId, input.fileId))).limit(1);
     if (existing) {
-      return mapDocument(existing);
+      const [updated] = await scopedDb.update(schema.documents).set({
+        title: cleanString(input.title) ?? cleanString(input.sourceTitle) ?? existing.title,
+        ...input.sourceKind !== void 0 ? { sourceKind: input.sourceKind } : {},
+        sourceUri: cleanString(input.sourceUri) ?? existing.sourceUri,
+        sourceExternalId: cleanString(input.sourceExternalId) ?? existing.sourceExternalId,
+        sourceTitle: cleanString(input.sourceTitle) ?? existing.sourceTitle,
+        sourceAuthor: cleanString(input.sourceAuthor) ?? existing.sourceAuthor,
+        sourceCreatedAt: parseOptionalDate(input.sourceCreatedAt) ?? existing.sourceCreatedAt,
+        sourceUpdatedAt: parseOptionalDate(input.sourceUpdatedAt) ?? existing.sourceUpdatedAt,
+        sourceVersion: cleanString(input.sourceVersion) ?? existing.sourceVersion,
+        ...input.aclTags !== void 0 ? { aclTags: cleanStringArray(input.aclTags) } : {},
+        updatedAt: now
+      }).where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, existing.id))).returning();
+      return mapDocument(updated ?? existing);
     }
     const [row] = await scopedDb.insert(schema.documents).values({
       accountId: input.accountId,
@@ -229,8 +242,17 @@ async function addDocumentToBase(db, input) {
       baseId: input.baseId,
       fileId: input.fileId,
       status: "queued",
-      title: file.filename,
+      title: cleanString(input.title) ?? cleanString(input.sourceTitle) ?? file.filename,
       parser: DEFAULT_DOCUMENT_PARSER,
+      sourceKind: input.sourceKind ?? "manual_upload",
+      sourceUri: cleanString(input.sourceUri) ?? null,
+      sourceExternalId: cleanString(input.sourceExternalId) ?? null,
+      sourceTitle: cleanString(input.sourceTitle) ?? null,
+      sourceAuthor: cleanString(input.sourceAuthor) ?? null,
+      sourceCreatedAt: parseOptionalDate(input.sourceCreatedAt),
+      sourceUpdatedAt: parseOptionalDate(input.sourceUpdatedAt),
+      sourceVersion: cleanString(input.sourceVersion) ?? null,
+      aclTags: cleanStringArray(input.aclTags),
       updatedAt: now
     }).returning();
     if (!row) throw new Error("Failed to create document");
@@ -313,7 +335,19 @@ async function indexDocumentNow(db, objectStorage, workspaceId, documentId, serv
           fileId: file.id,
           chunkIndex: index,
           text: chunk.text,
-          metadata: chunk.metadata,
+          metadata: {
+            ...chunk.metadata,
+            documentTitle: document.title,
+            sourceKind: document.sourceKind,
+            sourceUri: document.sourceUri,
+            sourceExternalId: document.sourceExternalId,
+            sourceTitle: document.sourceTitle,
+            sourceAuthor: document.sourceAuthor,
+            sourceCreatedAt: document.sourceCreatedAt?.toISOString() ?? null,
+            sourceUpdatedAt: document.sourceUpdatedAt?.toISOString() ?? null,
+            sourceVersion: document.sourceVersion,
+            aclTags: document.aclTags
+          },
           embedding: validateEmbedding(embeddings[index] ?? [], services.embedder.dimensions, services.embedder.model),
           embeddingModel: services.embedder.model
         })));
@@ -344,7 +378,29 @@ async function indexDocumentNow(db, objectStorage, workspaceId, documentId, serv
   return updated;
 }
 async function searchDocuments(db, input, services = createDocumentServices()) {
-  const limit = Math.min(Math.max(input.limit ?? 5, 1), 20);
+  const mode = input.mode ?? "hybrid";
+  const limit = Math.min(Math.max(input.limit ?? 5, 1), 50);
+  const candidateLimit = mode === "hybrid" ? Math.min(limit * 4, 100) : limit;
+  const rows = [];
+  if (mode === "vector" || mode === "hybrid") {
+    try {
+      rows.push(...await vectorSearchDocuments(db, input, candidateLimit, services));
+    } catch (error) {
+      if (mode === "vector") {
+        throw error;
+      }
+      console.warn("document hybrid search vector component failed; falling back to keyword search", {
+        workspaceId: input.workspaceId,
+        error: error instanceof Error ? error.message : String(error)
+      });
+    }
+  }
+  if (mode === "keyword" || mode === "hybrid") {
+    rows.push(...await keywordSearchDocuments(db, input, candidateLimit));
+  }
+  return mergeDocumentSearchRows(rows, mode).slice(0, limit);
+}
+async function vectorSearchDocuments(db, input, limit, services) {
   const queryEmbedding = await services.embedder.embedQuery(input.query);
   validateEmbedding(queryEmbedding, services.embedder.dimensions, services.embedder.model);
   const distance = sql`${schema.documentChunks.embedding} <=> ${vectorLiteral(queryEmbedding)}::vector`;
@@ -360,25 +416,57 @@ async function searchDocuments(db, input, services = createDocumentServices()) {
       text: schema.documentChunks.text,
       chunkIndex: schema.documentChunks.chunkIndex,
       metadata: schema.documentChunks.metadata,
+      sourceKind: schema.documents.sourceKind,
+      sourceUri: schema.documents.sourceUri,
+      sourceExternalId: schema.documents.sourceExternalId,
+      sourceTitle: schema.documents.sourceTitle,
+      sourceAuthor: schema.documents.sourceAuthor,
+      sourceCreatedAt: schema.documents.sourceCreatedAt,
+      sourceUpdatedAt: schema.documents.sourceUpdatedAt,
+      sourceVersion: schema.documents.sourceVersion,
+      aclTags: schema.documents.aclTags,
       distance
+    }).from(schema.documentChunks).innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id)).where(and(...documentSearchConditions(input, services.embedder.model))).orderBy(distance).limit(limit)
+  );
+  return rows.map((row) => ({
+    ...mapSearchRowBase(row, input.workspaceId),
+    vectorScore: 1 / (1 + Number(row.distance)),
+    keywordScore: null
+  }));
+}
+async function keywordSearchDocuments(db, input, limit) {
+  const rank = sql`ts_rank_cd(to_tsvector('simple', ${schema.documentChunks.text}), plainto_tsquery('simple', ${input.query}))`;
+  const rows = await withWorkspaceRls(
+    db,
+    input.workspaceId,
+    async (scopedDb) => await scopedDb.select({
+      chunkId: schema.documentChunks.id,
+      documentId: schema.documentChunks.documentId,
+      baseId: schema.documentChunks.baseId,
+      fileId: schema.documentChunks.fileId,
+      title: schema.documents.title,
+      text: schema.documentChunks.text,
+      chunkIndex: schema.documentChunks.chunkIndex,
+      metadata: schema.documentChunks.metadata,
+      sourceKind: schema.documents.sourceKind,
+      sourceUri: schema.documents.sourceUri,
+      sourceExternalId: schema.documents.sourceExternalId,
+      sourceTitle: schema.documents.sourceTitle,
+      sourceAuthor: schema.documents.sourceAuthor,
+      sourceCreatedAt: schema.documents.sourceCreatedAt,
+      sourceUpdatedAt: schema.documents.sourceUpdatedAt,
+      sourceVersion: schema.documents.sourceVersion,
+      aclTags: schema.documents.aclTags,
+      rank
     }).from(schema.documentChunks).innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id)).where(and(
-      eq(schema.documents.status, "ready"),
-      eq(schema.documentChunks.workspaceId, input.workspaceId),
-      eq(schema.documentChunks.embeddingModel, services.embedder.model),
-      input.baseIds && input.baseIds.length > 0 ? inArray(schema.documentChunks.baseId, input.baseIds) : void 0
-    )).orderBy(distance).limit(limit)
+      ...documentSearchConditions(input),
+      sql`to_tsvector('simple', ${schema.documentChunks.text}) @@ plainto_tsquery('simple', ${input.query})`
+    )).orderBy(desc(rank)).limit(limit)
   );
   return rows.map((row) => ({
-    chunkId: row.chunkId,
-    workspaceId: input.workspaceId,
-    documentId: row.documentId,
-    baseId: row.baseId,
-    fileId: row.fileId,
-    title: row.title,
-    text: row.text,
-    score: 1 / (1 + Number(row.distance)),
-    chunkIndex: row.chunkIndex,
-    metadata: row.metadata
+    ...mapSearchRowBase(row, input.workspaceId),
+    vectorScore: null,
+    keywordScore: normalizeKeywordScore(Number(row.rank))
   }));
 }
 async function getDocumentChunk(db, workspaceId, chunkId) {
@@ -393,10 +481,48 @@ async function getDocumentChunk(db, workspaceId, chunkId) {
       title: schema.documents.title,
       text: schema.documentChunks.text,
       chunkIndex: schema.documentChunks.chunkIndex,
-      metadata: schema.documentChunks.metadata
+      metadata: schema.documentChunks.metadata,
+      sourceKind: schema.documents.sourceKind,
+      sourceUri: schema.documents.sourceUri,
+      sourceExternalId: schema.documents.sourceExternalId,
+      sourceTitle: schema.documents.sourceTitle,
+      sourceAuthor: schema.documents.sourceAuthor,
+      sourceCreatedAt: schema.documents.sourceCreatedAt,
+      sourceUpdatedAt: schema.documents.sourceUpdatedAt,
+      sourceVersion: schema.documents.sourceVersion,
+      aclTags: schema.documents.aclTags
     }).from(schema.documentChunks).innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id)).where(and(eq(schema.documentChunks.workspaceId, workspaceId), eq(schema.documentChunks.id, chunkId), eq(schema.documents.status, "ready"))).limit(1)
   );
   if (!row) return null;
+  return {
+    ...mapSearchRowBase(row, workspaceId),
+    score: 1,
+    matchType: "hybrid",
+    vectorScore: null,
+    keywordScore: null
+  };
+}
+function documentSearchConditions(input, embeddingModel) {
+  const conditions = [
+    eq(schema.documents.status, "ready"),
+    eq(schema.documentChunks.workspaceId, input.workspaceId)
+  ];
+  if (embeddingModel) {
+    conditions.push(eq(schema.documentChunks.embeddingModel, embeddingModel));
+  }
+  if (input.baseIds?.length) {
+    conditions.push(inArray(schema.documentChunks.baseId, input.baseIds));
+  }
+  if (input.sourceKinds?.length) {
+    conditions.push(inArray(schema.documents.sourceKind, input.sourceKinds));
+  }
+  const aclTags = cleanStringArray(input.aclTags);
+  if (aclTags.length > 0) {
+    conditions.push(sql`${schema.documents.aclTags} @> ${JSON.stringify(aclTags)}::jsonb`);
+  }
+  return conditions;
+}
+function mapSearchRowBase(row, workspaceId) {
   return {
     chunkId: row.chunkId,
     workspaceId,
@@ -405,11 +531,62 @@ async function getDocumentChunk(db, workspaceId, chunkId) {
     fileId: row.fileId,
     title: row.title,
     text: row.text,
-    score: 1,
     chunkIndex: row.chunkIndex,
-    metadata: row.metadata
+    metadata: row.metadata,
+    sourceKind: normalizeKnowledgeSourceKind(row.sourceKind),
+    sourceUri: row.sourceUri,
+    sourceExternalId: row.sourceExternalId,
+    sourceTitle: row.sourceTitle,
+    sourceAuthor: row.sourceAuthor,
+    sourceCreatedAt: row.sourceCreatedAt?.toISOString() ?? null,
+    sourceUpdatedAt: row.sourceUpdatedAt?.toISOString() ?? null,
+    sourceVersion: row.sourceVersion,
+    aclTags: cleanStringArray(row.aclTags)
   };
 }
+function mergeDocumentSearchRows(rows, mode) {
+  const byChunk = /* @__PURE__ */ new Map();
+  for (const row of rows) {
+    const existing = byChunk.get(row.chunkId);
+    if (!existing) {
+      byChunk.set(row.chunkId, row);
+      continue;
+    }
+    byChunk.set(row.chunkId, {
+      ...existing,
+      vectorScore: Math.max(existing.vectorScore ?? 0, row.vectorScore ?? 0) || null,
+      keywordScore: Math.max(existing.keywordScore ?? 0, row.keywordScore ?? 0) || null
+    });
+  }
+  return [...byChunk.values()].map((row) => {
+    const vectorScore = row.vectorScore;
+    const keywordScore = row.keywordScore;
+    const matchType = vectorScore !== null && keywordScore !== null ? "hybrid" : vectorScore !== null ? "vector" : "keyword";
+    return {
+      ...row,
+      score: combinedSearchScore(mode, vectorScore, keywordScore, matchType),
+      matchType
+    };
+  }).sort(
+    (left, right) => right.score - left.score || (right.vectorScore ?? 0) - (left.vectorScore ?? 0) || (right.keywordScore ?? 0) - (left.keywordScore ?? 0) || left.chunkIndex - right.chunkIndex
+  );
+}
+function combinedSearchScore(mode, vectorScore, keywordScore, matchType) {
+  const vector = vectorScore ?? 0;
+  const keyword = keywordScore ?? 0;
+  if (mode === "vector") return roundScore(vector);
+  if (mode === "keyword") return roundScore(keyword);
+  return roundScore(Math.min(1, 0.65 * vector + 0.35 * keyword + (matchType === "hybrid" ? 0.1 : 0)));
+}
+function normalizeKeywordScore(rank) {
+  if (!Number.isFinite(rank) || rank <= 0) {
+    return 0;
+  }
+  return rank / (rank + 1);
+}
+function roundScore(value) {
+  return Number(value.toFixed(6));
+}
 async function parseDocumentBytes(bytes, file, parser = new LiteParseDocumentParser()) {
   return await parser.parse(bytes, file);
 }
@@ -499,6 +676,39 @@ function withOverlap(previous, overlapChars, next, maxChars) {
   const candidate = overlap ? `${overlap} ${next}` : next;
   return candidate.length <= maxChars ? candidate : next;
 }
+function cleanString(value) {
+  const trimmed = value?.trim();
+  return trimmed ? trimmed : void 0;
+}
+function cleanStringArray(values) {
+  return [...new Set((values ?? []).map((value) => value.trim()).filter(Boolean))];
+}
+function parseOptionalDate(value) {
+  const trimmed = cleanString(value);
+  if (!trimmed) {
+    return null;
+  }
+  const date = new Date(trimmed);
+  if (Number.isNaN(date.getTime())) {
+    throw new Error(`Invalid source timestamp: ${trimmed}`);
+  }
+  return date;
+}
+function normalizeKnowledgeSourceKind(value) {
+  switch (value) {
+    case "manual_upload":
+    case "meeting_transcript":
+    case "repository":
+    case "email":
+    case "chat":
+    case "document":
+    case "web":
+    case "other":
+      return value;
+    default:
+      return "other";
+  }
+}
 function vectorLiteral(values) {
   return `[${values.join(",")}]`;
 }
@@ -523,6 +733,15 @@ function mapDocument(row) {
     parser: row.parser,
     chunkCount: row.chunkCount,
     error: row.error,
+    sourceKind: normalizeKnowledgeSourceKind(row.sourceKind),
+    sourceUri: row.sourceUri,
+    sourceExternalId: row.sourceExternalId,
+    sourceTitle: row.sourceTitle,
+    sourceAuthor: row.sourceAuthor,
+    sourceCreatedAt: row.sourceCreatedAt?.toISOString() ?? null,
+    sourceUpdatedAt: row.sourceUpdatedAt?.toISOString() ?? null,
+    sourceVersion: row.sourceVersion,
+    aclTags: cleanStringArray(row.aclTags),
     createdAt: row.createdAt.toISOString(),
     updatedAt: row.updatedAt.toISOString()
   };

package/dist/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/index.ts"],"sourcesContent":["import type { Settings } from \"@opengeni/config\";\nimport type { AddDocumentRequest, CreateDocumentBaseRequest, Document, DocumentBase, DocumentSearchResult, FileAsset } from \"@opengeni/contracts\";\nimport { requireFile, withRlsContext, withWorkspaceRls, type Database } from \"@opengeni/db\";\nimport * as schema from \"@opengeni/db/schema\";\nimport type { ObjectStorage } from \"@opengeni/storage\";\nimport { LiteParse } from \"@llamaindex/liteparse\";\nimport { and, asc, desc, eq, inArray, sql } from \"drizzle-orm\";\nimport OpenAI from \"openai\";\n\nexport const DEFAULT_DOCUMENT_PARSER = \"liteparse\";\nexport const DEFAULT_DOCUMENT_EMBEDDING_MODEL = \"text-embedding-3-large\";\nexport const DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS = 3072;\nexport const DEFAULT_DOCUMENT_CHUNK_SIZE = 1200;\nexport const DEFAULT_DOCUMENT_CHUNK_OVERLAP = 160;\n\nexport type ParsedDocument = {\n  text: string;\n  metadata?: Record<string, unknown>;\n};\n\nexport type DocumentChunk = {\n  text: string;\n  metadata: Record<string, unknown>;\n};\n\nexport type DocumentParser = {\n  name: string;\n  parse: (bytes: Uint8Array, file: FileAsset) => Promise<ParsedDocument>;\n};\n\nexport type DocumentChunker = {\n  chunk: (parsed: ParsedDocument, file: FileAsset) => DocumentChunk[];\n};\n\nexport type DocumentEmbedder = {\n  model: string;\n  dimensions: number;\n  embedMany: (texts: string[]) => Promise<number[][]>;\n  embedQuery: (text: string) => Promise<number[]>;\n};\n\nexport type DocumentServices = {\n  parser: DocumentParser;\n  chunker: DocumentChunker;\n  embedder: DocumentEmbedder;\n};\n\nexport type DocumentIndexHooks = {\n  beforeEmbed?: (input: {\n    accountId: string;\n    workspaceId: string;\n    documentId: string;\n    chunkCount: number;\n  }) => Promise<void>;\n};\n\nexport class LiteParseDocumentParser implements DocumentParser {\n  readonly name = DEFAULT_DOCUMENT_PARSER;\n  private parseQueue: Promise<void> = Promise.resolve();\n\n  async parse(bytes: Uint8Array, file: FileAsset): Promise<ParsedDocument> {\n    const text = isTextLike(file)\n      ? Buffer.from(bytes).toString(\"utf8\").replace(/\\0/g, \" \").trim()\n      : await this.parseWithLiteParse(bytes);\n    if (!text.trim()) {\n      throw new Error(`Parsed document is empty: ${file.filename}`);\n    }\n    return {\n      text: text.trim(),\n      metadata: {\n        parser: this.name,\n        filename: file.filename,\n        contentType: file.contentType,\n      },\n    };\n  }\n\n  private async parseWithLiteParse(bytes: Uint8Array): Promise<string> {\n    return await this.enqueueParse(async () => {\n      const parser = new LiteParse({ ocrEnabled: true, numWorkers: 1 });\n      const result = await parser.parse(Buffer.from(bytes), true);\n      const text = typeof result?.text === \"string\" ? result.text : \"\";\n      return text.replace(/\\0/g, \" \").trim();\n    });\n  }\n\n  private async enqueueParse<T>(task: () => Promise<T>): Promise<T> {\n    const previous = this.parseQueue;\n    let release: () => void = () => undefined;\n    this.parseQueue = new Promise<void>((resolve) => {\n      release = resolve;\n    });\n    await previous.catch(() => undefined);\n    try {\n      return await task();\n    } finally {\n      release();\n    }\n  }\n}\n\nexport class RecursiveTextChunker implements DocumentChunker {\n  constructor(\n    private readonly maxChars = DEFAULT_DOCUMENT_CHUNK_SIZE,\n    private readonly overlapChars = DEFAULT_DOCUMENT_CHUNK_OVERLAP,\n  ) {\n    if (overlapChars >= maxChars) {\n      throw new Error(\"document chunk overlap must be smaller than chunk size\");\n    }\n  }\n\n  chunk(parsed: ParsedDocument, file: FileAsset): DocumentChunk[] {\n    return chunkText(parsed.text, this.maxChars, this.overlapChars).map((text, index) => ({\n      text,\n      metadata: {\n        ...parsed.metadata,\n        filename: file.filename,\n        contentType: file.contentType,\n        chunkIndex: index,\n      },\n    }));\n  }\n}\n\nexport class OpenAIEmbeddingProvider implements DocumentEmbedder {\n  private client: OpenAI | null = null;\n  private readonly apiKey: string | undefined;\n  private readonly baseURL: string | undefined;\n\n  constructor(args: {\n    apiKey?: string | undefined;\n    baseURL?: string | undefined;\n    defaultHeaders?: Record<string, string> | undefined;\n    defaultQuery?: Record<string, string> | undefined;\n    model?: string | undefined;\n    dimensions?: number | undefined;\n  }) {\n    this.apiKey = args.apiKey ?? process.env.OPENAI_API_KEY;\n    this.baseURL = args.baseURL;\n    this.defaultHeaders = args.defaultHeaders;\n    this.defaultQuery = args.defaultQuery;\n    this.model = args.model ?? DEFAULT_DOCUMENT_EMBEDDING_MODEL;\n    this.dimensions = args.dimensions ?? DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS;\n  }\n\n  readonly model: string;\n  readonly dimensions: number;\n  private readonly defaultHeaders: Record<string, string> | undefined;\n  private readonly defaultQuery: Record<string, string> | undefined;\n\n  async embedMany(texts: string[]): Promise<number[][]> {\n    if (texts.length === 0) return [];\n    const out: number[][] = [];\n    for (let start = 0; start < texts.length; start += 64) {\n      const batch = texts.slice(start, start + 64);\n      const response = await this.openai().embeddings.create({\n        model: this.model,\n        input: batch,\n        dimensions: this.dimensions,\n      });\n      for (const item of response.data) {\n        out.push(validateEmbedding(item.embedding, this.dimensions, this.model));\n      }\n    }\n    return out;\n  }\n\n  async embedQuery(text: string): Promise<number[]> {\n    const [embedding] = await this.embedMany([text]);\n    if (!embedding) {\n      throw new Error(\"Embedding provider returned no query embedding\");\n    }\n    return embedding;\n  }\n\n  private openai(): OpenAI {\n    if (!this.apiKey) {\n      throw new Error(\"OpenAI document embeddings require an API key\");\n    }\n    this.client ??= new OpenAI({\n      apiKey: this.apiKey,\n      ...(this.baseURL ? { baseURL: this.baseURL } : {}),\n      ...(this.defaultQuery ? { defaultQuery: this.defaultQuery } : {}),\n      ...(this.defaultHeaders ? { defaultHeaders: this.defaultHeaders } : {}),\n    });\n    return this.client;\n  }\n}\n\nexport class DeterministicEmbeddingProvider implements DocumentEmbedder {\n  constructor(\n    readonly dimensions = DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS,\n    readonly model = `deterministic-local-${dimensions}`,\n  ) {}\n\n  async embedMany(texts: string[]): Promise<number[][]> {\n    return texts.map((text) => deterministicEmbedding(text, this.dimensions));\n  }\n\n  async embedQuery(text: string): Promise<number[]> {\n    return deterministicEmbedding(text, this.dimensions);\n  }\n}\n\nexport function createDocumentServices(settings?: Settings, overrides: Partial<DocumentServices> = {}): DocumentServices {\n  const dimensions = settings?.documentEmbeddingDimensions ?? DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS;\n  const openAIEmbeddingConfig = documentOpenAIEmbeddingConfig(settings);\n  return {\n    parser: overrides.parser ?? new LiteParseDocumentParser(),\n    chunker: overrides.chunker ?? new RecursiveTextChunker(\n      settings?.documentChunkSize ?? DEFAULT_DOCUMENT_CHUNK_SIZE,\n      settings?.documentChunkOverlap ?? DEFAULT_DOCUMENT_CHUNK_OVERLAP,\n    ),\n    embedder: overrides.embedder ?? (\n      settings?.documentEmbeddingProvider === \"deterministic\"\n        ? new DeterministicEmbeddingProvider(dimensions, settings.documentEmbeddingModel)\n        : new OpenAIEmbeddingProvider({\n          ...openAIEmbeddingConfig,\n          model: settings?.documentEmbeddingModel ?? DEFAULT_DOCUMENT_EMBEDDING_MODEL,\n          dimensions,\n        })\n    ),\n  };\n}\n\nexport function documentOpenAIEmbeddingConfig(settings?: Settings): {\n  apiKey?: string | undefined;\n  baseURL?: string | undefined;\n  defaultHeaders?: Record<string, string> | undefined;\n  defaultQuery?: Record<string, string> | undefined;\n} {\n  if (!settings) return {};\n  if (settings.documentEmbeddingApiKey || settings.documentEmbeddingBaseUrl) {\n    return {\n      apiKey: settings.documentEmbeddingApiKey ?? settings.openaiApiKey ?? settings.azureOpenaiApiKey,\n      baseURL: settings.documentEmbeddingBaseUrl ?? settings.openaiBaseUrl ?? settings.azureOpenaiBaseUrl,\n    };\n  }\n  if (settings.openaiProvider === \"azure\") {\n    const baseURL = settings.azureOpenaiBaseUrl ?? azureDeploymentBaseUrl(settings);\n    return {\n      apiKey: settings.azureOpenaiApiKey ?? settings.azureOpenaiAdToken ?? \"azure-ad-token\",\n      baseURL,\n      defaultQuery: azureOpenAIDefaultQuery(settings, baseURL),\n      defaultHeaders: settings.azureOpenaiAdToken && !settings.azureOpenaiApiKey\n        ? { Authorization: `Bearer ${settings.azureOpenaiAdToken}` }\n        : undefined,\n    };\n  }\n  return {\n    apiKey: settings.openaiApiKey,\n    baseURL: settings.openaiBaseUrl,\n  };\n}\n\nfunction azureDeploymentBaseUrl(settings: Settings): string {\n  const endpoint = settings.azureOpenaiEndpoint?.replace(/\\/+$/, \"\");\n  if (!endpoint || !settings.azureOpenaiDeployment) {\n    throw new Error(\"Azure OpenAI endpoint/deployment settings are incomplete\");\n  }\n  return `${endpoint}/openai/deployments/${settings.azureOpenaiDeployment}`;\n}\n\nfunction azureOpenAIDefaultQuery(\n  settings: Pick<Settings, \"azureOpenaiApiVersion\">,\n  baseURL: string,\n): Record<string, string> | undefined {\n  if (!settings.azureOpenaiApiVersion) return undefined;\n  const normalized = baseURL.replace(/\\/+$/, \"\").toLowerCase();\n  if (normalized.endsWith(\"/openai/v1\")) {\n    return undefined;\n  }\n  return { \"api-version\": settings.azureOpenaiApiVersion };\n}\n\nexport async function createDocumentBase(db: Database, input: CreateDocumentBaseRequest & { accountId: string; workspaceId: string }): Promise<DocumentBase> {\n  return await withRlsContext(db, { accountId: input.accountId, workspaceId: input.workspaceId }, async (scopedDb) => {\n    const [row] = await scopedDb.insert(schema.documentBases).values({\n      accountId: input.accountId,\n      workspaceId: input.workspaceId,\n      name: input.name.trim(),\n      description: input.description?.trim() || null,\n    }).returning();\n    if (!row) throw new Error(\"Failed to create document base\");\n    return mapDocumentBase(row);\n  });\n}\n\nexport async function listDocumentBases(db: Database, workspaceId: string): Promise<DocumentBase[]> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const rows = await scopedDb.select().from(schema.documentBases)\n      .where(eq(schema.documentBases.workspaceId, workspaceId))\n      .orderBy(desc(schema.documentBases.createdAt));\n    return rows.map(mapDocumentBase);\n  });\n}\n\nexport async function getDocumentBase(db: Database, workspaceId: string, baseId: string): Promise<DocumentBase | null> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const [row] = await scopedDb.select().from(schema.documentBases).where(and(eq(schema.documentBases.workspaceId, workspaceId), eq(schema.documentBases.id, baseId))).limit(1);\n    return row ? mapDocumentBase(row) : null;\n  });\n}\n\nexport async function addDocumentToBase(db: Database, input: AddDocumentRequest & { accountId: string; workspaceId: string; baseId: string }): Promise<Document> {\n  return await withRlsContext(db, { accountId: input.accountId, workspaceId: input.workspaceId }, async (scopedDb) => {\n    const base = await getDocumentBase(scopedDb, input.workspaceId, input.baseId);\n    if (!base) throw new Error(`Document base not found: ${input.baseId}`);\n    const file = await requireReadyFile(scopedDb, input.workspaceId, input.fileId);\n    const now = new Date();\n    const [existing] = await scopedDb.select().from(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.baseId, input.baseId), eq(schema.documents.fileId, input.fileId)))\n      .limit(1);\n    if (existing) {\n      return mapDocument(existing);\n    }\n    const [row] = await scopedDb.insert(schema.documents).values({\n      accountId: input.accountId,\n      workspaceId: input.workspaceId,\n      baseId: input.baseId,\n      fileId: input.fileId,\n      status: \"queued\",\n      title: file.filename,\n      parser: DEFAULT_DOCUMENT_PARSER,\n      updatedAt: now,\n    }).returning();\n    if (!row) throw new Error(\"Failed to create document\");\n    return mapDocument(row);\n  });\n}\n\nexport async function deleteDocumentFromBase(\n  db: Database,\n  input: { accountId: string; workspaceId: string; baseId: string; documentId: string },\n): Promise<void> {\n  await withRlsContext(db, { accountId: input.accountId, workspaceId: input.workspaceId }, async (scopedDb) => {\n    const [document] = await scopedDb.select().from(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, input.documentId)))\n      .limit(1);\n    if (!document) {\n      throw new Error(`Document not found: ${input.documentId}`);\n    }\n    if (document.baseId !== input.baseId) {\n      throw new Error(`Document not found: ${input.documentId}`);\n    }\n    await scopedDb.delete(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, input.documentId)));\n  });\n}\n\nexport async function listDocuments(db: Database, workspaceId: string, baseId: string): Promise<Document[]> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const rows = await scopedDb.select().from(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.baseId, baseId)))\n      .orderBy(asc(schema.documents.createdAt));\n    return rows.map(mapDocument);\n  });\n}\n\nexport async function getDocument(db: Database, workspaceId: string, documentId: string): Promise<Document | null> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const [row] = await scopedDb.select().from(schema.documents).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).limit(1);\n    return row ? mapDocument(row) : null;\n  });\n}\n\nexport async function queueDocumentForReindex(db: Database, workspaceId: string, documentId: string): Promise<Document> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const [row] = await scopedDb.update(schema.documents).set({\n      status: \"queued\",\n      error: null,\n      updatedAt: new Date(),\n    }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).returning();\n    if (!row) throw new Error(`Document not found: ${documentId}`);\n    return mapDocument(row);\n  });\n}\n\nexport async function indexDocumentNow(\n  db: Database,\n  objectStorage: ObjectStorage,\n  workspaceId: string,\n  documentId: string,\n  services: DocumentServices = createDocumentServices(),\n  hooks: DocumentIndexHooks = {},\n): Promise<Document> {\n  const [document] = await withWorkspaceRls(db, workspaceId, async (scopedDb) =>\n    await scopedDb.select().from(schema.documents).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).limit(1)\n  );\n  if (!document) throw new Error(`Document not found: ${documentId}`);\n  const file = await requireReadyFile(db, workspaceId, document.fileId);\n  await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    await scopedDb.update(schema.documents).set({\n      status: \"indexing\",\n      parser: services.parser.name,\n      error: null,\n      updatedAt: new Date(),\n    }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId)));\n  });\n  try {\n    const bytes = await objectStorage.getFileBytes(file);\n    const parsed = await services.parser.parse(bytes, file);\n    const chunks = services.chunker.chunk(parsed, file);\n    await hooks.beforeEmbed?.({\n      accountId: document.accountId,\n      workspaceId: document.workspaceId,\n      documentId,\n      chunkCount: chunks.length,\n    });\n    const embeddings = await services.embedder.embedMany(chunks.map((chunk) => chunk.text));\n    if (embeddings.length !== chunks.length) {\n      throw new Error(`Embedding provider returned ${embeddings.length} embeddings for ${chunks.length} chunks`);\n    }\n    await withWorkspaceRls(db, workspaceId, async (scopedDb) => await scopedDb.transaction(async (tx) => {\n      await tx.delete(schema.documentChunks).where(and(eq(schema.documentChunks.workspaceId, workspaceId), eq(schema.documentChunks.documentId, documentId)));\n      if (chunks.length > 0) {\n        await tx.insert(schema.documentChunks).values(chunks.map((chunk, index) => ({\n          accountId: document.accountId,\n          workspaceId: document.workspaceId,\n          documentId,\n          baseId: document.baseId,\n          fileId: file.id,\n          chunkIndex: index,\n          text: chunk.text,\n          metadata: chunk.metadata,\n          embedding: validateEmbedding(embeddings[index] ?? [], services.embedder.dimensions, services.embedder.model),\n          embeddingModel: services.embedder.model,\n        })));\n      }\n      await tx.update(schema.documents).set({\n        status: \"ready\",\n        parser: services.parser.name,\n        chunkCount: chunks.length,\n        error: null,\n        updatedAt: new Date(),\n      }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId)));\n    }));\n  } catch (error) {\n    const [failed] = await withWorkspaceRls(db, workspaceId, async (scopedDb) =>\n      await scopedDb.update(schema.documents).set({\n        status: \"failed\",\n        error: error instanceof Error ? error.message : String(error),\n        updatedAt: new Date(),\n      }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).returning()\n    );\n    if (!failed) throw error;\n    return mapDocument(failed);\n  }\n  const updated = await getDocument(db, workspaceId, documentId);\n  if (!updated) throw new Error(`Document disappeared after indexing: ${documentId}`);\n  return updated;\n}\n\nexport async function searchDocuments(\n  db: Database,\n  input: { workspaceId: string; query: string; baseIds?: string[]; limit?: number },\n  services: Pick<DocumentServices, \"embedder\"> = createDocumentServices(),\n): Promise<DocumentSearchResult[]> {\n  const limit = Math.min(Math.max(input.limit ?? 5, 1), 20);\n  const queryEmbedding = await services.embedder.embedQuery(input.query);\n  validateEmbedding(queryEmbedding, services.embedder.dimensions, services.embedder.model);\n  const distance = sql<number>`${schema.documentChunks.embedding} <=> ${vectorLiteral(queryEmbedding)}::vector`;\n  const rows = await withWorkspaceRls(db, input.workspaceId, async (scopedDb) =>\n    await scopedDb.select({\n      chunkId: schema.documentChunks.id,\n      documentId: schema.documentChunks.documentId,\n      baseId: schema.documentChunks.baseId,\n      fileId: schema.documentChunks.fileId,\n      title: schema.documents.title,\n      text: schema.documentChunks.text,\n      chunkIndex: schema.documentChunks.chunkIndex,\n      metadata: schema.documentChunks.metadata,\n      distance,\n    }).from(schema.documentChunks)\n      .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))\n      .where(and(\n        eq(schema.documents.status, \"ready\"),\n        eq(schema.documentChunks.workspaceId, input.workspaceId),\n        eq(schema.documentChunks.embeddingModel, services.embedder.model),\n        input.baseIds && input.baseIds.length > 0 ? inArray(schema.documentChunks.baseId, input.baseIds) : undefined,\n      ))\n      .orderBy(distance)\n      .limit(limit)\n  );\n  return rows.map((row) => ({\n    chunkId: row.chunkId,\n    workspaceId: input.workspaceId,\n    documentId: row.documentId,\n    baseId: row.baseId,\n    fileId: row.fileId,\n    title: row.title,\n    text: row.text,\n    score: 1 / (1 + Number(row.distance)),\n    chunkIndex: row.chunkIndex,\n    metadata: row.metadata,\n  }));\n}\n\nexport async function getDocumentChunk(db: Database, workspaceId: string, chunkId: string): Promise<DocumentSearchResult | null> {\n  const [row] = await withWorkspaceRls(db, workspaceId, async (scopedDb) =>\n    await scopedDb.select({\n      chunkId: schema.documentChunks.id,\n      documentId: schema.documentChunks.documentId,\n      baseId: schema.documentChunks.baseId,\n      fileId: schema.documentChunks.fileId,\n      title: schema.documents.title,\n      text: schema.documentChunks.text,\n      chunkIndex: schema.documentChunks.chunkIndex,\n      metadata: schema.documentChunks.metadata,\n    }).from(schema.documentChunks)\n      .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))\n      .where(and(eq(schema.documentChunks.workspaceId, workspaceId), eq(schema.documentChunks.id, chunkId), eq(schema.documents.status, \"ready\")))\n      .limit(1)\n  );\n  if (!row) return null;\n  return {\n    chunkId: row.chunkId,\n    workspaceId,\n    documentId: row.documentId,\n    baseId: row.baseId,\n    fileId: row.fileId,\n    title: row.title,\n    text: row.text,\n    score: 1,\n    chunkIndex: row.chunkIndex,\n    metadata: row.metadata,\n  };\n}\n\nexport async function parseDocumentBytes(bytes: Uint8Array, file: FileAsset, parser: DocumentParser = new LiteParseDocumentParser()): Promise<ParsedDocument> {\n  return await parser.parse(bytes, file);\n}\n\nexport function chunkText(text: string, maxChars = DEFAULT_DOCUMENT_CHUNK_SIZE, overlapChars = DEFAULT_DOCUMENT_CHUNK_OVERLAP): string[] {\n  if (overlapChars >= maxChars) {\n    throw new Error(\"chunk overlap must be smaller than chunk size\");\n  }\n  const normalized = text.replace(/\\r\\n/g, \"\\n\").replace(/[ \\t]+/g, \" \").trim();\n  if (!normalized) return [];\n  const paragraphs = normalized.split(/\\n{2,}/).map((part) => part.replace(/\\s+/g, \" \").trim()).filter(Boolean);\n  const chunks: string[] = [];\n  let current = \"\";\n  for (const paragraph of paragraphs.length > 0 ? paragraphs : [normalized.replace(/\\s+/g, \" \")]) {\n    for (const part of splitOversizedText(paragraph, maxChars)) {\n      if (!current) {\n        current = part;\n      } else if (current.length + 1 + part.length <= maxChars) {\n        current = `${current} ${part}`;\n      } else {\n        chunks.push(current);\n        current = withOverlap(current, overlapChars, part, maxChars);\n      }\n    }\n  }\n  if (current) chunks.push(current);\n  return chunks.map((chunk) => chunk.trim()).filter(Boolean);\n}\n\nexport function deterministicEmbedding(text: string, dimensions = DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS): number[] {\n  const values = new Array(dimensions).fill(0);\n  const tokens = text.toLowerCase().match(/[\\p{L}\\p{N}_-]+/gu) ?? [];\n  for (const token of tokens) {\n    let hash = 2166136261;\n    for (const char of token) {\n      hash ^= char.codePointAt(0) ?? 0;\n      hash = Math.imul(hash, 16777619);\n    }\n    values[Math.abs(hash) % dimensions] += 1;\n  }\n  const norm = Math.hypot(...values) || 1;\n  return values.map((value) => Number((value / norm).toFixed(6)));\n}\n\nasync function requireReadyFile(db: Database, workspaceId: string, fileId: string): Promise<FileAsset> {\n  const file = await requireFile(db, workspaceId, fileId);\n  if (file.status !== \"ready\") {\n    throw new Error(`File ${fileId} is ${file.status}`);\n  }\n  return file;\n}\n\nfunction validateEmbedding(values: number[], dimensions: number, model: string): number[] {\n  if (values.length !== dimensions) {\n    throw new Error(`Embedding model ${model} returned ${values.length} dimensions; expected ${dimensions}`);\n  }\n  if (values.some((value) => !Number.isFinite(value))) {\n    throw new Error(`Embedding model ${model} returned non-finite values`);\n  }\n  return values;\n}\n\nfunction isTextLike(file: FileAsset): boolean {\n  const contentType = file.contentType.toLowerCase();\n  const filename = file.filename.toLowerCase();\n  return contentType.startsWith(\"text/\")\n    || contentType === \"application/json\"\n    || contentType === \"application/xml\"\n    || contentType === \"application/x-yaml\"\n    || filename.endsWith(\".md\")\n    || filename.endsWith(\".markdown\")\n    || filename.endsWith(\".json\")\n    || filename.endsWith(\".yaml\")\n    || filename.endsWith(\".yml\")\n    || filename.endsWith(\".csv\")\n    || filename.endsWith(\".tsv\")\n    || filename.endsWith(\".xml\");\n}\n\nfunction splitOversizedText(text: string, maxChars: number): string[] {\n  if (text.length <= maxChars) return [text];\n  const out: string[] = [];\n  let remaining = text;\n  while (remaining.length > maxChars) {\n    const window = remaining.slice(0, maxChars + 1);\n    const breakAt = Math.max(\n      window.lastIndexOf(\". \"),\n      window.lastIndexOf(\"? \"),\n      window.lastIndexOf(\"! \"),\n      window.lastIndexOf(\"; \"),\n      window.lastIndexOf(\", \"),\n      window.lastIndexOf(\" \"),\n    );\n    const end = breakAt > Math.floor(maxChars * 0.5) ? breakAt + 1 : maxChars;\n    out.push(remaining.slice(0, end).trim());\n    remaining = remaining.slice(end).trim();\n  }\n  if (remaining) out.push(remaining);\n  return out;\n}\n\nfunction withOverlap(previous: string, overlapChars: number, next: string, maxChars: number): string {\n  if (overlapChars <= 0) return next;\n  const overlap = previous.slice(Math.max(0, previous.length - overlapChars)).replace(/^\\S+\\s+/, \"\").trim();\n  const candidate = overlap ? `${overlap} ${next}` : next;\n  return candidate.length <= maxChars ? candidate : next;\n}\n\nfunction vectorLiteral(values: number[]): string {\n  return `[${values.join(\",\")}]`;\n}\n\nfunction mapDocumentBase(row: typeof schema.documentBases.$inferSelect): DocumentBase {\n  return {\n    id: row.id,\n    workspaceId: row.workspaceId,\n    name: row.name,\n    description: row.description,\n    createdAt: row.createdAt.toISOString(),\n    updatedAt: row.updatedAt.toISOString(),\n  };\n}\n\nfunction mapDocument(row: typeof schema.documents.$inferSelect): Document {\n  return {\n    id: row.id,\n    workspaceId: row.workspaceId,\n    baseId: row.baseId,\n    fileId: row.fileId,\n    status: row.status as Document[\"status\"],\n    title: row.title,\n    parser: row.parser,\n    chunkCount: row.chunkCount,\n    error: row.error,\n    createdAt: row.createdAt.toISOString(),\n    updatedAt: row.updatedAt.toISOString(),\n  };\n}\n"],"mappings":";AAEA,SAAS,aAAa,gBAAgB,wBAAuC;AAC7E,YAAY,YAAY;AAExB,SAAS,iBAAiB;AAC1B,SAAS,KAAK,KAAK,MAAM,IAAI,SAAS,WAAW;AACjD,OAAO,YAAY;AAEZ,IAAM,0BAA0B;AAChC,IAAM,mCAAmC;AACzC,IAAM,wCAAwC;AAC9C,IAAM,8BAA8B;AACpC,IAAM,iCAAiC;AA2CvC,IAAM,0BAAN,MAAwD;AAAA,EACpD,OAAO;AAAA,EACR,aAA4B,QAAQ,QAAQ;AAAA,EAEpD,MAAM,MAAM,OAAmB,MAA0C;AACvE,UAAM,OAAO,WAAW,IAAI,IACxB,OAAO,KAAK,KAAK,EAAE,SAAS,MAAM,EAAE,QAAQ,OAAO,GAAG,EAAE,KAAK,IAC7D,MAAM,KAAK,mBAAmB,KAAK;AACvC,QAAI,CAAC,KAAK,KAAK,GAAG;AAChB,YAAM,IAAI,MAAM,6BAA6B,KAAK,QAAQ,EAAE;AAAA,IAC9D;AACA,WAAO;AAAA,MACL,MAAM,KAAK,KAAK;AAAA,MAChB,UAAU;AAAA,QACR,QAAQ,KAAK;AAAA,QACb,UAAU,KAAK;AAAA,QACf,aAAa,KAAK;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,mBAAmB,OAAoC;AACnE,WAAO,MAAM,KAAK,aAAa,YAAY;AACzC,YAAM,SAAS,IAAI,UAAU,EAAE,YAAY,MAAM,YAAY,EAAE,CAAC;AAChE,YAAM,SAAS,MAAM,OAAO,MAAM,OAAO,KAAK,KAAK,GAAG,IAAI;AAC1D,YAAM,OAAO,OAAO,QAAQ,SAAS,WAAW,OAAO,OAAO;AAC9D,aAAO,KAAK,QAAQ,OAAO,GAAG,EAAE,KAAK;AAAA,IACvC,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,aAAgB,MAAoC;AAChE,UAAM,WAAW,KAAK;AACtB,QAAI,UAAsB,MAAM;AAChC,SAAK,aAAa,IAAI,QAAc,CAAC,YAAY;AAC/C,gBAAU;AAAA,IACZ,CAAC;AACD,UAAM,SAAS,MAAM,MAAM,MAAS;AACpC,QAAI;AACF,aAAO,MAAM,KAAK;AAAA,IACpB,UAAE;AACA,cAAQ;AAAA,IACV;AAAA,EACF;AACF;AAEO,IAAM,uBAAN,MAAsD;AAAA,EAC3D,YACmB,WAAW,6BACX,eAAe,gCAChC;AAFiB;AACA;AAEjB,QAAI,gBAAgB,UAAU;AAC5B,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAAA,EACF;AAAA,EANmB;AAAA,EACA;AAAA,EAOnB,MAAM,QAAwB,MAAkC;AAC9D,WAAO,UAAU,OAAO,MAAM,KAAK,UAAU,KAAK,YAAY,EAAE,IAAI,CAAC,MAAM,WAAW;AAAA,MACpF;AAAA,MACA,UAAU;AAAA,QACR,GAAG,OAAO;AAAA,QACV,UAAU,KAAK;AAAA,QACf,aAAa,KAAK;AAAA,QAClB,YAAY;AAAA,MACd;AAAA,IACF,EAAE;AAAA,EACJ;AACF;AAEO,IAAM,0BAAN,MAA0D;AAAA,EACvD,SAAwB;AAAA,EACf;AAAA,EACA;AAAA,EAEjB,YAAY,MAOT;AACD,SAAK,SAAS,KAAK,UAAU,QAAQ,IAAI;AACzC,SAAK,UAAU,KAAK;AACpB,SAAK,iBAAiB,KAAK;AAC3B,SAAK,eAAe,KAAK;AACzB,SAAK,QAAQ,KAAK,SAAS;AAC3B,SAAK,aAAa,KAAK,cAAc;AAAA,EACvC;AAAA,EAES;AAAA,EACA;AAAA,EACQ;AAAA,EACA;AAAA,EAEjB,MAAM,UAAU,OAAsC;AACpD,QAAI,MAAM,WAAW,EAAG,QAAO,CAAC;AAChC,UAAM,MAAkB,CAAC;AACzB,aAAS,QAAQ,GAAG,QAAQ,MAAM,QAAQ,SAAS,IAAI;AACrD,YAAM,QAAQ,MAAM,MAAM,OAAO,QAAQ,EAAE;AAC3C,YAAM,WAAW,MAAM,KAAK,OAAO,EAAE,WAAW,OAAO;AAAA,QACrD,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,QACP,YAAY,KAAK;AAAA,MACnB,CAAC;AACD,iBAAW,QAAQ,SAAS,MAAM;AAChC,YAAI,KAAK,kBAAkB,KAAK,WAAW,KAAK,YAAY,KAAK,KAAK,CAAC;AAAA,MACzE;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,WAAW,MAAiC;AAChD,UAAM,CAAC,SAAS,IAAI,MAAM,KAAK,UAAU,CAAC,IAAI,CAAC;AAC/C,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,gDAAgD;AAAA,IAClE;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,SAAiB;AACvB,QAAI,CAAC,KAAK,QAAQ;AAChB,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,SAAK,WAAW,IAAI,OAAO;AAAA,MACzB,QAAQ,KAAK;AAAA,MACb,GAAI,KAAK,UAAU,EAAE,SAAS,KAAK,QAAQ,IAAI,CAAC;AAAA,MAChD,GAAI,KAAK,eAAe,EAAE,cAAc,KAAK,aAAa,IAAI,CAAC;AAAA,MAC/D,GAAI,KAAK,iBAAiB,EAAE,gBAAgB,KAAK,eAAe,IAAI,CAAC;AAAA,IACvE,CAAC;AACD,WAAO,KAAK;AAAA,EACd;AACF;AAEO,IAAM,iCAAN,MAAiE;AAAA,EACtE,YACW,aAAa,uCACb,QAAQ,uBAAuB,UAAU,IAClD;AAFS;AACA;AAAA,EACR;AAAA,EAFQ;AAAA,EACA;AAAA,EAGX,MAAM,UAAU,OAAsC;AACpD,WAAO,MAAM,IAAI,CAAC,SAAS,uBAAuB,MAAM,KAAK,UAAU,CAAC;AAAA,EAC1E;AAAA,EAEA,MAAM,WAAW,MAAiC;AAChD,WAAO,uBAAuB,MAAM,KAAK,UAAU;AAAA,EACrD;AACF;AAEO,SAAS,uBAAuB,UAAqB,YAAuC,CAAC,GAAqB;AACvH,QAAM,aAAa,UAAU,+BAA+B;AAC5D,QAAM,wBAAwB,8BAA8B,QAAQ;AACpE,SAAO;AAAA,IACL,QAAQ,UAAU,UAAU,IAAI,wBAAwB;AAAA,IACxD,SAAS,UAAU,WAAW,IAAI;AAAA,MAChC,UAAU,qBAAqB;AAAA,MAC/B,UAAU,wBAAwB;AAAA,IACpC;AAAA,IACA,UAAU,UAAU,aAClB,UAAU,8BAA8B,kBACpC,IAAI,+BAA+B,YAAY,SAAS,sBAAsB,IAC9E,IAAI,wBAAwB;AAAA,MAC5B,GAAG;AAAA,MACH,OAAO,UAAU,0BAA0B;AAAA,MAC3C;AAAA,IACF,CAAC;AAAA,EAEP;AACF;AAEO,SAAS,8BAA8B,UAK5C;AACA,MAAI,CAAC,SAAU,QAAO,CAAC;AACvB,MAAI,SAAS,2BAA2B,SAAS,0BAA0B;AACzE,WAAO;AAAA,MACL,QAAQ,SAAS,2BAA2B,SAAS,gBAAgB,SAAS;AAAA,MAC9E,SAAS,SAAS,4BAA4B,SAAS,iBAAiB,SAAS;AAAA,IACnF;AAAA,EACF;AACA,MAAI,SAAS,mBAAmB,SAAS;AACvC,UAAM,UAAU,SAAS,sBAAsB,uBAAuB,QAAQ;AAC9E,WAAO;AAAA,MACL,QAAQ,SAAS,qBAAqB,SAAS,sBAAsB;AAAA,MACrE;AAAA,MACA,cAAc,wBAAwB,UAAU,OAAO;AAAA,MACvD,gBAAgB,SAAS,sBAAsB,CAAC,SAAS,oBACrD,EAAE,eAAe,UAAU,SAAS,kBAAkB,GAAG,IACzD;AAAA,IACN;AAAA,EACF;AACA,SAAO;AAAA,IACL,QAAQ,SAAS;AAAA,IACjB,SAAS,SAAS;AAAA,EACpB;AACF;AAEA,SAAS,uBAAuB,UAA4B;AAC1D,QAAM,WAAW,SAAS,qBAAqB,QAAQ,QAAQ,EAAE;AACjE,MAAI,CAAC,YAAY,CAAC,SAAS,uBAAuB;AAChD,UAAM,IAAI,MAAM,0DAA0D;AAAA,EAC5E;AACA,SAAO,GAAG,QAAQ,uBAAuB,SAAS,qBAAqB;AACzE;AAEA,SAAS,wBACP,UACA,SACoC;AACpC,MAAI,CAAC,SAAS,sBAAuB,QAAO;AAC5C,QAAM,aAAa,QAAQ,QAAQ,QAAQ,EAAE,EAAE,YAAY;AAC3D,MAAI,WAAW,SAAS,YAAY,GAAG;AACrC,WAAO;AAAA,EACT;AACA,SAAO,EAAE,eAAe,SAAS,sBAAsB;AACzD;AAEA,eAAsB,mBAAmB,IAAc,OAAsG;AAC3J,SAAO,MAAM,eAAe,IAAI,EAAE,WAAW,MAAM,WAAW,aAAa,MAAM,YAAY,GAAG,OAAO,aAAa;AAClH,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAc,oBAAa,EAAE,OAAO;AAAA,MAC/D,WAAW,MAAM;AAAA,MACjB,aAAa,MAAM;AAAA,MACnB,MAAM,MAAM,KAAK,KAAK;AAAA,MACtB,aAAa,MAAM,aAAa,KAAK,KAAK;AAAA,IAC5C,CAAC,EAAE,UAAU;AACb,QAAI,CAAC,IAAK,OAAM,IAAI,MAAM,gCAAgC;AAC1D,WAAO,gBAAgB,GAAG;AAAA,EAC5B,CAAC;AACH;AAEA,eAAsB,kBAAkB,IAAc,aAA8C;AAClG,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,OAAO,MAAM,SAAS,OAAO,EAAE,KAAY,oBAAa,EAC3D,MAAM,GAAU,qBAAc,aAAa,WAAW,CAAC,EACvD,QAAQ,KAAY,qBAAc,SAAS,CAAC;AAC/C,WAAO,KAAK,IAAI,eAAe;AAAA,EACjC,CAAC;AACH;AAEA,eAAsB,gBAAgB,IAAc,aAAqB,QAA8C;AACrH,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,oBAAa,EAAE,MAAM,IAAI,GAAU,qBAAc,aAAa,WAAW,GAAG,GAAU,qBAAc,IAAI,MAAM,CAAC,CAAC,EAAE,MAAM,CAAC;AAC3K,WAAO,MAAM,gBAAgB,GAAG,IAAI;AAAA,EACtC,CAAC;AACH;AAEA,eAAsB,kBAAkB,IAAc,OAA2G;AAC/J,SAAO,MAAM,eAAe,IAAI,EAAE,WAAW,MAAM,WAAW,aAAa,MAAM,YAAY,GAAG,OAAO,aAAa;AAClH,UAAM,OAAO,MAAM,gBAAgB,UAAU,MAAM,aAAa,MAAM,MAAM;AAC5E,QAAI,CAAC,KAAM,OAAM,IAAI,MAAM,4BAA4B,MAAM,MAAM,EAAE;AACrE,UAAM,OAAO,MAAM,iBAAiB,UAAU,MAAM,aAAa,MAAM,MAAM;AAC7E,UAAM,MAAM,oBAAI,KAAK;AACrB,UAAM,CAAC,QAAQ,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAC7D,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,QAAQ,MAAM,MAAM,GAAG,GAAU,iBAAU,QAAQ,MAAM,MAAM,CAAC,CAAC,EACpJ,MAAM,CAAC;AACV,QAAI,UAAU;AACZ,aAAO,YAAY,QAAQ;AAAA,IAC7B;AACA,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAc,gBAAS,EAAE,OAAO;AAAA,MAC3D,WAAW,MAAM;AAAA,MACjB,aAAa,MAAM;AAAA,MACnB,QAAQ,MAAM;AAAA,MACd,QAAQ,MAAM;AAAA,MACd,QAAQ;AAAA,MACR,OAAO,KAAK;AAAA,MACZ,QAAQ;AAAA,MACR,WAAW;AAAA,IACb,CAAC,EAAE,UAAU;AACb,QAAI,CAAC,IAAK,OAAM,IAAI,MAAM,2BAA2B;AACrD,WAAO,YAAY,GAAG;AAAA,EACxB,CAAC;AACH;AAEA,eAAsB,uBACpB,IACA,OACe;AACf,QAAM,eAAe,IAAI,EAAE,WAAW,MAAM,WAAW,aAAa,MAAM,YAAY,GAAG,OAAO,aAAa;AAC3G,UAAM,CAAC,QAAQ,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAC7D,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,IAAI,MAAM,UAAU,CAAC,CAAC,EACzG,MAAM,CAAC;AACV,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,uBAAuB,MAAM,UAAU,EAAE;AAAA,IAC3D;AACA,QAAI,SAAS,WAAW,MAAM,QAAQ;AACpC,YAAM,IAAI,MAAM,uBAAuB,MAAM,UAAU,EAAE;AAAA,IAC3D;AACA,UAAM,SAAS,OAAc,gBAAS,EACnC,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,IAAI,MAAM,UAAU,CAAC,CAAC;AAAA,EAC9G,CAAC;AACH;AAEA,eAAsB,cAAc,IAAc,aAAqB,QAAqC;AAC1G,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,OAAO,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EACvD,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,QAAQ,MAAM,CAAC,CAAC,EAC7F,QAAQ,IAAW,iBAAU,SAAS,CAAC;AAC1C,WAAO,KAAK,IAAI,WAAW;AAAA,EAC7B,CAAC;AACH;AAEA,eAAsB,YAAY,IAAc,aAAqB,YAA8C;AACjH,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,MAAM,CAAC;AACnK,WAAO,MAAM,YAAY,GAAG,IAAI;AAAA,EAClC,CAAC;AACH;AAEA,eAAsB,wBAAwB,IAAc,aAAqB,YAAuC;AACtH,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,MACxD,QAAQ;AAAA,MACR,OAAO;AAAA,MACP,WAAW,oBAAI,KAAK;AAAA,IACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,UAAU;AAC5G,QAAI,CAAC,IAAK,OAAM,IAAI,MAAM,uBAAuB,UAAU,EAAE;AAC7D,WAAO,YAAY,GAAG;AAAA,EACxB,CAAC;AACH;AAEA,eAAsB,iBACpB,IACA,eACA,aACA,YACA,WAA6B,uBAAuB,GACpD,QAA4B,CAAC,GACV;AACnB,QAAM,CAAC,QAAQ,IAAI,MAAM;AAAA,IAAiB;AAAA,IAAI;AAAA,IAAa,OAAO,aAChE,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,MAAM,CAAC;AAAA,EACvJ;AACA,MAAI,CAAC,SAAU,OAAM,IAAI,MAAM,uBAAuB,UAAU,EAAE;AAClE,QAAM,OAAO,MAAM,iBAAiB,IAAI,aAAa,SAAS,MAAM;AACpE,QAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AAC1D,UAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,MAC1C,QAAQ;AAAA,MACR,QAAQ,SAAS,OAAO;AAAA,MACxB,OAAO;AAAA,MACP,WAAW,oBAAI,KAAK;AAAA,IACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC;AAAA,EAClG,CAAC;AACD,MAAI;AACF,UAAM,QAAQ,MAAM,cAAc,aAAa,IAAI;AACnD,UAAM,SAAS,MAAM,SAAS,OAAO,MAAM,OAAO,IAAI;AACtD,UAAM,SAAS,SAAS,QAAQ,MAAM,QAAQ,IAAI;AAClD,UAAM,MAAM,cAAc;AAAA,MACxB,WAAW,SAAS;AAAA,MACpB,aAAa,SAAS;AAAA,MACtB;AAAA,MACA,YAAY,OAAO;AAAA,IACrB,CAAC;AACD,UAAM,aAAa,MAAM,SAAS,SAAS,UAAU,OAAO,IAAI,CAAC,UAAU,MAAM,IAAI,CAAC;AACtF,QAAI,WAAW,WAAW,OAAO,QAAQ;AACvC,YAAM,IAAI,MAAM,+BAA+B,WAAW,MAAM,mBAAmB,OAAO,MAAM,SAAS;AAAA,IAC3G;AACA,UAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa,MAAM,SAAS,YAAY,OAAO,OAAO;AACnG,YAAM,GAAG,OAAc,qBAAc,EAAE,MAAM,IAAI,GAAU,sBAAe,aAAa,WAAW,GAAG,GAAU,sBAAe,YAAY,UAAU,CAAC,CAAC;AACtJ,UAAI,OAAO,SAAS,GAAG;AACrB,cAAM,GAAG,OAAc,qBAAc,EAAE,OAAO,OAAO,IAAI,CAAC,OAAO,WAAW;AAAA,UAC1E,WAAW,SAAS;AAAA,UACpB,aAAa,SAAS;AAAA,UACtB;AAAA,UACA,QAAQ,SAAS;AAAA,UACjB,QAAQ,KAAK;AAAA,UACb,YAAY;AAAA,UACZ,MAAM,MAAM;AAAA,UACZ,UAAU,MAAM;AAAA,UAChB,WAAW,kBAAkB,WAAW,KAAK,KAAK,CAAC,GAAG,SAAS,SAAS,YAAY,SAAS,SAAS,KAAK;AAAA,UAC3G,gBAAgB,SAAS,SAAS;AAAA,QACpC,EAAE,CAAC;AAAA,MACL;AACA,YAAM,GAAG,OAAc,gBAAS,EAAE,IAAI;AAAA,QACpC,QAAQ;AAAA,QACR,QAAQ,SAAS,OAAO;AAAA,QACxB,YAAY,OAAO;AAAA,QACnB,OAAO;AAAA,QACP,WAAW,oBAAI,KAAK;AAAA,MACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC;AAAA,IAClG,CAAC,CAAC;AAAA,EACJ,SAAS,OAAO;AACd,UAAM,CAAC,MAAM,IAAI,MAAM;AAAA,MAAiB;AAAA,MAAI;AAAA,MAAa,OAAO,aAC9D,MAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,QAC1C,QAAQ;AAAA,QACR,OAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAAA,QAC5D,WAAW,oBAAI,KAAK;AAAA,MACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,UAAU;AAAA,IAC9G;AACA,QAAI,CAAC,OAAQ,OAAM;AACnB,WAAO,YAAY,MAAM;AAAA,EAC3B;AACA,QAAM,UAAU,MAAM,YAAY,IAAI,aAAa,UAAU;AAC7D,MAAI,CAAC,QAAS,OAAM,IAAI,MAAM,wCAAwC,UAAU,EAAE;AAClF,SAAO;AACT;AAEA,eAAsB,gBACpB,IACA,OACA,WAA+C,uBAAuB,GACrC;AACjC,QAAM,QAAQ,KAAK,IAAI,KAAK,IAAI,MAAM,SAAS,GAAG,CAAC,GAAG,EAAE;AACxD,QAAM,iBAAiB,MAAM,SAAS,SAAS,WAAW,MAAM,KAAK;AACrE,oBAAkB,gBAAgB,SAAS,SAAS,YAAY,SAAS,SAAS,KAAK;AACvF,QAAM,WAAW,MAAqB,sBAAe,SAAS,QAAQ,cAAc,cAAc,CAAC;AACnG,QAAM,OAAO,MAAM;AAAA,IAAiB;AAAA,IAAI,MAAM;AAAA,IAAa,OAAO,aAChE,MAAM,SAAS,OAAO;AAAA,MACpB,SAAgB,sBAAe;AAAA,MAC/B,YAAmB,sBAAe;AAAA,MAClC,QAAe,sBAAe;AAAA,MAC9B,QAAe,sBAAe;AAAA,MAC9B,OAAc,iBAAU;AAAA,MACxB,MAAa,sBAAe;AAAA,MAC5B,YAAmB,sBAAe;AAAA,MAClC,UAAiB,sBAAe;AAAA,MAChC;AAAA,IACF,CAAC,EAAE,KAAY,qBAAc,EAC1B,UAAiB,kBAAW,GAAU,sBAAe,YAAmB,iBAAU,EAAE,CAAC,EACrF,MAAM;AAAA,MACL,GAAU,iBAAU,QAAQ,OAAO;AAAA,MACnC,GAAU,sBAAe,aAAa,MAAM,WAAW;AAAA,MACvD,GAAU,sBAAe,gBAAgB,SAAS,SAAS,KAAK;AAAA,MAChE,MAAM,WAAW,MAAM,QAAQ,SAAS,IAAI,QAAe,sBAAe,QAAQ,MAAM,OAAO,IAAI;AAAA,IACrG,CAAC,EACA,QAAQ,QAAQ,EAChB,MAAM,KAAK;AAAA,EAChB;AACA,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB,SAAS,IAAI;AAAA,IACb,aAAa,MAAM;AAAA,IACnB,YAAY,IAAI;AAAA,IAChB,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,OAAO,IAAI;AAAA,IACX,MAAM,IAAI;AAAA,IACV,OAAO,KAAK,IAAI,OAAO,IAAI,QAAQ;AAAA,IACnC,YAAY,IAAI;AAAA,IAChB,UAAU,IAAI;AAAA,EAChB,EAAE;AACJ;AAEA,eAAsB,iBAAiB,IAAc,aAAqB,SAAuD;AAC/H,QAAM,CAAC,GAAG,IAAI,MAAM;AAAA,IAAiB;AAAA,IAAI;AAAA,IAAa,OAAO,aAC3D,MAAM,SAAS,OAAO;AAAA,MACpB,SAAgB,sBAAe;AAAA,MAC/B,YAAmB,sBAAe;AAAA,MAClC,QAAe,sBAAe;AAAA,MAC9B,QAAe,sBAAe;AAAA,MAC9B,OAAc,iBAAU;AAAA,MACxB,MAAa,sBAAe;AAAA,MAC5B,YAAmB,sBAAe;AAAA,MAClC,UAAiB,sBAAe;AAAA,IAClC,CAAC,EAAE,KAAY,qBAAc,EAC1B,UAAiB,kBAAW,GAAU,sBAAe,YAAmB,iBAAU,EAAE,CAAC,EACrF,MAAM,IAAI,GAAU,sBAAe,aAAa,WAAW,GAAG,GAAU,sBAAe,IAAI,OAAO,GAAG,GAAU,iBAAU,QAAQ,OAAO,CAAC,CAAC,EAC1I,MAAM,CAAC;AAAA,EACZ;AACA,MAAI,CAAC,IAAK,QAAO;AACjB,SAAO;AAAA,IACL,SAAS,IAAI;AAAA,IACb;AAAA,IACA,YAAY,IAAI;AAAA,IAChB,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,OAAO,IAAI;AAAA,IACX,MAAM,IAAI;AAAA,IACV,OAAO;AAAA,IACP,YAAY,IAAI;AAAA,IAChB,UAAU,IAAI;AAAA,EAChB;AACF;AAEA,eAAsB,mBAAmB,OAAmB,MAAiB,SAAyB,IAAI,wBAAwB,GAA4B;AAC5J,SAAO,MAAM,OAAO,MAAM,OAAO,IAAI;AACvC;AAEO,SAAS,UAAU,MAAc,WAAW,6BAA6B,eAAe,gCAA0C;AACvI,MAAI,gBAAgB,UAAU;AAC5B,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,aAAa,KAAK,QAAQ,SAAS,IAAI,EAAE,QAAQ,WAAW,GAAG,EAAE,KAAK;AAC5E,MAAI,CAAC,WAAY,QAAO,CAAC;AACzB,QAAM,aAAa,WAAW,MAAM,QAAQ,EAAE,IAAI,CAAC,SAAS,KAAK,QAAQ,QAAQ,GAAG,EAAE,KAAK,CAAC,EAAE,OAAO,OAAO;AAC5G,QAAM,SAAmB,CAAC;AAC1B,MAAI,UAAU;AACd,aAAW,aAAa,WAAW,SAAS,IAAI,aAAa,CAAC,WAAW,QAAQ,QAAQ,GAAG,CAAC,GAAG;AAC9F,eAAW,QAAQ,mBAAmB,WAAW,QAAQ,GAAG;AAC1D,UAAI,CAAC,SAAS;AACZ,kBAAU;AAAA,MACZ,WAAW,QAAQ,SAAS,IAAI,KAAK,UAAU,UAAU;AACvD,kBAAU,GAAG,OAAO,IAAI,IAAI;AAAA,MAC9B,OAAO;AACL,eAAO,KAAK,OAAO;AACnB,kBAAU,YAAY,SAAS,cAAc,MAAM,QAAQ;AAAA,MAC7D;AAAA,IACF;AAAA,EACF;AACA,MAAI,QAAS,QAAO,KAAK,OAAO;AAChC,SAAO,OAAO,IAAI,CAAC,UAAU,MAAM,KAAK,CAAC,EAAE,OAAO,OAAO;AAC3D;AAEO,SAAS,uBAAuB,MAAc,aAAa,uCAAiD;AACjH,QAAM,SAAS,IAAI,MAAM,UAAU,EAAE,KAAK,CAAC;AAC3C,QAAM,SAAS,KAAK,YAAY,EAAE,MAAM,mBAAmB,KAAK,CAAC;AACjE,aAAW,SAAS,QAAQ;AAC1B,QAAI,OAAO;AACX,eAAW,QAAQ,OAAO;AACxB,cAAQ,KAAK,YAAY,CAAC,KAAK;AAC/B,aAAO,KAAK,KAAK,MAAM,QAAQ;AAAA,IACjC;AACA,WAAO,KAAK,IAAI,IAAI,IAAI,UAAU,KAAK;AAAA,EACzC;AACA,QAAM,OAAO,KAAK,MAAM,GAAG,MAAM,KAAK;AACtC,SAAO,OAAO,IAAI,CAAC,UAAU,QAAQ,QAAQ,MAAM,QAAQ,CAAC,CAAC,CAAC;AAChE;AAEA,eAAe,iBAAiB,IAAc,aAAqB,QAAoC;AACrG,QAAM,OAAO,MAAM,YAAY,IAAI,aAAa,MAAM;AACtD,MAAI,KAAK,WAAW,SAAS;AAC3B,UAAM,IAAI,MAAM,QAAQ,MAAM,OAAO,KAAK,MAAM,EAAE;AAAA,EACpD;AACA,SAAO;AACT;AAEA,SAAS,kBAAkB,QAAkB,YAAoB,OAAyB;AACxF,MAAI,OAAO,WAAW,YAAY;AAChC,UAAM,IAAI,MAAM,mBAAmB,KAAK,aAAa,OAAO,MAAM,yBAAyB,UAAU,EAAE;AAAA,EACzG;AACA,MAAI,OAAO,KAAK,CAAC,UAAU,CAAC,OAAO,SAAS,KAAK,CAAC,GAAG;AACnD,UAAM,IAAI,MAAM,mBAAmB,KAAK,6BAA6B;AAAA,EACvE;AACA,SAAO;AACT;AAEA,SAAS,WAAW,MAA0B;AAC5C,QAAM,cAAc,KAAK,YAAY,YAAY;AACjD,QAAM,WAAW,KAAK,SAAS,YAAY;AAC3C,SAAO,YAAY,WAAW,OAAO,KAChC,gBAAgB,sBAChB,gBAAgB,qBAChB,gBAAgB,wBAChB,SAAS,SAAS,KAAK,KACvB,SAAS,SAAS,WAAW,KAC7B,SAAS,SAAS,OAAO,KACzB,SAAS,SAAS,OAAO,KACzB,SAAS,SAAS,MAAM,KACxB,SAAS,SAAS,MAAM,KACxB,SAAS,SAAS,MAAM,KACxB,SAAS,SAAS,MAAM;AAC/B;AAEA,SAAS,mBAAmB,MAAc,UAA4B;AACpE,MAAI,KAAK,UAAU,SAAU,QAAO,CAAC,IAAI;AACzC,QAAM,MAAgB,CAAC;AACvB,MAAI,YAAY;AAChB,SAAO,UAAU,SAAS,UAAU;AAClC,UAAM,SAAS,UAAU,MAAM,GAAG,WAAW,CAAC;AAC9C,UAAM,UAAU,KAAK;AAAA,MACnB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,GAAG;AAAA,IACxB;AACA,UAAM,MAAM,UAAU,KAAK,MAAM,WAAW,GAAG,IAAI,UAAU,IAAI;AACjE,QAAI,KAAK,UAAU,MAAM,GAAG,GAAG,EAAE,KAAK,CAAC;AACvC,gBAAY,UAAU,MAAM,GAAG,EAAE,KAAK;AAAA,EACxC;AACA,MAAI,UAAW,KAAI,KAAK,SAAS;AACjC,SAAO;AACT;AAEA,SAAS,YAAY,UAAkB,cAAsB,MAAc,UAA0B;AACnG,MAAI,gBAAgB,EAAG,QAAO;AAC9B,QAAM,UAAU,SAAS,MAAM,KAAK,IAAI,GAAG,SAAS,SAAS,YAAY,CAAC,EAAE,QAAQ,WAAW,EAAE,EAAE,KAAK;AACxG,QAAM,YAAY,UAAU,GAAG,OAAO,IAAI,IAAI,KAAK;AACnD,SAAO,UAAU,UAAU,WAAW,YAAY;AACpD;AAEA,SAAS,cAAc,QAA0B;AAC/C,SAAO,IAAI,OAAO,KAAK,GAAG,CAAC;AAC7B;AAEA,SAAS,gBAAgB,KAA6D;AACpF,SAAO;AAAA,IACL,IAAI,IAAI;AAAA,IACR,aAAa,IAAI;AAAA,IACjB,MAAM,IAAI;AAAA,IACV,aAAa,IAAI;AAAA,IACjB,WAAW,IAAI,UAAU,YAAY;AAAA,IACrC,WAAW,IAAI,UAAU,YAAY;AAAA,EACvC;AACF;AAEA,SAAS,YAAY,KAAqD;AACxE,SAAO;AAAA,IACL,IAAI,IAAI;AAAA,IACR,aAAa,IAAI;AAAA,IACjB,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,OAAO,IAAI;AAAA,IACX,QAAQ,IAAI;AAAA,IACZ,YAAY,IAAI;AAAA,IAChB,OAAO,IAAI;AAAA,IACX,WAAW,IAAI,UAAU,YAAY;AAAA,IACrC,WAAW,IAAI,UAAU,YAAY;AAAA,EACvC;AACF;","names":[]}
+{"version":3,"sources":["../src/index.ts"],"sourcesContent":["import type { Settings } from \"@opengeni/config\";\nimport type {\n  AddDocumentRequest,\n  CreateDocumentBaseRequest,\n  Document,\n  DocumentBase,\n  DocumentSearchMode,\n  DocumentSearchResult,\n  FileAsset,\n  KnowledgeSourceKind,\n} from \"@opengeni/contracts\";\nimport { requireFile, withRlsContext, withWorkspaceRls, type Database } from \"@opengeni/db\";\nimport * as schema from \"@opengeni/db/schema\";\nimport type { ObjectStorage } from \"@opengeni/storage\";\nimport { LiteParse } from \"@llamaindex/liteparse\";\nimport { and, asc, desc, eq, inArray, sql, type SQL } from \"drizzle-orm\";\nimport OpenAI from \"openai\";\n\nexport const DEFAULT_DOCUMENT_PARSER = \"liteparse\";\nexport const DEFAULT_DOCUMENT_EMBEDDING_MODEL = \"text-embedding-3-large\";\nexport const DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS = 3072;\nexport const DEFAULT_DOCUMENT_CHUNK_SIZE = 1200;\nexport const DEFAULT_DOCUMENT_CHUNK_OVERLAP = 160;\n\nexport type ParsedDocument = {\n  text: string;\n  metadata?: Record<string, unknown>;\n};\n\nexport type DocumentChunk = {\n  text: string;\n  metadata: Record<string, unknown>;\n};\n\nexport type DocumentParser = {\n  name: string;\n  parse: (bytes: Uint8Array, file: FileAsset) => Promise<ParsedDocument>;\n};\n\nexport type DocumentChunker = {\n  chunk: (parsed: ParsedDocument, file: FileAsset) => DocumentChunk[];\n};\n\nexport type DocumentEmbedder = {\n  model: string;\n  dimensions: number;\n  embedMany: (texts: string[]) => Promise<number[][]>;\n  embedQuery: (text: string) => Promise<number[]>;\n};\n\nexport type DocumentServices = {\n  parser: DocumentParser;\n  chunker: DocumentChunker;\n  embedder: DocumentEmbedder;\n};\n\nexport type DocumentSearchInput = {\n  workspaceId: string;\n  query: string;\n  baseIds?: string[] | undefined;\n  limit?: number | undefined;\n  mode?: DocumentSearchMode | undefined;\n  sourceKinds?: KnowledgeSourceKind[] | undefined;\n  aclTags?: string[] | undefined;\n};\n\nexport type DocumentIndexHooks = {\n  beforeEmbed?: (input: {\n    accountId: string;\n    workspaceId: string;\n    documentId: string;\n    chunkCount: number;\n  }) => Promise<void>;\n};\n\nexport class LiteParseDocumentParser implements DocumentParser {\n  readonly name = DEFAULT_DOCUMENT_PARSER;\n  private parseQueue: Promise<void> = Promise.resolve();\n\n  async parse(bytes: Uint8Array, file: FileAsset): Promise<ParsedDocument> {\n    const text = isTextLike(file)\n      ? Buffer.from(bytes).toString(\"utf8\").replace(/\\0/g, \" \").trim()\n      : await this.parseWithLiteParse(bytes);\n    if (!text.trim()) {\n      throw new Error(`Parsed document is empty: ${file.filename}`);\n    }\n    return {\n      text: text.trim(),\n      metadata: {\n        parser: this.name,\n        filename: file.filename,\n        contentType: file.contentType,\n      },\n    };\n  }\n\n  private async parseWithLiteParse(bytes: Uint8Array): Promise<string> {\n    return await this.enqueueParse(async () => {\n      const parser = new LiteParse({ ocrEnabled: true, numWorkers: 1 });\n      const result = await parser.parse(Buffer.from(bytes), true);\n      const text = typeof result?.text === \"string\" ? result.text : \"\";\n      return text.replace(/\\0/g, \" \").trim();\n    });\n  }\n\n  private async enqueueParse<T>(task: () => Promise<T>): Promise<T> {\n    const previous = this.parseQueue;\n    let release: () => void = () => undefined;\n    this.parseQueue = new Promise<void>((resolve) => {\n      release = resolve;\n    });\n    await previous.catch(() => undefined);\n    try {\n      return await task();\n    } finally {\n      release();\n    }\n  }\n}\n\nexport class RecursiveTextChunker implements DocumentChunker {\n  constructor(\n    private readonly maxChars = DEFAULT_DOCUMENT_CHUNK_SIZE,\n    private readonly overlapChars = DEFAULT_DOCUMENT_CHUNK_OVERLAP,\n  ) {\n    if (overlapChars >= maxChars) {\n      throw new Error(\"document chunk overlap must be smaller than chunk size\");\n    }\n  }\n\n  chunk(parsed: ParsedDocument, file: FileAsset): DocumentChunk[] {\n    return chunkText(parsed.text, this.maxChars, this.overlapChars).map((text, index) => ({\n      text,\n      metadata: {\n        ...parsed.metadata,\n        filename: file.filename,\n        contentType: file.contentType,\n        chunkIndex: index,\n      },\n    }));\n  }\n}\n\nexport class OpenAIEmbeddingProvider implements DocumentEmbedder {\n  private client: OpenAI | null = null;\n  private readonly apiKey: string | undefined;\n  private readonly baseURL: string | undefined;\n\n  constructor(args: {\n    apiKey?: string | undefined;\n    baseURL?: string | undefined;\n    defaultHeaders?: Record<string, string> | undefined;\n    defaultQuery?: Record<string, string> | undefined;\n    model?: string | undefined;\n    dimensions?: number | undefined;\n  }) {\n    this.apiKey = args.apiKey ?? process.env.OPENAI_API_KEY;\n    this.baseURL = args.baseURL;\n    this.defaultHeaders = args.defaultHeaders;\n    this.defaultQuery = args.defaultQuery;\n    this.model = args.model ?? DEFAULT_DOCUMENT_EMBEDDING_MODEL;\n    this.dimensions = args.dimensions ?? DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS;\n  }\n\n  readonly model: string;\n  readonly dimensions: number;\n  private readonly defaultHeaders: Record<string, string> | undefined;\n  private readonly defaultQuery: Record<string, string> | undefined;\n\n  async embedMany(texts: string[]): Promise<number[][]> {\n    if (texts.length === 0) return [];\n    const out: number[][] = [];\n    for (let start = 0; start < texts.length; start += 64) {\n      const batch = texts.slice(start, start + 64);\n      const response = await this.openai().embeddings.create({\n        model: this.model,\n        input: batch,\n        dimensions: this.dimensions,\n      });\n      for (const item of response.data) {\n        out.push(validateEmbedding(item.embedding, this.dimensions, this.model));\n      }\n    }\n    return out;\n  }\n\n  async embedQuery(text: string): Promise<number[]> {\n    const [embedding] = await this.embedMany([text]);\n    if (!embedding) {\n      throw new Error(\"Embedding provider returned no query embedding\");\n    }\n    return embedding;\n  }\n\n  private openai(): OpenAI {\n    if (!this.apiKey) {\n      throw new Error(\"OpenAI document embeddings require an API key\");\n    }\n    this.client ??= new OpenAI({\n      apiKey: this.apiKey,\n      ...(this.baseURL ? { baseURL: this.baseURL } : {}),\n      ...(this.defaultQuery ? { defaultQuery: this.defaultQuery } : {}),\n      ...(this.defaultHeaders ? { defaultHeaders: this.defaultHeaders } : {}),\n    });\n    return this.client;\n  }\n}\n\nexport class DeterministicEmbeddingProvider implements DocumentEmbedder {\n  constructor(\n    readonly dimensions = DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS,\n    readonly model = `deterministic-local-${dimensions}`,\n  ) {}\n\n  async embedMany(texts: string[]): Promise<number[][]> {\n    return texts.map((text) => deterministicEmbedding(text, this.dimensions));\n  }\n\n  async embedQuery(text: string): Promise<number[]> {\n    return deterministicEmbedding(text, this.dimensions);\n  }\n}\n\nexport function createDocumentServices(settings?: Settings, overrides: Partial<DocumentServices> = {}): DocumentServices {\n  const dimensions = settings?.documentEmbeddingDimensions ?? DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS;\n  const openAIEmbeddingConfig = documentOpenAIEmbeddingConfig(settings);\n  return {\n    parser: overrides.parser ?? new LiteParseDocumentParser(),\n    chunker: overrides.chunker ?? new RecursiveTextChunker(\n      settings?.documentChunkSize ?? DEFAULT_DOCUMENT_CHUNK_SIZE,\n      settings?.documentChunkOverlap ?? DEFAULT_DOCUMENT_CHUNK_OVERLAP,\n    ),\n    embedder: overrides.embedder ?? (\n      settings?.documentEmbeddingProvider === \"deterministic\"\n        ? new DeterministicEmbeddingProvider(dimensions, settings.documentEmbeddingModel)\n        : new OpenAIEmbeddingProvider({\n          ...openAIEmbeddingConfig,\n          model: settings?.documentEmbeddingModel ?? DEFAULT_DOCUMENT_EMBEDDING_MODEL,\n          dimensions,\n        })\n    ),\n  };\n}\n\nexport function documentOpenAIEmbeddingConfig(settings?: Settings): {\n  apiKey?: string | undefined;\n  baseURL?: string | undefined;\n  defaultHeaders?: Record<string, string> | undefined;\n  defaultQuery?: Record<string, string> | undefined;\n} {\n  if (!settings) return {};\n  if (settings.documentEmbeddingApiKey || settings.documentEmbeddingBaseUrl) {\n    return {\n      apiKey: settings.documentEmbeddingApiKey ?? settings.openaiApiKey ?? settings.azureOpenaiApiKey,\n      baseURL: settings.documentEmbeddingBaseUrl ?? settings.openaiBaseUrl ?? settings.azureOpenaiBaseUrl,\n    };\n  }\n  if (settings.openaiProvider === \"azure\") {\n    const baseURL = settings.azureOpenaiBaseUrl ?? azureDeploymentBaseUrl(settings);\n    return {\n      apiKey: settings.azureOpenaiApiKey ?? settings.azureOpenaiAdToken ?? \"azure-ad-token\",\n      baseURL,\n      defaultQuery: azureOpenAIDefaultQuery(settings, baseURL),\n      defaultHeaders: settings.azureOpenaiAdToken && !settings.azureOpenaiApiKey\n        ? { Authorization: `Bearer ${settings.azureOpenaiAdToken}` }\n        : undefined,\n    };\n  }\n  return {\n    apiKey: settings.openaiApiKey,\n    baseURL: settings.openaiBaseUrl,\n  };\n}\n\nfunction azureDeploymentBaseUrl(settings: Settings): string {\n  const endpoint = settings.azureOpenaiEndpoint?.replace(/\\/+$/, \"\");\n  if (!endpoint || !settings.azureOpenaiDeployment) {\n    throw new Error(\"Azure OpenAI endpoint/deployment settings are incomplete\");\n  }\n  return `${endpoint}/openai/deployments/${settings.azureOpenaiDeployment}`;\n}\n\nfunction azureOpenAIDefaultQuery(\n  settings: Pick<Settings, \"azureOpenaiApiVersion\">,\n  baseURL: string,\n): Record<string, string> | undefined {\n  if (!settings.azureOpenaiApiVersion) return undefined;\n  const normalized = baseURL.replace(/\\/+$/, \"\").toLowerCase();\n  if (normalized.endsWith(\"/openai/v1\")) {\n    return undefined;\n  }\n  return { \"api-version\": settings.azureOpenaiApiVersion };\n}\n\nexport async function createDocumentBase(db: Database, input: CreateDocumentBaseRequest & { accountId: string; workspaceId: string }): Promise<DocumentBase> {\n  return await withRlsContext(db, { accountId: input.accountId, workspaceId: input.workspaceId }, async (scopedDb) => {\n    const [row] = await scopedDb.insert(schema.documentBases).values({\n      accountId: input.accountId,\n      workspaceId: input.workspaceId,\n      name: input.name.trim(),\n      description: input.description?.trim() || null,\n    }).returning();\n    if (!row) throw new Error(\"Failed to create document base\");\n    return mapDocumentBase(row);\n  });\n}\n\nexport async function listDocumentBases(db: Database, workspaceId: string): Promise<DocumentBase[]> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const rows = await scopedDb.select().from(schema.documentBases)\n      .where(eq(schema.documentBases.workspaceId, workspaceId))\n      .orderBy(desc(schema.documentBases.createdAt));\n    return rows.map(mapDocumentBase);\n  });\n}\n\nexport async function getDocumentBase(db: Database, workspaceId: string, baseId: string): Promise<DocumentBase | null> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const [row] = await scopedDb.select().from(schema.documentBases).where(and(eq(schema.documentBases.workspaceId, workspaceId), eq(schema.documentBases.id, baseId))).limit(1);\n    return row ? mapDocumentBase(row) : null;\n  });\n}\n\nexport async function addDocumentToBase(db: Database, input: AddDocumentRequest & { accountId: string; workspaceId: string; baseId: string }): Promise<Document> {\n  return await withRlsContext(db, { accountId: input.accountId, workspaceId: input.workspaceId }, async (scopedDb) => {\n    const base = await getDocumentBase(scopedDb, input.workspaceId, input.baseId);\n    if (!base) throw new Error(`Document base not found: ${input.baseId}`);\n    const file = await requireReadyFile(scopedDb, input.workspaceId, input.fileId);\n    const now = new Date();\n    const [existing] = await scopedDb.select().from(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.baseId, input.baseId), eq(schema.documents.fileId, input.fileId)))\n      .limit(1);\n    if (existing) {\n      // Idempotent re-add: refresh caller-supplied source metadata on the\n      // existing row instead of silently discarding it (aclTags especially —\n      // a re-add that tightens tags must not be a no-op).\n      const [updated] = await scopedDb.update(schema.documents).set({\n        title: cleanString(input.title) ?? cleanString(input.sourceTitle) ?? existing.title,\n        ...(input.sourceKind !== undefined ? { sourceKind: input.sourceKind } : {}),\n        sourceUri: cleanString(input.sourceUri) ?? existing.sourceUri,\n        sourceExternalId: cleanString(input.sourceExternalId) ?? existing.sourceExternalId,\n        sourceTitle: cleanString(input.sourceTitle) ?? existing.sourceTitle,\n        sourceAuthor: cleanString(input.sourceAuthor) ?? existing.sourceAuthor,\n        sourceCreatedAt: parseOptionalDate(input.sourceCreatedAt) ?? existing.sourceCreatedAt,\n        sourceUpdatedAt: parseOptionalDate(input.sourceUpdatedAt) ?? existing.sourceUpdatedAt,\n        sourceVersion: cleanString(input.sourceVersion) ?? existing.sourceVersion,\n        ...(input.aclTags !== undefined ? { aclTags: cleanStringArray(input.aclTags) } : {}),\n        updatedAt: now,\n      }).where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, existing.id))).returning();\n      return mapDocument(updated ?? existing);\n    }\n    const [row] = await scopedDb.insert(schema.documents).values({\n      accountId: input.accountId,\n      workspaceId: input.workspaceId,\n      baseId: input.baseId,\n      fileId: input.fileId,\n      status: \"queued\",\n      title: cleanString(input.title) ?? cleanString(input.sourceTitle) ?? file.filename,\n      parser: DEFAULT_DOCUMENT_PARSER,\n      sourceKind: input.sourceKind ?? \"manual_upload\",\n      sourceUri: cleanString(input.sourceUri) ?? null,\n      sourceExternalId: cleanString(input.sourceExternalId) ?? null,\n      sourceTitle: cleanString(input.sourceTitle) ?? null,\n      sourceAuthor: cleanString(input.sourceAuthor) ?? null,\n      sourceCreatedAt: parseOptionalDate(input.sourceCreatedAt),\n      sourceUpdatedAt: parseOptionalDate(input.sourceUpdatedAt),\n      sourceVersion: cleanString(input.sourceVersion) ?? null,\n      aclTags: cleanStringArray(input.aclTags),\n      updatedAt: now,\n    }).returning();\n    if (!row) throw new Error(\"Failed to create document\");\n    return mapDocument(row);\n  });\n}\n\nexport async function deleteDocumentFromBase(\n  db: Database,\n  input: { accountId: string; workspaceId: string; baseId: string; documentId: string },\n): Promise<void> {\n  await withRlsContext(db, { accountId: input.accountId, workspaceId: input.workspaceId }, async (scopedDb) => {\n    const [document] = await scopedDb.select().from(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, input.documentId)))\n      .limit(1);\n    if (!document) {\n      throw new Error(`Document not found: ${input.documentId}`);\n    }\n    if (document.baseId !== input.baseId) {\n      throw new Error(`Document not found: ${input.documentId}`);\n    }\n    await scopedDb.delete(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, input.documentId)));\n  });\n}\n\nexport async function listDocuments(db: Database, workspaceId: string, baseId: string): Promise<Document[]> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const rows = await scopedDb.select().from(schema.documents)\n      .where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.baseId, baseId)))\n      .orderBy(asc(schema.documents.createdAt));\n    return rows.map(mapDocument);\n  });\n}\n\nexport async function getDocument(db: Database, workspaceId: string, documentId: string): Promise<Document | null> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const [row] = await scopedDb.select().from(schema.documents).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).limit(1);\n    return row ? mapDocument(row) : null;\n  });\n}\n\nexport async function queueDocumentForReindex(db: Database, workspaceId: string, documentId: string): Promise<Document> {\n  return await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    const [row] = await scopedDb.update(schema.documents).set({\n      status: \"queued\",\n      error: null,\n      updatedAt: new Date(),\n    }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).returning();\n    if (!row) throw new Error(`Document not found: ${documentId}`);\n    return mapDocument(row);\n  });\n}\n\nexport async function indexDocumentNow(\n  db: Database,\n  objectStorage: ObjectStorage,\n  workspaceId: string,\n  documentId: string,\n  services: DocumentServices = createDocumentServices(),\n  hooks: DocumentIndexHooks = {},\n): Promise<Document> {\n  const [document] = await withWorkspaceRls(db, workspaceId, async (scopedDb) =>\n    await scopedDb.select().from(schema.documents).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).limit(1)\n  );\n  if (!document) throw new Error(`Document not found: ${documentId}`);\n  const file = await requireReadyFile(db, workspaceId, document.fileId);\n  await withWorkspaceRls(db, workspaceId, async (scopedDb) => {\n    await scopedDb.update(schema.documents).set({\n      status: \"indexing\",\n      parser: services.parser.name,\n      error: null,\n      updatedAt: new Date(),\n    }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId)));\n  });\n  try {\n    const bytes = await objectStorage.getFileBytes(file);\n    const parsed = await services.parser.parse(bytes, file);\n    const chunks = services.chunker.chunk(parsed, file);\n    await hooks.beforeEmbed?.({\n      accountId: document.accountId,\n      workspaceId: document.workspaceId,\n      documentId,\n      chunkCount: chunks.length,\n    });\n    const embeddings = await services.embedder.embedMany(chunks.map((chunk) => chunk.text));\n    if (embeddings.length !== chunks.length) {\n      throw new Error(`Embedding provider returned ${embeddings.length} embeddings for ${chunks.length} chunks`);\n    }\n    await withWorkspaceRls(db, workspaceId, async (scopedDb) => await scopedDb.transaction(async (tx) => {\n      await tx.delete(schema.documentChunks).where(and(eq(schema.documentChunks.workspaceId, workspaceId), eq(schema.documentChunks.documentId, documentId)));\n      if (chunks.length > 0) {\n        await tx.insert(schema.documentChunks).values(chunks.map((chunk, index) => ({\n          accountId: document.accountId,\n          workspaceId: document.workspaceId,\n          documentId,\n          baseId: document.baseId,\n          fileId: file.id,\n          chunkIndex: index,\n          text: chunk.text,\n          metadata: {\n            ...chunk.metadata,\n            documentTitle: document.title,\n            sourceKind: document.sourceKind,\n            sourceUri: document.sourceUri,\n            sourceExternalId: document.sourceExternalId,\n            sourceTitle: document.sourceTitle,\n            sourceAuthor: document.sourceAuthor,\n            sourceCreatedAt: document.sourceCreatedAt?.toISOString() ?? null,\n            sourceUpdatedAt: document.sourceUpdatedAt?.toISOString() ?? null,\n            sourceVersion: document.sourceVersion,\n            aclTags: document.aclTags,\n          },\n          embedding: validateEmbedding(embeddings[index] ?? [], services.embedder.dimensions, services.embedder.model),\n          embeddingModel: services.embedder.model,\n        })));\n      }\n      await tx.update(schema.documents).set({\n        status: \"ready\",\n        parser: services.parser.name,\n        chunkCount: chunks.length,\n        error: null,\n        updatedAt: new Date(),\n      }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId)));\n    }));\n  } catch (error) {\n    const [failed] = await withWorkspaceRls(db, workspaceId, async (scopedDb) =>\n      await scopedDb.update(schema.documents).set({\n        status: \"failed\",\n        error: error instanceof Error ? error.message : String(error),\n        updatedAt: new Date(),\n      }).where(and(eq(schema.documents.workspaceId, workspaceId), eq(schema.documents.id, documentId))).returning()\n    );\n    if (!failed) throw error;\n    return mapDocument(failed);\n  }\n  const updated = await getDocument(db, workspaceId, documentId);\n  if (!updated) throw new Error(`Document disappeared after indexing: ${documentId}`);\n  return updated;\n}\n\nexport async function searchDocuments(\n  db: Database,\n  input: DocumentSearchInput,\n  services: Pick<DocumentServices, \"embedder\"> = createDocumentServices(),\n): Promise<DocumentSearchResult[]> {\n  const mode = input.mode ?? \"hybrid\";\n  const limit = Math.min(Math.max(input.limit ?? 5, 1), 50);\n  const candidateLimit = mode === \"hybrid\" ? Math.min(limit * 4, 100) : limit;\n  const rows: CombinedSearchRow[] = [];\n  if (mode === \"vector\" || mode === \"hybrid\") {\n    try {\n      rows.push(...await vectorSearchDocuments(db, input, candidateLimit, services));\n    } catch (error) {\n      if (mode === \"vector\") {\n        throw error;\n      }\n      console.warn(\"document hybrid search vector component failed; falling back to keyword search\", {\n        workspaceId: input.workspaceId,\n        error: error instanceof Error ? error.message : String(error),\n      });\n    }\n  }\n  if (mode === \"keyword\" || mode === \"hybrid\") {\n    rows.push(...await keywordSearchDocuments(db, input, candidateLimit));\n  }\n  return mergeDocumentSearchRows(rows, mode).slice(0, limit);\n}\n\nasync function vectorSearchDocuments(\n  db: Database,\n  input: DocumentSearchInput,\n  limit: number,\n  services: Pick<DocumentServices, \"embedder\">,\n): Promise<CombinedSearchRow[]> {\n  const queryEmbedding = await services.embedder.embedQuery(input.query);\n  validateEmbedding(queryEmbedding, services.embedder.dimensions, services.embedder.model);\n  const distance = sql<number>`${schema.documentChunks.embedding} <=> ${vectorLiteral(queryEmbedding)}::vector`;\n  const rows = await withWorkspaceRls(db, input.workspaceId, async (scopedDb) =>\n    await scopedDb.select({\n      chunkId: schema.documentChunks.id,\n      documentId: schema.documentChunks.documentId,\n      baseId: schema.documentChunks.baseId,\n      fileId: schema.documentChunks.fileId,\n      title: schema.documents.title,\n      text: schema.documentChunks.text,\n      chunkIndex: schema.documentChunks.chunkIndex,\n      metadata: schema.documentChunks.metadata,\n      sourceKind: schema.documents.sourceKind,\n      sourceUri: schema.documents.sourceUri,\n      sourceExternalId: schema.documents.sourceExternalId,\n      sourceTitle: schema.documents.sourceTitle,\n      sourceAuthor: schema.documents.sourceAuthor,\n      sourceCreatedAt: schema.documents.sourceCreatedAt,\n      sourceUpdatedAt: schema.documents.sourceUpdatedAt,\n      sourceVersion: schema.documents.sourceVersion,\n      aclTags: schema.documents.aclTags,\n      distance,\n    }).from(schema.documentChunks)\n      .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))\n      .where(and(...documentSearchConditions(input, services.embedder.model)))\n      .orderBy(distance)\n      .limit(limit)\n  );\n  return rows.map((row) => ({\n    ...mapSearchRowBase(row, input.workspaceId),\n    vectorScore: 1 / (1 + Number(row.distance)),\n    keywordScore: null,\n  }));\n}\n\nasync function keywordSearchDocuments(db: Database, input: DocumentSearchInput, limit: number): Promise<CombinedSearchRow[]> {\n  const rank = sql<number>`ts_rank_cd(to_tsvector('simple', ${schema.documentChunks.text}), plainto_tsquery('simple', ${input.query}))`;\n  const rows = await withWorkspaceRls(db, input.workspaceId, async (scopedDb) =>\n    await scopedDb.select({\n      chunkId: schema.documentChunks.id,\n      documentId: schema.documentChunks.documentId,\n      baseId: schema.documentChunks.baseId,\n      fileId: schema.documentChunks.fileId,\n      title: schema.documents.title,\n      text: schema.documentChunks.text,\n      chunkIndex: schema.documentChunks.chunkIndex,\n      metadata: schema.documentChunks.metadata,\n      sourceKind: schema.documents.sourceKind,\n      sourceUri: schema.documents.sourceUri,\n      sourceExternalId: schema.documents.sourceExternalId,\n      sourceTitle: schema.documents.sourceTitle,\n      sourceAuthor: schema.documents.sourceAuthor,\n      sourceCreatedAt: schema.documents.sourceCreatedAt,\n      sourceUpdatedAt: schema.documents.sourceUpdatedAt,\n      sourceVersion: schema.documents.sourceVersion,\n      aclTags: schema.documents.aclTags,\n      rank,\n    }).from(schema.documentChunks)\n      .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))\n      .where(and(\n        ...documentSearchConditions(input),\n        sql`to_tsvector('simple', ${schema.documentChunks.text}) @@ plainto_tsquery('simple', ${input.query})`,\n      ))\n      .orderBy(desc(rank))\n      .limit(limit)\n  );\n  return rows.map((row) => ({\n    ...mapSearchRowBase(row, input.workspaceId),\n    vectorScore: null,\n    keywordScore: normalizeKeywordScore(Number(row.rank)),\n  }));\n}\n\nexport async function getDocumentChunk(db: Database, workspaceId: string, chunkId: string): Promise<DocumentSearchResult | null> {\n  const [row] = await withWorkspaceRls(db, workspaceId, async (scopedDb) =>\n    await scopedDb.select({\n      chunkId: schema.documentChunks.id,\n      documentId: schema.documentChunks.documentId,\n      baseId: schema.documentChunks.baseId,\n      fileId: schema.documentChunks.fileId,\n      title: schema.documents.title,\n      text: schema.documentChunks.text,\n      chunkIndex: schema.documentChunks.chunkIndex,\n      metadata: schema.documentChunks.metadata,\n      sourceKind: schema.documents.sourceKind,\n      sourceUri: schema.documents.sourceUri,\n      sourceExternalId: schema.documents.sourceExternalId,\n      sourceTitle: schema.documents.sourceTitle,\n      sourceAuthor: schema.documents.sourceAuthor,\n      sourceCreatedAt: schema.documents.sourceCreatedAt,\n      sourceUpdatedAt: schema.documents.sourceUpdatedAt,\n      sourceVersion: schema.documents.sourceVersion,\n      aclTags: schema.documents.aclTags,\n    }).from(schema.documentChunks)\n      .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))\n      .where(and(eq(schema.documentChunks.workspaceId, workspaceId), eq(schema.documentChunks.id, chunkId), eq(schema.documents.status, \"ready\")))\n      .limit(1)\n  );\n  if (!row) return null;\n  return {\n    ...mapSearchRowBase(row, workspaceId),\n    score: 1,\n    matchType: \"hybrid\",\n    vectorScore: null,\n    keywordScore: null,\n  };\n}\n\ntype SearchRowBase = Omit<DocumentSearchResult, \"score\" | \"matchType\" | \"vectorScore\" | \"keywordScore\">;\ntype CombinedSearchRow = SearchRowBase & {\n  vectorScore: number | null;\n  keywordScore: number | null;\n};\n\nfunction documentSearchConditions(input: DocumentSearchInput, embeddingModel?: string): SQL[] {\n  const conditions: SQL[] = [\n    eq(schema.documents.status, \"ready\"),\n    eq(schema.documentChunks.workspaceId, input.workspaceId),\n  ];\n  if (embeddingModel) {\n    conditions.push(eq(schema.documentChunks.embeddingModel, embeddingModel));\n  }\n  if (input.baseIds?.length) {\n    conditions.push(inArray(schema.documentChunks.baseId, input.baseIds));\n  }\n  if (input.sourceKinds?.length) {\n    conditions.push(inArray(schema.documents.sourceKind, input.sourceKinds));\n  }\n  const aclTags = cleanStringArray(input.aclTags);\n  if (aclTags.length > 0) {\n    conditions.push(sql`${schema.documents.aclTags} @> ${JSON.stringify(aclTags)}::jsonb`);\n  }\n  return conditions;\n}\n\nfunction mapSearchRowBase(row: {\n  chunkId: string;\n  documentId: string;\n  baseId: string;\n  fileId: string;\n  title: string;\n  text: string;\n  chunkIndex: number;\n  metadata: Record<string, unknown>;\n  sourceKind: string;\n  sourceUri: string | null;\n  sourceExternalId: string | null;\n  sourceTitle: string | null;\n  sourceAuthor: string | null;\n  sourceCreatedAt: Date | null;\n  sourceUpdatedAt: Date | null;\n  sourceVersion: string | null;\n  aclTags: string[];\n}, workspaceId: string): SearchRowBase {\n  return {\n    chunkId: row.chunkId,\n    workspaceId,\n    documentId: row.documentId,\n    baseId: row.baseId,\n    fileId: row.fileId,\n    title: row.title,\n    text: row.text,\n    chunkIndex: row.chunkIndex,\n    metadata: row.metadata,\n    sourceKind: normalizeKnowledgeSourceKind(row.sourceKind),\n    sourceUri: row.sourceUri,\n    sourceExternalId: row.sourceExternalId,\n    sourceTitle: row.sourceTitle,\n    sourceAuthor: row.sourceAuthor,\n    sourceCreatedAt: row.sourceCreatedAt?.toISOString() ?? null,\n    sourceUpdatedAt: row.sourceUpdatedAt?.toISOString() ?? null,\n    sourceVersion: row.sourceVersion,\n    aclTags: cleanStringArray(row.aclTags),\n  };\n}\n\nfunction mergeDocumentSearchRows(rows: CombinedSearchRow[], mode: DocumentSearchMode): DocumentSearchResult[] {\n  const byChunk = new Map<string, CombinedSearchRow>();\n  for (const row of rows) {\n    const existing = byChunk.get(row.chunkId);\n    if (!existing) {\n      byChunk.set(row.chunkId, row);\n      continue;\n    }\n    byChunk.set(row.chunkId, {\n      ...existing,\n      vectorScore: Math.max(existing.vectorScore ?? 0, row.vectorScore ?? 0) || null,\n      keywordScore: Math.max(existing.keywordScore ?? 0, row.keywordScore ?? 0) || null,\n    });\n  }\n  return [...byChunk.values()].map((row) => {\n    const vectorScore = row.vectorScore;\n    const keywordScore = row.keywordScore;\n    const matchType: DocumentSearchMode = vectorScore !== null && keywordScore !== null\n      ? \"hybrid\"\n      : vectorScore !== null\n        ? \"vector\"\n        : \"keyword\";\n    return {\n      ...row,\n      score: combinedSearchScore(mode, vectorScore, keywordScore, matchType),\n      matchType,\n    };\n  }).sort((left, right) =>\n    right.score - left.score\n    || (right.vectorScore ?? 0) - (left.vectorScore ?? 0)\n    || (right.keywordScore ?? 0) - (left.keywordScore ?? 0)\n    || left.chunkIndex - right.chunkIndex\n  );\n}\n\nfunction combinedSearchScore(\n  mode: DocumentSearchMode,\n  vectorScore: number | null,\n  keywordScore: number | null,\n  matchType: DocumentSearchMode,\n): number {\n  const vector = vectorScore ?? 0;\n  const keyword = keywordScore ?? 0;\n  if (mode === \"vector\") return roundScore(vector);\n  if (mode === \"keyword\") return roundScore(keyword);\n  return roundScore(Math.min(1, (0.65 * vector) + (0.35 * keyword) + (matchType === \"hybrid\" ? 0.1 : 0)));\n}\n\nfunction normalizeKeywordScore(rank: number): number {\n  if (!Number.isFinite(rank) || rank <= 0) {\n    return 0;\n  }\n  return rank / (rank + 1);\n}\n\nfunction roundScore(value: number): number {\n  return Number(value.toFixed(6));\n}\n\nexport async function parseDocumentBytes(bytes: Uint8Array, file: FileAsset, parser: DocumentParser = new LiteParseDocumentParser()): Promise<ParsedDocument> {\n  return await parser.parse(bytes, file);\n}\n\nexport function chunkText(text: string, maxChars = DEFAULT_DOCUMENT_CHUNK_SIZE, overlapChars = DEFAULT_DOCUMENT_CHUNK_OVERLAP): string[] {\n  if (overlapChars >= maxChars) {\n    throw new Error(\"chunk overlap must be smaller than chunk size\");\n  }\n  const normalized = text.replace(/\\r\\n/g, \"\\n\").replace(/[ \\t]+/g, \" \").trim();\n  if (!normalized) return [];\n  const paragraphs = normalized.split(/\\n{2,}/).map((part) => part.replace(/\\s+/g, \" \").trim()).filter(Boolean);\n  const chunks: string[] = [];\n  let current = \"\";\n  for (const paragraph of paragraphs.length > 0 ? paragraphs : [normalized.replace(/\\s+/g, \" \")]) {\n    for (const part of splitOversizedText(paragraph, maxChars)) {\n      if (!current) {\n        current = part;\n      } else if (current.length + 1 + part.length <= maxChars) {\n        current = `${current} ${part}`;\n      } else {\n        chunks.push(current);\n        current = withOverlap(current, overlapChars, part, maxChars);\n      }\n    }\n  }\n  if (current) chunks.push(current);\n  return chunks.map((chunk) => chunk.trim()).filter(Boolean);\n}\n\nexport function deterministicEmbedding(text: string, dimensions = DEFAULT_DOCUMENT_EMBEDDING_DIMENSIONS): number[] {\n  const values = new Array(dimensions).fill(0);\n  const tokens = text.toLowerCase().match(/[\\p{L}\\p{N}_-]+/gu) ?? [];\n  for (const token of tokens) {\n    let hash = 2166136261;\n    for (const char of token) {\n      hash ^= char.codePointAt(0) ?? 0;\n      hash = Math.imul(hash, 16777619);\n    }\n    values[Math.abs(hash) % dimensions] += 1;\n  }\n  const norm = Math.hypot(...values) || 1;\n  return values.map((value) => Number((value / norm).toFixed(6)));\n}\n\nasync function requireReadyFile(db: Database, workspaceId: string, fileId: string): Promise<FileAsset> {\n  const file = await requireFile(db, workspaceId, fileId);\n  if (file.status !== \"ready\") {\n    throw new Error(`File ${fileId} is ${file.status}`);\n  }\n  return file;\n}\n\nfunction validateEmbedding(values: number[], dimensions: number, model: string): number[] {\n  if (values.length !== dimensions) {\n    throw new Error(`Embedding model ${model} returned ${values.length} dimensions; expected ${dimensions}`);\n  }\n  if (values.some((value) => !Number.isFinite(value))) {\n    throw new Error(`Embedding model ${model} returned non-finite values`);\n  }\n  return values;\n}\n\nfunction isTextLike(file: FileAsset): boolean {\n  const contentType = file.contentType.toLowerCase();\n  const filename = file.filename.toLowerCase();\n  return contentType.startsWith(\"text/\")\n    || contentType === \"application/json\"\n    || contentType === \"application/xml\"\n    || contentType === \"application/x-yaml\"\n    || filename.endsWith(\".md\")\n    || filename.endsWith(\".markdown\")\n    || filename.endsWith(\".json\")\n    || filename.endsWith(\".yaml\")\n    || filename.endsWith(\".yml\")\n    || filename.endsWith(\".csv\")\n    || filename.endsWith(\".tsv\")\n    || filename.endsWith(\".xml\");\n}\n\nfunction splitOversizedText(text: string, maxChars: number): string[] {\n  if (text.length <= maxChars) return [text];\n  const out: string[] = [];\n  let remaining = text;\n  while (remaining.length > maxChars) {\n    const window = remaining.slice(0, maxChars + 1);\n    const breakAt = Math.max(\n      window.lastIndexOf(\". \"),\n      window.lastIndexOf(\"? \"),\n      window.lastIndexOf(\"! \"),\n      window.lastIndexOf(\"; \"),\n      window.lastIndexOf(\", \"),\n      window.lastIndexOf(\" \"),\n    );\n    const end = breakAt > Math.floor(maxChars * 0.5) ? breakAt + 1 : maxChars;\n    out.push(remaining.slice(0, end).trim());\n    remaining = remaining.slice(end).trim();\n  }\n  if (remaining) out.push(remaining);\n  return out;\n}\n\nfunction withOverlap(previous: string, overlapChars: number, next: string, maxChars: number): string {\n  if (overlapChars <= 0) return next;\n  const overlap = previous.slice(Math.max(0, previous.length - overlapChars)).replace(/^\\S+\\s+/, \"\").trim();\n  const candidate = overlap ? `${overlap} ${next}` : next;\n  return candidate.length <= maxChars ? candidate : next;\n}\n\nfunction cleanString(value: string | undefined | null): string | undefined {\n  const trimmed = value?.trim();\n  return trimmed ? trimmed : undefined;\n}\n\nfunction cleanStringArray(values: string[] | undefined | null): string[] {\n  return [...new Set((values ?? []).map((value) => value.trim()).filter(Boolean))];\n}\n\nfunction parseOptionalDate(value: string | undefined): Date | null {\n  const trimmed = cleanString(value);\n  if (!trimmed) {\n    return null;\n  }\n  const date = new Date(trimmed);\n  if (Number.isNaN(date.getTime())) {\n    throw new Error(`Invalid source timestamp: ${trimmed}`);\n  }\n  return date;\n}\n\nfunction normalizeKnowledgeSourceKind(value: string): KnowledgeSourceKind {\n  switch (value) {\n    case \"manual_upload\":\n    case \"meeting_transcript\":\n    case \"repository\":\n    case \"email\":\n    case \"chat\":\n    case \"document\":\n    case \"web\":\n    case \"other\":\n      return value;\n    default:\n      return \"other\";\n  }\n}\n\nfunction vectorLiteral(values: number[]): string {\n  return `[${values.join(\",\")}]`;\n}\n\nfunction mapDocumentBase(row: typeof schema.documentBases.$inferSelect): DocumentBase {\n  return {\n    id: row.id,\n    workspaceId: row.workspaceId,\n    name: row.name,\n    description: row.description,\n    createdAt: row.createdAt.toISOString(),\n    updatedAt: row.updatedAt.toISOString(),\n  };\n}\n\nfunction mapDocument(row: typeof schema.documents.$inferSelect): Document {\n  return {\n    id: row.id,\n    workspaceId: row.workspaceId,\n    baseId: row.baseId,\n    fileId: row.fileId,\n    status: row.status as Document[\"status\"],\n    title: row.title,\n    parser: row.parser,\n    chunkCount: row.chunkCount,\n    error: row.error,\n    sourceKind: normalizeKnowledgeSourceKind(row.sourceKind),\n    sourceUri: row.sourceUri,\n    sourceExternalId: row.sourceExternalId,\n    sourceTitle: row.sourceTitle,\n    sourceAuthor: row.sourceAuthor,\n    sourceCreatedAt: row.sourceCreatedAt?.toISOString() ?? null,\n    sourceUpdatedAt: row.sourceUpdatedAt?.toISOString() ?? null,\n    sourceVersion: row.sourceVersion,\n    aclTags: cleanStringArray(row.aclTags),\n    createdAt: row.createdAt.toISOString(),\n    updatedAt: row.updatedAt.toISOString(),\n  };\n}\n"],"mappings":";AAWA,SAAS,aAAa,gBAAgB,wBAAuC;AAC7E,YAAY,YAAY;AAExB,SAAS,iBAAiB;AAC1B,SAAS,KAAK,KAAK,MAAM,IAAI,SAAS,WAAqB;AAC3D,OAAO,YAAY;AAEZ,IAAM,0BAA0B;AAChC,IAAM,mCAAmC;AACzC,IAAM,wCAAwC;AAC9C,IAAM,8BAA8B;AACpC,IAAM,iCAAiC;AAqDvC,IAAM,0BAAN,MAAwD;AAAA,EACpD,OAAO;AAAA,EACR,aAA4B,QAAQ,QAAQ;AAAA,EAEpD,MAAM,MAAM,OAAmB,MAA0C;AACvE,UAAM,OAAO,WAAW,IAAI,IACxB,OAAO,KAAK,KAAK,EAAE,SAAS,MAAM,EAAE,QAAQ,OAAO,GAAG,EAAE,KAAK,IAC7D,MAAM,KAAK,mBAAmB,KAAK;AACvC,QAAI,CAAC,KAAK,KAAK,GAAG;AAChB,YAAM,IAAI,MAAM,6BAA6B,KAAK,QAAQ,EAAE;AAAA,IAC9D;AACA,WAAO;AAAA,MACL,MAAM,KAAK,KAAK;AAAA,MAChB,UAAU;AAAA,QACR,QAAQ,KAAK;AAAA,QACb,UAAU,KAAK;AAAA,QACf,aAAa,KAAK;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,mBAAmB,OAAoC;AACnE,WAAO,MAAM,KAAK,aAAa,YAAY;AACzC,YAAM,SAAS,IAAI,UAAU,EAAE,YAAY,MAAM,YAAY,EAAE,CAAC;AAChE,YAAM,SAAS,MAAM,OAAO,MAAM,OAAO,KAAK,KAAK,GAAG,IAAI;AAC1D,YAAM,OAAO,OAAO,QAAQ,SAAS,WAAW,OAAO,OAAO;AAC9D,aAAO,KAAK,QAAQ,OAAO,GAAG,EAAE,KAAK;AAAA,IACvC,CAAC;AAAA,EACH;AAAA,EAEA,MAAc,aAAgB,MAAoC;AAChE,UAAM,WAAW,KAAK;AACtB,QAAI,UAAsB,MAAM;AAChC,SAAK,aAAa,IAAI,QAAc,CAAC,YAAY;AAC/C,gBAAU;AAAA,IACZ,CAAC;AACD,UAAM,SAAS,MAAM,MAAM,MAAS;AACpC,QAAI;AACF,aAAO,MAAM,KAAK;AAAA,IACpB,UAAE;AACA,cAAQ;AAAA,IACV;AAAA,EACF;AACF;AAEO,IAAM,uBAAN,MAAsD;AAAA,EAC3D,YACmB,WAAW,6BACX,eAAe,gCAChC;AAFiB;AACA;AAEjB,QAAI,gBAAgB,UAAU;AAC5B,YAAM,IAAI,MAAM,wDAAwD;AAAA,IAC1E;AAAA,EACF;AAAA,EANmB;AAAA,EACA;AAAA,EAOnB,MAAM,QAAwB,MAAkC;AAC9D,WAAO,UAAU,OAAO,MAAM,KAAK,UAAU,KAAK,YAAY,EAAE,IAAI,CAAC,MAAM,WAAW;AAAA,MACpF;AAAA,MACA,UAAU;AAAA,QACR,GAAG,OAAO;AAAA,QACV,UAAU,KAAK;AAAA,QACf,aAAa,KAAK;AAAA,QAClB,YAAY;AAAA,MACd;AAAA,IACF,EAAE;AAAA,EACJ;AACF;AAEO,IAAM,0BAAN,MAA0D;AAAA,EACvD,SAAwB;AAAA,EACf;AAAA,EACA;AAAA,EAEjB,YAAY,MAOT;AACD,SAAK,SAAS,KAAK,UAAU,QAAQ,IAAI;AACzC,SAAK,UAAU,KAAK;AACpB,SAAK,iBAAiB,KAAK;AAC3B,SAAK,eAAe,KAAK;AACzB,SAAK,QAAQ,KAAK,SAAS;AAC3B,SAAK,aAAa,KAAK,cAAc;AAAA,EACvC;AAAA,EAES;AAAA,EACA;AAAA,EACQ;AAAA,EACA;AAAA,EAEjB,MAAM,UAAU,OAAsC;AACpD,QAAI,MAAM,WAAW,EAAG,QAAO,CAAC;AAChC,UAAM,MAAkB,CAAC;AACzB,aAAS,QAAQ,GAAG,QAAQ,MAAM,QAAQ,SAAS,IAAI;AACrD,YAAM,QAAQ,MAAM,MAAM,OAAO,QAAQ,EAAE;AAC3C,YAAM,WAAW,MAAM,KAAK,OAAO,EAAE,WAAW,OAAO;AAAA,QACrD,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,QACP,YAAY,KAAK;AAAA,MACnB,CAAC;AACD,iBAAW,QAAQ,SAAS,MAAM;AAChC,YAAI,KAAK,kBAAkB,KAAK,WAAW,KAAK,YAAY,KAAK,KAAK,CAAC;AAAA,MACzE;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA,EAEA,MAAM,WAAW,MAAiC;AAChD,UAAM,CAAC,SAAS,IAAI,MAAM,KAAK,UAAU,CAAC,IAAI,CAAC;AAC/C,QAAI,CAAC,WAAW;AACd,YAAM,IAAI,MAAM,gDAAgD;AAAA,IAClE;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,SAAiB;AACvB,QAAI,CAAC,KAAK,QAAQ;AAChB,YAAM,IAAI,MAAM,+CAA+C;AAAA,IACjE;AACA,SAAK,WAAW,IAAI,OAAO;AAAA,MACzB,QAAQ,KAAK;AAAA,MACb,GAAI,KAAK,UAAU,EAAE,SAAS,KAAK,QAAQ,IAAI,CAAC;AAAA,MAChD,GAAI,KAAK,eAAe,EAAE,cAAc,KAAK,aAAa,IAAI,CAAC;AAAA,MAC/D,GAAI,KAAK,iBAAiB,EAAE,gBAAgB,KAAK,eAAe,IAAI,CAAC;AAAA,IACvE,CAAC;AACD,WAAO,KAAK;AAAA,EACd;AACF;AAEO,IAAM,iCAAN,MAAiE;AAAA,EACtE,YACW,aAAa,uCACb,QAAQ,uBAAuB,UAAU,IAClD;AAFS;AACA;AAAA,EACR;AAAA,EAFQ;AAAA,EACA;AAAA,EAGX,MAAM,UAAU,OAAsC;AACpD,WAAO,MAAM,IAAI,CAAC,SAAS,uBAAuB,MAAM,KAAK,UAAU,CAAC;AAAA,EAC1E;AAAA,EAEA,MAAM,WAAW,MAAiC;AAChD,WAAO,uBAAuB,MAAM,KAAK,UAAU;AAAA,EACrD;AACF;AAEO,SAAS,uBAAuB,UAAqB,YAAuC,CAAC,GAAqB;AACvH,QAAM,aAAa,UAAU,+BAA+B;AAC5D,QAAM,wBAAwB,8BAA8B,QAAQ;AACpE,SAAO;AAAA,IACL,QAAQ,UAAU,UAAU,IAAI,wBAAwB;AAAA,IACxD,SAAS,UAAU,WAAW,IAAI;AAAA,MAChC,UAAU,qBAAqB;AAAA,MAC/B,UAAU,wBAAwB;AAAA,IACpC;AAAA,IACA,UAAU,UAAU,aAClB,UAAU,8BAA8B,kBACpC,IAAI,+BAA+B,YAAY,SAAS,sBAAsB,IAC9E,IAAI,wBAAwB;AAAA,MAC5B,GAAG;AAAA,MACH,OAAO,UAAU,0BAA0B;AAAA,MAC3C;AAAA,IACF,CAAC;AAAA,EAEP;AACF;AAEO,SAAS,8BAA8B,UAK5C;AACA,MAAI,CAAC,SAAU,QAAO,CAAC;AACvB,MAAI,SAAS,2BAA2B,SAAS,0BAA0B;AACzE,WAAO;AAAA,MACL,QAAQ,SAAS,2BAA2B,SAAS,gBAAgB,SAAS;AAAA,MAC9E,SAAS,SAAS,4BAA4B,SAAS,iBAAiB,SAAS;AAAA,IACnF;AAAA,EACF;AACA,MAAI,SAAS,mBAAmB,SAAS;AACvC,UAAM,UAAU,SAAS,sBAAsB,uBAAuB,QAAQ;AAC9E,WAAO;AAAA,MACL,QAAQ,SAAS,qBAAqB,SAAS,sBAAsB;AAAA,MACrE;AAAA,MACA,cAAc,wBAAwB,UAAU,OAAO;AAAA,MACvD,gBAAgB,SAAS,sBAAsB,CAAC,SAAS,oBACrD,EAAE,eAAe,UAAU,SAAS,kBAAkB,GAAG,IACzD;AAAA,IACN;AAAA,EACF;AACA,SAAO;AAAA,IACL,QAAQ,SAAS;AAAA,IACjB,SAAS,SAAS;AAAA,EACpB;AACF;AAEA,SAAS,uBAAuB,UAA4B;AAC1D,QAAM,WAAW,SAAS,qBAAqB,QAAQ,QAAQ,EAAE;AACjE,MAAI,CAAC,YAAY,CAAC,SAAS,uBAAuB;AAChD,UAAM,IAAI,MAAM,0DAA0D;AAAA,EAC5E;AACA,SAAO,GAAG,QAAQ,uBAAuB,SAAS,qBAAqB;AACzE;AAEA,SAAS,wBACP,UACA,SACoC;AACpC,MAAI,CAAC,SAAS,sBAAuB,QAAO;AAC5C,QAAM,aAAa,QAAQ,QAAQ,QAAQ,EAAE,EAAE,YAAY;AAC3D,MAAI,WAAW,SAAS,YAAY,GAAG;AACrC,WAAO;AAAA,EACT;AACA,SAAO,EAAE,eAAe,SAAS,sBAAsB;AACzD;AAEA,eAAsB,mBAAmB,IAAc,OAAsG;AAC3J,SAAO,MAAM,eAAe,IAAI,EAAE,WAAW,MAAM,WAAW,aAAa,MAAM,YAAY,GAAG,OAAO,aAAa;AAClH,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAc,oBAAa,EAAE,OAAO;AAAA,MAC/D,WAAW,MAAM;AAAA,MACjB,aAAa,MAAM;AAAA,MACnB,MAAM,MAAM,KAAK,KAAK;AAAA,MACtB,aAAa,MAAM,aAAa,KAAK,KAAK;AAAA,IAC5C,CAAC,EAAE,UAAU;AACb,QAAI,CAAC,IAAK,OAAM,IAAI,MAAM,gCAAgC;AAC1D,WAAO,gBAAgB,GAAG;AAAA,EAC5B,CAAC;AACH;AAEA,eAAsB,kBAAkB,IAAc,aAA8C;AAClG,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,OAAO,MAAM,SAAS,OAAO,EAAE,KAAY,oBAAa,EAC3D,MAAM,GAAU,qBAAc,aAAa,WAAW,CAAC,EACvD,QAAQ,KAAY,qBAAc,SAAS,CAAC;AAC/C,WAAO,KAAK,IAAI,eAAe;AAAA,EACjC,CAAC;AACH;AAEA,eAAsB,gBAAgB,IAAc,aAAqB,QAA8C;AACrH,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,oBAAa,EAAE,MAAM,IAAI,GAAU,qBAAc,aAAa,WAAW,GAAG,GAAU,qBAAc,IAAI,MAAM,CAAC,CAAC,EAAE,MAAM,CAAC;AAC3K,WAAO,MAAM,gBAAgB,GAAG,IAAI;AAAA,EACtC,CAAC;AACH;AAEA,eAAsB,kBAAkB,IAAc,OAA2G;AAC/J,SAAO,MAAM,eAAe,IAAI,EAAE,WAAW,MAAM,WAAW,aAAa,MAAM,YAAY,GAAG,OAAO,aAAa;AAClH,UAAM,OAAO,MAAM,gBAAgB,UAAU,MAAM,aAAa,MAAM,MAAM;AAC5E,QAAI,CAAC,KAAM,OAAM,IAAI,MAAM,4BAA4B,MAAM,MAAM,EAAE;AACrE,UAAM,OAAO,MAAM,iBAAiB,UAAU,MAAM,aAAa,MAAM,MAAM;AAC7E,UAAM,MAAM,oBAAI,KAAK;AACrB,UAAM,CAAC,QAAQ,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAC7D,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,QAAQ,MAAM,MAAM,GAAG,GAAU,iBAAU,QAAQ,MAAM,MAAM,CAAC,CAAC,EACpJ,MAAM,CAAC;AACV,QAAI,UAAU;AAIZ,YAAM,CAAC,OAAO,IAAI,MAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,QAC5D,OAAO,YAAY,MAAM,KAAK,KAAK,YAAY,MAAM,WAAW,KAAK,SAAS;AAAA,QAC9E,GAAI,MAAM,eAAe,SAAY,EAAE,YAAY,MAAM,WAAW,IAAI,CAAC;AAAA,QACzE,WAAW,YAAY,MAAM,SAAS,KAAK,SAAS;AAAA,QACpD,kBAAkB,YAAY,MAAM,gBAAgB,KAAK,SAAS;AAAA,QAClE,aAAa,YAAY,MAAM,WAAW,KAAK,SAAS;AAAA,QACxD,cAAc,YAAY,MAAM,YAAY,KAAK,SAAS;AAAA,QAC1D,iBAAiB,kBAAkB,MAAM,eAAe,KAAK,SAAS;AAAA,QACtE,iBAAiB,kBAAkB,MAAM,eAAe,KAAK,SAAS;AAAA,QACtE,eAAe,YAAY,MAAM,aAAa,KAAK,SAAS;AAAA,QAC5D,GAAI,MAAM,YAAY,SAAY,EAAE,SAAS,iBAAiB,MAAM,OAAO,EAAE,IAAI,CAAC;AAAA,QAClF,WAAW;AAAA,MACb,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,IAAI,SAAS,EAAE,CAAC,CAAC,EAAE,UAAU;AACnH,aAAO,YAAY,WAAW,QAAQ;AAAA,IACxC;AACA,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAc,gBAAS,EAAE,OAAO;AAAA,MAC3D,WAAW,MAAM;AAAA,MACjB,aAAa,MAAM;AAAA,MACnB,QAAQ,MAAM;AAAA,MACd,QAAQ,MAAM;AAAA,MACd,QAAQ;AAAA,MACR,OAAO,YAAY,MAAM,KAAK,KAAK,YAAY,MAAM,WAAW,KAAK,KAAK;AAAA,MAC1E,QAAQ;AAAA,MACR,YAAY,MAAM,cAAc;AAAA,MAChC,WAAW,YAAY,MAAM,SAAS,KAAK;AAAA,MAC3C,kBAAkB,YAAY,MAAM,gBAAgB,KAAK;AAAA,MACzD,aAAa,YAAY,MAAM,WAAW,KAAK;AAAA,MAC/C,cAAc,YAAY,MAAM,YAAY,KAAK;AAAA,MACjD,iBAAiB,kBAAkB,MAAM,eAAe;AAAA,MACxD,iBAAiB,kBAAkB,MAAM,eAAe;AAAA,MACxD,eAAe,YAAY,MAAM,aAAa,KAAK;AAAA,MACnD,SAAS,iBAAiB,MAAM,OAAO;AAAA,MACvC,WAAW;AAAA,IACb,CAAC,EAAE,UAAU;AACb,QAAI,CAAC,IAAK,OAAM,IAAI,MAAM,2BAA2B;AACrD,WAAO,YAAY,GAAG;AAAA,EACxB,CAAC;AACH;AAEA,eAAsB,uBACpB,IACA,OACe;AACf,QAAM,eAAe,IAAI,EAAE,WAAW,MAAM,WAAW,aAAa,MAAM,YAAY,GAAG,OAAO,aAAa;AAC3G,UAAM,CAAC,QAAQ,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAC7D,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,IAAI,MAAM,UAAU,CAAC,CAAC,EACzG,MAAM,CAAC;AACV,QAAI,CAAC,UAAU;AACb,YAAM,IAAI,MAAM,uBAAuB,MAAM,UAAU,EAAE;AAAA,IAC3D;AACA,QAAI,SAAS,WAAW,MAAM,QAAQ;AACpC,YAAM,IAAI,MAAM,uBAAuB,MAAM,UAAU,EAAE;AAAA,IAC3D;AACA,UAAM,SAAS,OAAc,gBAAS,EACnC,MAAM,IAAI,GAAU,iBAAU,aAAa,MAAM,WAAW,GAAG,GAAU,iBAAU,IAAI,MAAM,UAAU,CAAC,CAAC;AAAA,EAC9G,CAAC;AACH;AAEA,eAAsB,cAAc,IAAc,aAAqB,QAAqC;AAC1G,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,OAAO,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EACvD,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,QAAQ,MAAM,CAAC,CAAC,EAC7F,QAAQ,IAAW,iBAAU,SAAS,CAAC;AAC1C,WAAO,KAAK,IAAI,WAAW;AAAA,EAC7B,CAAC;AACH;AAEA,eAAsB,YAAY,IAAc,aAAqB,YAA8C;AACjH,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,MAAM,CAAC;AACnK,WAAO,MAAM,YAAY,GAAG,IAAI;AAAA,EAClC,CAAC;AACH;AAEA,eAAsB,wBAAwB,IAAc,aAAqB,YAAuC;AACtH,SAAO,MAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AACjE,UAAM,CAAC,GAAG,IAAI,MAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,MACxD,QAAQ;AAAA,MACR,OAAO;AAAA,MACP,WAAW,oBAAI,KAAK;AAAA,IACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,UAAU;AAC5G,QAAI,CAAC,IAAK,OAAM,IAAI,MAAM,uBAAuB,UAAU,EAAE;AAC7D,WAAO,YAAY,GAAG;AAAA,EACxB,CAAC;AACH;AAEA,eAAsB,iBACpB,IACA,eACA,aACA,YACA,WAA6B,uBAAuB,GACpD,QAA4B,CAAC,GACV;AACnB,QAAM,CAAC,QAAQ,IAAI,MAAM;AAAA,IAAiB;AAAA,IAAI;AAAA,IAAa,OAAO,aAChE,MAAM,SAAS,OAAO,EAAE,KAAY,gBAAS,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,MAAM,CAAC;AAAA,EACvJ;AACA,MAAI,CAAC,SAAU,OAAM,IAAI,MAAM,uBAAuB,UAAU,EAAE;AAClE,QAAM,OAAO,MAAM,iBAAiB,IAAI,aAAa,SAAS,MAAM;AACpE,QAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa;AAC1D,UAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,MAC1C,QAAQ;AAAA,MACR,QAAQ,SAAS,OAAO;AAAA,MACxB,OAAO;AAAA,MACP,WAAW,oBAAI,KAAK;AAAA,IACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC;AAAA,EAClG,CAAC;AACD,MAAI;AACF,UAAM,QAAQ,MAAM,cAAc,aAAa,IAAI;AACnD,UAAM,SAAS,MAAM,SAAS,OAAO,MAAM,OAAO,IAAI;AACtD,UAAM,SAAS,SAAS,QAAQ,MAAM,QAAQ,IAAI;AAClD,UAAM,MAAM,cAAc;AAAA,MACxB,WAAW,SAAS;AAAA,MACpB,aAAa,SAAS;AAAA,MACtB;AAAA,MACA,YAAY,OAAO;AAAA,IACrB,CAAC;AACD,UAAM,aAAa,MAAM,SAAS,SAAS,UAAU,OAAO,IAAI,CAAC,UAAU,MAAM,IAAI,CAAC;AACtF,QAAI,WAAW,WAAW,OAAO,QAAQ;AACvC,YAAM,IAAI,MAAM,+BAA+B,WAAW,MAAM,mBAAmB,OAAO,MAAM,SAAS;AAAA,IAC3G;AACA,UAAM,iBAAiB,IAAI,aAAa,OAAO,aAAa,MAAM,SAAS,YAAY,OAAO,OAAO;AACnG,YAAM,GAAG,OAAc,qBAAc,EAAE,MAAM,IAAI,GAAU,sBAAe,aAAa,WAAW,GAAG,GAAU,sBAAe,YAAY,UAAU,CAAC,CAAC;AACtJ,UAAI,OAAO,SAAS,GAAG;AACrB,cAAM,GAAG,OAAc,qBAAc,EAAE,OAAO,OAAO,IAAI,CAAC,OAAO,WAAW;AAAA,UAC1E,WAAW,SAAS;AAAA,UACpB,aAAa,SAAS;AAAA,UACtB;AAAA,UACA,QAAQ,SAAS;AAAA,UACjB,QAAQ,KAAK;AAAA,UACb,YAAY;AAAA,UACZ,MAAM,MAAM;AAAA,UACZ,UAAU;AAAA,YACR,GAAG,MAAM;AAAA,YACT,eAAe,SAAS;AAAA,YACxB,YAAY,SAAS;AAAA,YACrB,WAAW,SAAS;AAAA,YACpB,kBAAkB,SAAS;AAAA,YAC3B,aAAa,SAAS;AAAA,YACtB,cAAc,SAAS;AAAA,YACvB,iBAAiB,SAAS,iBAAiB,YAAY,KAAK;AAAA,YAC5D,iBAAiB,SAAS,iBAAiB,YAAY,KAAK;AAAA,YAC5D,eAAe,SAAS;AAAA,YACxB,SAAS,SAAS;AAAA,UACpB;AAAA,UACA,WAAW,kBAAkB,WAAW,KAAK,KAAK,CAAC,GAAG,SAAS,SAAS,YAAY,SAAS,SAAS,KAAK;AAAA,UAC3G,gBAAgB,SAAS,SAAS;AAAA,QACpC,EAAE,CAAC;AAAA,MACL;AACA,YAAM,GAAG,OAAc,gBAAS,EAAE,IAAI;AAAA,QACpC,QAAQ;AAAA,QACR,QAAQ,SAAS,OAAO;AAAA,QACxB,YAAY,OAAO;AAAA,QACnB,OAAO;AAAA,QACP,WAAW,oBAAI,KAAK;AAAA,MACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC;AAAA,IAClG,CAAC,CAAC;AAAA,EACJ,SAAS,OAAO;AACd,UAAM,CAAC,MAAM,IAAI,MAAM;AAAA,MAAiB;AAAA,MAAI;AAAA,MAAa,OAAO,aAC9D,MAAM,SAAS,OAAc,gBAAS,EAAE,IAAI;AAAA,QAC1C,QAAQ;AAAA,QACR,OAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAAA,QAC5D,WAAW,oBAAI,KAAK;AAAA,MACtB,CAAC,EAAE,MAAM,IAAI,GAAU,iBAAU,aAAa,WAAW,GAAG,GAAU,iBAAU,IAAI,UAAU,CAAC,CAAC,EAAE,UAAU;AAAA,IAC9G;AACA,QAAI,CAAC,OAAQ,OAAM;AACnB,WAAO,YAAY,MAAM;AAAA,EAC3B;AACA,QAAM,UAAU,MAAM,YAAY,IAAI,aAAa,UAAU;AAC7D,MAAI,CAAC,QAAS,OAAM,IAAI,MAAM,wCAAwC,UAAU,EAAE;AAClF,SAAO;AACT;AAEA,eAAsB,gBACpB,IACA,OACA,WAA+C,uBAAuB,GACrC;AACjC,QAAM,OAAO,MAAM,QAAQ;AAC3B,QAAM,QAAQ,KAAK,IAAI,KAAK,IAAI,MAAM,SAAS,GAAG,CAAC,GAAG,EAAE;AACxD,QAAM,iBAAiB,SAAS,WAAW,KAAK,IAAI,QAAQ,GAAG,GAAG,IAAI;AACtE,QAAM,OAA4B,CAAC;AACnC,MAAI,SAAS,YAAY,SAAS,UAAU;AAC1C,QAAI;AACF,WAAK,KAAK,GAAG,MAAM,sBAAsB,IAAI,OAAO,gBAAgB,QAAQ,CAAC;AAAA,IAC/E,SAAS,OAAO;AACd,UAAI,SAAS,UAAU;AACrB,cAAM;AAAA,MACR;AACA,cAAQ,KAAK,kFAAkF;AAAA,QAC7F,aAAa,MAAM;AAAA,QACnB,OAAO,iBAAiB,QAAQ,MAAM,UAAU,OAAO,KAAK;AAAA,MAC9D,CAAC;AAAA,IACH;AAAA,EACF;AACA,MAAI,SAAS,aAAa,SAAS,UAAU;AAC3C,SAAK,KAAK,GAAG,MAAM,uBAAuB,IAAI,OAAO,cAAc,CAAC;AAAA,EACtE;AACA,SAAO,wBAAwB,MAAM,IAAI,EAAE,MAAM,GAAG,KAAK;AAC3D;AAEA,eAAe,sBACb,IACA,OACA,OACA,UAC8B;AAC9B,QAAM,iBAAiB,MAAM,SAAS,SAAS,WAAW,MAAM,KAAK;AACrE,oBAAkB,gBAAgB,SAAS,SAAS,YAAY,SAAS,SAAS,KAAK;AACvF,QAAM,WAAW,MAAqB,sBAAe,SAAS,QAAQ,cAAc,cAAc,CAAC;AACnG,QAAM,OAAO,MAAM;AAAA,IAAiB;AAAA,IAAI,MAAM;AAAA,IAAa,OAAO,aAChE,MAAM,SAAS,OAAO;AAAA,MACpB,SAAgB,sBAAe;AAAA,MAC/B,YAAmB,sBAAe;AAAA,MAClC,QAAe,sBAAe;AAAA,MAC9B,QAAe,sBAAe;AAAA,MAC9B,OAAc,iBAAU;AAAA,MACxB,MAAa,sBAAe;AAAA,MAC5B,YAAmB,sBAAe;AAAA,MAClC,UAAiB,sBAAe;AAAA,MAChC,YAAmB,iBAAU;AAAA,MAC7B,WAAkB,iBAAU;AAAA,MAC5B,kBAAyB,iBAAU;AAAA,MACnC,aAAoB,iBAAU;AAAA,MAC9B,cAAqB,iBAAU;AAAA,MAC/B,iBAAwB,iBAAU;AAAA,MAClC,iBAAwB,iBAAU;AAAA,MAClC,eAAsB,iBAAU;AAAA,MAChC,SAAgB,iBAAU;AAAA,MAC1B;AAAA,IACF,CAAC,EAAE,KAAY,qBAAc,EAC1B,UAAiB,kBAAW,GAAU,sBAAe,YAAmB,iBAAU,EAAE,CAAC,EACrF,MAAM,IAAI,GAAG,yBAAyB,OAAO,SAAS,SAAS,KAAK,CAAC,CAAC,EACtE,QAAQ,QAAQ,EAChB,MAAM,KAAK;AAAA,EAChB;AACA,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB,GAAG,iBAAiB,KAAK,MAAM,WAAW;AAAA,IAC1C,aAAa,KAAK,IAAI,OAAO,IAAI,QAAQ;AAAA,IACzC,cAAc;AAAA,EAChB,EAAE;AACJ;AAEA,eAAe,uBAAuB,IAAc,OAA4B,OAA6C;AAC3H,QAAM,OAAO,uCAAsD,sBAAe,IAAI,gCAAgC,MAAM,KAAK;AACjI,QAAM,OAAO,MAAM;AAAA,IAAiB;AAAA,IAAI,MAAM;AAAA,IAAa,OAAO,aAChE,MAAM,SAAS,OAAO;AAAA,MACpB,SAAgB,sBAAe;AAAA,MAC/B,YAAmB,sBAAe;AAAA,MAClC,QAAe,sBAAe;AAAA,MAC9B,QAAe,sBAAe;AAAA,MAC9B,OAAc,iBAAU;AAAA,MACxB,MAAa,sBAAe;AAAA,MAC5B,YAAmB,sBAAe;AAAA,MAClC,UAAiB,sBAAe;AAAA,MAChC,YAAmB,iBAAU;AAAA,MAC7B,WAAkB,iBAAU;AAAA,MAC5B,kBAAyB,iBAAU;AAAA,MACnC,aAAoB,iBAAU;AAAA,MAC9B,cAAqB,iBAAU;AAAA,MAC/B,iBAAwB,iBAAU;AAAA,MAClC,iBAAwB,iBAAU;AAAA,MAClC,eAAsB,iBAAU;AAAA,MAChC,SAAgB,iBAAU;AAAA,MAC1B;AAAA,IACF,CAAC,EAAE,KAAY,qBAAc,EAC1B,UAAiB,kBAAW,GAAU,sBAAe,YAAmB,iBAAU,EAAE,CAAC,EACrF,MAAM;AAAA,MACL,GAAG,yBAAyB,KAAK;AAAA,MACjC,4BAAmC,sBAAe,IAAI,kCAAkC,MAAM,KAAK;AAAA,IACrG,CAAC,EACA,QAAQ,KAAK,IAAI,CAAC,EAClB,MAAM,KAAK;AAAA,EAChB;AACA,SAAO,KAAK,IAAI,CAAC,SAAS;AAAA,IACxB,GAAG,iBAAiB,KAAK,MAAM,WAAW;AAAA,IAC1C,aAAa;AAAA,IACb,cAAc,sBAAsB,OAAO,IAAI,IAAI,CAAC;AAAA,EACtD,EAAE;AACJ;AAEA,eAAsB,iBAAiB,IAAc,aAAqB,SAAuD;AAC/H,QAAM,CAAC,GAAG,IAAI,MAAM;AAAA,IAAiB;AAAA,IAAI;AAAA,IAAa,OAAO,aAC3D,MAAM,SAAS,OAAO;AAAA,MACpB,SAAgB,sBAAe;AAAA,MAC/B,YAAmB,sBAAe;AAAA,MAClC,QAAe,sBAAe;AAAA,MAC9B,QAAe,sBAAe;AAAA,MAC9B,OAAc,iBAAU;AAAA,MACxB,MAAa,sBAAe;AAAA,MAC5B,YAAmB,sBAAe;AAAA,MAClC,UAAiB,sBAAe;AAAA,MAChC,YAAmB,iBAAU;AAAA,MAC7B,WAAkB,iBAAU;AAAA,MAC5B,kBAAyB,iBAAU;AAAA,MACnC,aAAoB,iBAAU;AAAA,MAC9B,cAAqB,iBAAU;AAAA,MAC/B,iBAAwB,iBAAU;AAAA,MAClC,iBAAwB,iBAAU;AAAA,MAClC,eAAsB,iBAAU;AAAA,MAChC,SAAgB,iBAAU;AAAA,IAC5B,CAAC,EAAE,KAAY,qBAAc,EAC1B,UAAiB,kBAAW,GAAU,sBAAe,YAAmB,iBAAU,EAAE,CAAC,EACrF,MAAM,IAAI,GAAU,sBAAe,aAAa,WAAW,GAAG,GAAU,sBAAe,IAAI,OAAO,GAAG,GAAU,iBAAU,QAAQ,OAAO,CAAC,CAAC,EAC1I,MAAM,CAAC;AAAA,EACZ;AACA,MAAI,CAAC,IAAK,QAAO;AACjB,SAAO;AAAA,IACL,GAAG,iBAAiB,KAAK,WAAW;AAAA,IACpC,OAAO;AAAA,IACP,WAAW;AAAA,IACX,aAAa;AAAA,IACb,cAAc;AAAA,EAChB;AACF;AAQA,SAAS,yBAAyB,OAA4B,gBAAgC;AAC5F,QAAM,aAAoB;AAAA,IACxB,GAAU,iBAAU,QAAQ,OAAO;AAAA,IACnC,GAAU,sBAAe,aAAa,MAAM,WAAW;AAAA,EACzD;AACA,MAAI,gBAAgB;AAClB,eAAW,KAAK,GAAU,sBAAe,gBAAgB,cAAc,CAAC;AAAA,EAC1E;AACA,MAAI,MAAM,SAAS,QAAQ;AACzB,eAAW,KAAK,QAAe,sBAAe,QAAQ,MAAM,OAAO,CAAC;AAAA,EACtE;AACA,MAAI,MAAM,aAAa,QAAQ;AAC7B,eAAW,KAAK,QAAe,iBAAU,YAAY,MAAM,WAAW,CAAC;AAAA,EACzE;AACA,QAAM,UAAU,iBAAiB,MAAM,OAAO;AAC9C,MAAI,QAAQ,SAAS,GAAG;AACtB,eAAW,KAAK,MAAa,iBAAU,OAAO,OAAO,KAAK,UAAU,OAAO,CAAC,SAAS;AAAA,EACvF;AACA,SAAO;AACT;AAEA,SAAS,iBAAiB,KAkBvB,aAAoC;AACrC,SAAO;AAAA,IACL,SAAS,IAAI;AAAA,IACb;AAAA,IACA,YAAY,IAAI;AAAA,IAChB,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,OAAO,IAAI;AAAA,IACX,MAAM,IAAI;AAAA,IACV,YAAY,IAAI;AAAA,IAChB,UAAU,IAAI;AAAA,IACd,YAAY,6BAA6B,IAAI,UAAU;AAAA,IACvD,WAAW,IAAI;AAAA,IACf,kBAAkB,IAAI;AAAA,IACtB,aAAa,IAAI;AAAA,IACjB,cAAc,IAAI;AAAA,IAClB,iBAAiB,IAAI,iBAAiB,YAAY,KAAK;AAAA,IACvD,iBAAiB,IAAI,iBAAiB,YAAY,KAAK;AAAA,IACvD,eAAe,IAAI;AAAA,IACnB,SAAS,iBAAiB,IAAI,OAAO;AAAA,EACvC;AACF;AAEA,SAAS,wBAAwB,MAA2B,MAAkD;AAC5G,QAAM,UAAU,oBAAI,IAA+B;AACnD,aAAW,OAAO,MAAM;AACtB,UAAM,WAAW,QAAQ,IAAI,IAAI,OAAO;AACxC,QAAI,CAAC,UAAU;AACb,cAAQ,IAAI,IAAI,SAAS,GAAG;AAC5B;AAAA,IACF;AACA,YAAQ,IAAI,IAAI,SAAS;AAAA,MACvB,GAAG;AAAA,MACH,aAAa,KAAK,IAAI,SAAS,eAAe,GAAG,IAAI,eAAe,CAAC,KAAK;AAAA,MAC1E,cAAc,KAAK,IAAI,SAAS,gBAAgB,GAAG,IAAI,gBAAgB,CAAC,KAAK;AAAA,IAC/E,CAAC;AAAA,EACH;AACA,SAAO,CAAC,GAAG,QAAQ,OAAO,CAAC,EAAE,IAAI,CAAC,QAAQ;AACxC,UAAM,cAAc,IAAI;AACxB,UAAM,eAAe,IAAI;AACzB,UAAM,YAAgC,gBAAgB,QAAQ,iBAAiB,OAC3E,WACA,gBAAgB,OACd,WACA;AACN,WAAO;AAAA,MACL,GAAG;AAAA,MACH,OAAO,oBAAoB,MAAM,aAAa,cAAc,SAAS;AAAA,MACrE;AAAA,IACF;AAAA,EACF,CAAC,EAAE;AAAA,IAAK,CAAC,MAAM,UACb,MAAM,QAAQ,KAAK,UACf,MAAM,eAAe,MAAM,KAAK,eAAe,OAC/C,MAAM,gBAAgB,MAAM,KAAK,gBAAgB,MAClD,KAAK,aAAa,MAAM;AAAA,EAC7B;AACF;AAEA,SAAS,oBACP,MACA,aACA,cACA,WACQ;AACR,QAAM,SAAS,eAAe;AAC9B,QAAM,UAAU,gBAAgB;AAChC,MAAI,SAAS,SAAU,QAAO,WAAW,MAAM;AAC/C,MAAI,SAAS,UAAW,QAAO,WAAW,OAAO;AACjD,SAAO,WAAW,KAAK,IAAI,GAAI,OAAO,SAAW,OAAO,WAAY,cAAc,WAAW,MAAM,EAAE,CAAC;AACxG;AAEA,SAAS,sBAAsB,MAAsB;AACnD,MAAI,CAAC,OAAO,SAAS,IAAI,KAAK,QAAQ,GAAG;AACvC,WAAO;AAAA,EACT;AACA,SAAO,QAAQ,OAAO;AACxB;AAEA,SAAS,WAAW,OAAuB;AACzC,SAAO,OAAO,MAAM,QAAQ,CAAC,CAAC;AAChC;AAEA,eAAsB,mBAAmB,OAAmB,MAAiB,SAAyB,IAAI,wBAAwB,GAA4B;AAC5J,SAAO,MAAM,OAAO,MAAM,OAAO,IAAI;AACvC;AAEO,SAAS,UAAU,MAAc,WAAW,6BAA6B,eAAe,gCAA0C;AACvI,MAAI,gBAAgB,UAAU;AAC5B,UAAM,IAAI,MAAM,+CAA+C;AAAA,EACjE;AACA,QAAM,aAAa,KAAK,QAAQ,SAAS,IAAI,EAAE,QAAQ,WAAW,GAAG,EAAE,KAAK;AAC5E,MAAI,CAAC,WAAY,QAAO,CAAC;AACzB,QAAM,aAAa,WAAW,MAAM,QAAQ,EAAE,IAAI,CAAC,SAAS,KAAK,QAAQ,QAAQ,GAAG,EAAE,KAAK,CAAC,EAAE,OAAO,OAAO;AAC5G,QAAM,SAAmB,CAAC;AAC1B,MAAI,UAAU;AACd,aAAW,aAAa,WAAW,SAAS,IAAI,aAAa,CAAC,WAAW,QAAQ,QAAQ,GAAG,CAAC,GAAG;AAC9F,eAAW,QAAQ,mBAAmB,WAAW,QAAQ,GAAG;AAC1D,UAAI,CAAC,SAAS;AACZ,kBAAU;AAAA,MACZ,WAAW,QAAQ,SAAS,IAAI,KAAK,UAAU,UAAU;AACvD,kBAAU,GAAG,OAAO,IAAI,IAAI;AAAA,MAC9B,OAAO;AACL,eAAO,KAAK,OAAO;AACnB,kBAAU,YAAY,SAAS,cAAc,MAAM,QAAQ;AAAA,MAC7D;AAAA,IACF;AAAA,EACF;AACA,MAAI,QAAS,QAAO,KAAK,OAAO;AAChC,SAAO,OAAO,IAAI,CAAC,UAAU,MAAM,KAAK,CAAC,EAAE,OAAO,OAAO;AAC3D;AAEO,SAAS,uBAAuB,MAAc,aAAa,uCAAiD;AACjH,QAAM,SAAS,IAAI,MAAM,UAAU,EAAE,KAAK,CAAC;AAC3C,QAAM,SAAS,KAAK,YAAY,EAAE,MAAM,mBAAmB,KAAK,CAAC;AACjE,aAAW,SAAS,QAAQ;AAC1B,QAAI,OAAO;AACX,eAAW,QAAQ,OAAO;AACxB,cAAQ,KAAK,YAAY,CAAC,KAAK;AAC/B,aAAO,KAAK,KAAK,MAAM,QAAQ;AAAA,IACjC;AACA,WAAO,KAAK,IAAI,IAAI,IAAI,UAAU,KAAK;AAAA,EACzC;AACA,QAAM,OAAO,KAAK,MAAM,GAAG,MAAM,KAAK;AACtC,SAAO,OAAO,IAAI,CAAC,UAAU,QAAQ,QAAQ,MAAM,QAAQ,CAAC,CAAC,CAAC;AAChE;AAEA,eAAe,iBAAiB,IAAc,aAAqB,QAAoC;AACrG,QAAM,OAAO,MAAM,YAAY,IAAI,aAAa,MAAM;AACtD,MAAI,KAAK,WAAW,SAAS;AAC3B,UAAM,IAAI,MAAM,QAAQ,MAAM,OAAO,KAAK,MAAM,EAAE;AAAA,EACpD;AACA,SAAO;AACT;AAEA,SAAS,kBAAkB,QAAkB,YAAoB,OAAyB;AACxF,MAAI,OAAO,WAAW,YAAY;AAChC,UAAM,IAAI,MAAM,mBAAmB,KAAK,aAAa,OAAO,MAAM,yBAAyB,UAAU,EAAE;AAAA,EACzG;AACA,MAAI,OAAO,KAAK,CAAC,UAAU,CAAC,OAAO,SAAS,KAAK,CAAC,GAAG;AACnD,UAAM,IAAI,MAAM,mBAAmB,KAAK,6BAA6B;AAAA,EACvE;AACA,SAAO;AACT;AAEA,SAAS,WAAW,MAA0B;AAC5C,QAAM,cAAc,KAAK,YAAY,YAAY;AACjD,QAAM,WAAW,KAAK,SAAS,YAAY;AAC3C,SAAO,YAAY,WAAW,OAAO,KAChC,gBAAgB,sBAChB,gBAAgB,qBAChB,gBAAgB,wBAChB,SAAS,SAAS,KAAK,KACvB,SAAS,SAAS,WAAW,KAC7B,SAAS,SAAS,OAAO,KACzB,SAAS,SAAS,OAAO,KACzB,SAAS,SAAS,MAAM,KACxB,SAAS,SAAS,MAAM,KACxB,SAAS,SAAS,MAAM,KACxB,SAAS,SAAS,MAAM;AAC/B;AAEA,SAAS,mBAAmB,MAAc,UAA4B;AACpE,MAAI,KAAK,UAAU,SAAU,QAAO,CAAC,IAAI;AACzC,QAAM,MAAgB,CAAC;AACvB,MAAI,YAAY;AAChB,SAAO,UAAU,SAAS,UAAU;AAClC,UAAM,SAAS,UAAU,MAAM,GAAG,WAAW,CAAC;AAC9C,UAAM,UAAU,KAAK;AAAA,MACnB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,IAAI;AAAA,MACvB,OAAO,YAAY,GAAG;AAAA,IACxB;AACA,UAAM,MAAM,UAAU,KAAK,MAAM,WAAW,GAAG,IAAI,UAAU,IAAI;AACjE,QAAI,KAAK,UAAU,MAAM,GAAG,GAAG,EAAE,KAAK,CAAC;AACvC,gBAAY,UAAU,MAAM,GAAG,EAAE,KAAK;AAAA,EACxC;AACA,MAAI,UAAW,KAAI,KAAK,SAAS;AACjC,SAAO;AACT;AAEA,SAAS,YAAY,UAAkB,cAAsB,MAAc,UAA0B;AACnG,MAAI,gBAAgB,EAAG,QAAO;AAC9B,QAAM,UAAU,SAAS,MAAM,KAAK,IAAI,GAAG,SAAS,SAAS,YAAY,CAAC,EAAE,QAAQ,WAAW,EAAE,EAAE,KAAK;AACxG,QAAM,YAAY,UAAU,GAAG,OAAO,IAAI,IAAI,KAAK;AACnD,SAAO,UAAU,UAAU,WAAW,YAAY;AACpD;AAEA,SAAS,YAAY,OAAsD;AACzE,QAAM,UAAU,OAAO,KAAK;AAC5B,SAAO,UAAU,UAAU;AAC7B;AAEA,SAAS,iBAAiB,QAA+C;AACvE,SAAO,CAAC,GAAG,IAAI,KAAK,UAAU,CAAC,GAAG,IAAI,CAAC,UAAU,MAAM,KAAK,CAAC,EAAE,OAAO,OAAO,CAAC,CAAC;AACjF;AAEA,SAAS,kBAAkB,OAAwC;AACjE,QAAM,UAAU,YAAY,KAAK;AACjC,MAAI,CAAC,SAAS;AACZ,WAAO;AAAA,EACT;AACA,QAAM,OAAO,IAAI,KAAK,OAAO;AAC7B,MAAI,OAAO,MAAM,KAAK,QAAQ,CAAC,GAAG;AAChC,UAAM,IAAI,MAAM,6BAA6B,OAAO,EAAE;AAAA,EACxD;AACA,SAAO;AACT;AAEA,SAAS,6BAA6B,OAAoC;AACxE,UAAQ,OAAO;AAAA,IACb,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AAAA,IACL,KAAK;AACH,aAAO;AAAA,IACT;AACE,aAAO;AAAA,EACX;AACF;AAEA,SAAS,cAAc,QAA0B;AAC/C,SAAO,IAAI,OAAO,KAAK,GAAG,CAAC;AAC7B;AAEA,SAAS,gBAAgB,KAA6D;AACpF,SAAO;AAAA,IACL,IAAI,IAAI;AAAA,IACR,aAAa,IAAI;AAAA,IACjB,MAAM,IAAI;AAAA,IACV,aAAa,IAAI;AAAA,IACjB,WAAW,IAAI,UAAU,YAAY;AAAA,IACrC,WAAW,IAAI,UAAU,YAAY;AAAA,EACvC;AACF;AAEA,SAAS,YAAY,KAAqD;AACxE,SAAO;AAAA,IACL,IAAI,IAAI;AAAA,IACR,aAAa,IAAI;AAAA,IACjB,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,QAAQ,IAAI;AAAA,IACZ,OAAO,IAAI;AAAA,IACX,QAAQ,IAAI;AAAA,IACZ,YAAY,IAAI;AAAA,IAChB,OAAO,IAAI;AAAA,IACX,YAAY,6BAA6B,IAAI,UAAU;AAAA,IACvD,WAAW,IAAI;AAAA,IACf,kBAAkB,IAAI;AAAA,IACtB,aAAa,IAAI;AAAA,IACjB,cAAc,IAAI;AAAA,IAClB,iBAAiB,IAAI,iBAAiB,YAAY,KAAK;AAAA,IACvD,iBAAiB,IAAI,iBAAiB,YAAY,KAAK;AAAA,IACvD,eAAe,IAAI;AAAA,IACnB,SAAS,iBAAiB,IAAI,OAAO;AAAA,IACrC,WAAW,IAAI,UAAU,YAAY;AAAA,IACrC,WAAW,IAAI,UAAU,YAAY;AAAA,EACvC;AACF;","names":[]}

package/package.json

@@ -1,7 +1,8 @@
 {
   "name": "@opengeni/documents",
-  "version": "0.2.5",
+  "version": "0.2.8",
   "type": "module",
+  "license": "Apache-2.0",
   "main": "./dist/index.js",
   "module": "./dist/index.js",
   "types": "./dist/index.d.ts",
@@ -21,21 +22,18 @@
   },
   "scripts": {
     "build": "tsup",
-    "typecheck": "tsc --noEmit"
+    "typecheck": "tsc --noEmit",
+    "prepublishOnly": "bash ../../scripts/prepublish-guard"
   },
   "dependencies": {
-    "@opengeni/config": "^0.2.5",
-    "@opengeni/contracts": "^0.7.0",
-    "@opengeni/db": "^0.4.1",
-    "@opengeni/storage": "^0.2.5",
+    "@opengeni/config": "^0.4.0",
+    "@opengeni/contracts": "^0.9.0",
+    "@opengeni/db": "^0.6.1",
+    "@opengeni/storage": "^0.2.8",
     "@llamaindex/liteparse": "^1.5.3",
     "drizzle-orm": "^0.45.2",
     "openai": "6.36.0"
   },
-  "devDependencies": {
-    "tsup": "^8.5.0",
-    "typescript": "^6.0.3"
-  },
   "repository": {
     "type": "git",
     "url": "git+https://github.com/Cloudgeni-ai/opengeni.git",

package/src/index.ts

@@ -1,10 +1,19 @@
 import type { Settings } from "@opengeni/config";
-import type { AddDocumentRequest, CreateDocumentBaseRequest, Document, DocumentBase, DocumentSearchResult, FileAsset } from "@opengeni/contracts";
+import type {
+  AddDocumentRequest,
+  CreateDocumentBaseRequest,
+  Document,
+  DocumentBase,
+  DocumentSearchMode,
+  DocumentSearchResult,
+  FileAsset,
+  KnowledgeSourceKind,
+} from "@opengeni/contracts";
 import { requireFile, withRlsContext, withWorkspaceRls, type Database } from "@opengeni/db";
 import * as schema from "@opengeni/db/schema";
 import type { ObjectStorage } from "@opengeni/storage";
 import { LiteParse } from "@llamaindex/liteparse";
-import { and, asc, desc, eq, inArray, sql } from "drizzle-orm";
+import { and, asc, desc, eq, inArray, sql, type SQL } from "drizzle-orm";
 import OpenAI from "openai";
 
 export const DEFAULT_DOCUMENT_PARSER = "liteparse";
@@ -45,6 +54,16 @@ export type DocumentServices = {
   embedder: DocumentEmbedder;
 };
 
+export type DocumentSearchInput = {
+  workspaceId: string;
+  query: string;
+  baseIds?: string[] | undefined;
+  limit?: number | undefined;
+  mode?: DocumentSearchMode | undefined;
+  sourceKinds?: KnowledgeSourceKind[] | undefined;
+  aclTags?: string[] | undefined;
+};
+
 export type DocumentIndexHooks = {
   beforeEmbed?: (input: {
     accountId: string;
@@ -312,7 +331,23 @@ export async function addDocumentToBase(db: Database, input: AddDocumentRequest
       .where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.baseId, input.baseId), eq(schema.documents.fileId, input.fileId)))
       .limit(1);
     if (existing) {
-      return mapDocument(existing);
+      // Idempotent re-add: refresh caller-supplied source metadata on the
+      // existing row instead of silently discarding it (aclTags especially —
+      // a re-add that tightens tags must not be a no-op).
+      const [updated] = await scopedDb.update(schema.documents).set({
+        title: cleanString(input.title) ?? cleanString(input.sourceTitle) ?? existing.title,
+        ...(input.sourceKind !== undefined ? { sourceKind: input.sourceKind } : {}),
+        sourceUri: cleanString(input.sourceUri) ?? existing.sourceUri,
+        sourceExternalId: cleanString(input.sourceExternalId) ?? existing.sourceExternalId,
+        sourceTitle: cleanString(input.sourceTitle) ?? existing.sourceTitle,
+        sourceAuthor: cleanString(input.sourceAuthor) ?? existing.sourceAuthor,
+        sourceCreatedAt: parseOptionalDate(input.sourceCreatedAt) ?? existing.sourceCreatedAt,
+        sourceUpdatedAt: parseOptionalDate(input.sourceUpdatedAt) ?? existing.sourceUpdatedAt,
+        sourceVersion: cleanString(input.sourceVersion) ?? existing.sourceVersion,
+        ...(input.aclTags !== undefined ? { aclTags: cleanStringArray(input.aclTags) } : {}),
+        updatedAt: now,
+      }).where(and(eq(schema.documents.workspaceId, input.workspaceId), eq(schema.documents.id, existing.id))).returning();
+      return mapDocument(updated ?? existing);
     }
     const [row] = await scopedDb.insert(schema.documents).values({
       accountId: input.accountId,
@@ -320,8 +355,17 @@ export async function addDocumentToBase(db: Database, input: AddDocumentRequest
       baseId: input.baseId,
       fileId: input.fileId,
       status: "queued",
-      title: file.filename,
+      title: cleanString(input.title) ?? cleanString(input.sourceTitle) ?? file.filename,
       parser: DEFAULT_DOCUMENT_PARSER,
+      sourceKind: input.sourceKind ?? "manual_upload",
+      sourceUri: cleanString(input.sourceUri) ?? null,
+      sourceExternalId: cleanString(input.sourceExternalId) ?? null,
+      sourceTitle: cleanString(input.sourceTitle) ?? null,
+      sourceAuthor: cleanString(input.sourceAuthor) ?? null,
+      sourceCreatedAt: parseOptionalDate(input.sourceCreatedAt),
+      sourceUpdatedAt: parseOptionalDate(input.sourceUpdatedAt),
+      sourceVersion: cleanString(input.sourceVersion) ?? null,
+      aclTags: cleanStringArray(input.aclTags),
       updatedAt: now,
     }).returning();
     if (!row) throw new Error("Failed to create document");
@@ -422,7 +466,19 @@ export async function indexDocumentNow(
           fileId: file.id,
           chunkIndex: index,
           text: chunk.text,
-          metadata: chunk.metadata,
+          metadata: {
+            ...chunk.metadata,
+            documentTitle: document.title,
+            sourceKind: document.sourceKind,
+            sourceUri: document.sourceUri,
+            sourceExternalId: document.sourceExternalId,
+            sourceTitle: document.sourceTitle,
+            sourceAuthor: document.sourceAuthor,
+            sourceCreatedAt: document.sourceCreatedAt?.toISOString() ?? null,
+            sourceUpdatedAt: document.sourceUpdatedAt?.toISOString() ?? null,
+            sourceVersion: document.sourceVersion,
+            aclTags: document.aclTags,
+          },
           embedding: validateEmbedding(embeddings[index] ?? [], services.embedder.dimensions, services.embedder.model),
           embeddingModel: services.embedder.model,
         })));
@@ -453,10 +509,38 @@ export async function indexDocumentNow(
 
 export async function searchDocuments(
   db: Database,
-  input: { workspaceId: string; query: string; baseIds?: string[]; limit?: number },
+  input: DocumentSearchInput,
   services: Pick<DocumentServices, "embedder"> = createDocumentServices(),
 ): Promise<DocumentSearchResult[]> {
-  const limit = Math.min(Math.max(input.limit ?? 5, 1), 20);
+  const mode = input.mode ?? "hybrid";
+  const limit = Math.min(Math.max(input.limit ?? 5, 1), 50);
+  const candidateLimit = mode === "hybrid" ? Math.min(limit * 4, 100) : limit;
+  const rows: CombinedSearchRow[] = [];
+  if (mode === "vector" || mode === "hybrid") {
+    try {
+      rows.push(...await vectorSearchDocuments(db, input, candidateLimit, services));
+    } catch (error) {
+      if (mode === "vector") {
+        throw error;
+      }
+      console.warn("document hybrid search vector component failed; falling back to keyword search", {
+        workspaceId: input.workspaceId,
+        error: error instanceof Error ? error.message : String(error),
+      });
+    }
+  }
+  if (mode === "keyword" || mode === "hybrid") {
+    rows.push(...await keywordSearchDocuments(db, input, candidateLimit));
+  }
+  return mergeDocumentSearchRows(rows, mode).slice(0, limit);
+}
+
+async function vectorSearchDocuments(
+  db: Database,
+  input: DocumentSearchInput,
+  limit: number,
+  services: Pick<DocumentServices, "embedder">,
+): Promise<CombinedSearchRow[]> {
   const queryEmbedding = await services.embedder.embedQuery(input.query);
   validateEmbedding(queryEmbedding, services.embedder.dimensions, services.embedder.model);
   const distance = sql<number>`${schema.documentChunks.embedding} <=> ${vectorLiteral(queryEmbedding)}::vector`;
@@ -470,29 +554,64 @@ export async function searchDocuments(
       text: schema.documentChunks.text,
       chunkIndex: schema.documentChunks.chunkIndex,
       metadata: schema.documentChunks.metadata,
+      sourceKind: schema.documents.sourceKind,
+      sourceUri: schema.documents.sourceUri,
+      sourceExternalId: schema.documents.sourceExternalId,
+      sourceTitle: schema.documents.sourceTitle,
+      sourceAuthor: schema.documents.sourceAuthor,
+      sourceCreatedAt: schema.documents.sourceCreatedAt,
+      sourceUpdatedAt: schema.documents.sourceUpdatedAt,
+      sourceVersion: schema.documents.sourceVersion,
+      aclTags: schema.documents.aclTags,
       distance,
+    }).from(schema.documentChunks)
+      .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))
+      .where(and(...documentSearchConditions(input, services.embedder.model)))
+      .orderBy(distance)
+      .limit(limit)
+  );
+  return rows.map((row) => ({
+    ...mapSearchRowBase(row, input.workspaceId),
+    vectorScore: 1 / (1 + Number(row.distance)),
+    keywordScore: null,
+  }));
+}
+
+async function keywordSearchDocuments(db: Database, input: DocumentSearchInput, limit: number): Promise<CombinedSearchRow[]> {
+  const rank = sql<number>`ts_rank_cd(to_tsvector('simple', ${schema.documentChunks.text}), plainto_tsquery('simple', ${input.query}))`;
+  const rows = await withWorkspaceRls(db, input.workspaceId, async (scopedDb) =>
+    await scopedDb.select({
+      chunkId: schema.documentChunks.id,
+      documentId: schema.documentChunks.documentId,
+      baseId: schema.documentChunks.baseId,
+      fileId: schema.documentChunks.fileId,
+      title: schema.documents.title,
+      text: schema.documentChunks.text,
+      chunkIndex: schema.documentChunks.chunkIndex,
+      metadata: schema.documentChunks.metadata,
+      sourceKind: schema.documents.sourceKind,
+      sourceUri: schema.documents.sourceUri,
+      sourceExternalId: schema.documents.sourceExternalId,
+      sourceTitle: schema.documents.sourceTitle,
+      sourceAuthor: schema.documents.sourceAuthor,
+      sourceCreatedAt: schema.documents.sourceCreatedAt,
+      sourceUpdatedAt: schema.documents.sourceUpdatedAt,
+      sourceVersion: schema.documents.sourceVersion,
+      aclTags: schema.documents.aclTags,
+      rank,
     }).from(schema.documentChunks)
       .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))
       .where(and(
-        eq(schema.documents.status, "ready"),
-        eq(schema.documentChunks.workspaceId, input.workspaceId),
-        eq(schema.documentChunks.embeddingModel, services.embedder.model),
-        input.baseIds && input.baseIds.length > 0 ? inArray(schema.documentChunks.baseId, input.baseIds) : undefined,
+        ...documentSearchConditions(input),
+        sql`to_tsvector('simple', ${schema.documentChunks.text}) @@ plainto_tsquery('simple', ${input.query})`,
       ))
-      .orderBy(distance)
+      .orderBy(desc(rank))
       .limit(limit)
   );
   return rows.map((row) => ({
-    chunkId: row.chunkId,
-    workspaceId: input.workspaceId,
-    documentId: row.documentId,
-    baseId: row.baseId,
-    fileId: row.fileId,
-    title: row.title,
-    text: row.text,
-    score: 1 / (1 + Number(row.distance)),
-    chunkIndex: row.chunkIndex,
-    metadata: row.metadata,
+    ...mapSearchRowBase(row, input.workspaceId),
+    vectorScore: null,
+    keywordScore: normalizeKeywordScore(Number(row.rank)),
   }));
 }
 
@@ -507,12 +626,76 @@ export async function getDocumentChunk(db: Database, workspaceId: string, chunkI
       text: schema.documentChunks.text,
       chunkIndex: schema.documentChunks.chunkIndex,
       metadata: schema.documentChunks.metadata,
+      sourceKind: schema.documents.sourceKind,
+      sourceUri: schema.documents.sourceUri,
+      sourceExternalId: schema.documents.sourceExternalId,
+      sourceTitle: schema.documents.sourceTitle,
+      sourceAuthor: schema.documents.sourceAuthor,
+      sourceCreatedAt: schema.documents.sourceCreatedAt,
+      sourceUpdatedAt: schema.documents.sourceUpdatedAt,
+      sourceVersion: schema.documents.sourceVersion,
+      aclTags: schema.documents.aclTags,
     }).from(schema.documentChunks)
       .innerJoin(schema.documents, eq(schema.documentChunks.documentId, schema.documents.id))
       .where(and(eq(schema.documentChunks.workspaceId, workspaceId), eq(schema.documentChunks.id, chunkId), eq(schema.documents.status, "ready")))
       .limit(1)
   );
   if (!row) return null;
+  return {
+    ...mapSearchRowBase(row, workspaceId),
+    score: 1,
+    matchType: "hybrid",
+    vectorScore: null,
+    keywordScore: null,
+  };
+}
+
+type SearchRowBase = Omit<DocumentSearchResult, "score" | "matchType" | "vectorScore" | "keywordScore">;
+type CombinedSearchRow = SearchRowBase & {
+  vectorScore: number | null;
+  keywordScore: number | null;
+};
+
+function documentSearchConditions(input: DocumentSearchInput, embeddingModel?: string): SQL[] {
+  const conditions: SQL[] = [
+    eq(schema.documents.status, "ready"),
+    eq(schema.documentChunks.workspaceId, input.workspaceId),
+  ];
+  if (embeddingModel) {
+    conditions.push(eq(schema.documentChunks.embeddingModel, embeddingModel));
+  }
+  if (input.baseIds?.length) {
+    conditions.push(inArray(schema.documentChunks.baseId, input.baseIds));
+  }
+  if (input.sourceKinds?.length) {
+    conditions.push(inArray(schema.documents.sourceKind, input.sourceKinds));
+  }
+  const aclTags = cleanStringArray(input.aclTags);
+  if (aclTags.length > 0) {
+    conditions.push(sql`${schema.documents.aclTags} @> ${JSON.stringify(aclTags)}::jsonb`);
+  }
+  return conditions;
+}
+
+function mapSearchRowBase(row: {
+  chunkId: string;
+  documentId: string;
+  baseId: string;
+  fileId: string;
+  title: string;
+  text: string;
+  chunkIndex: number;
+  metadata: Record<string, unknown>;
+  sourceKind: string;
+  sourceUri: string | null;
+  sourceExternalId: string | null;
+  sourceTitle: string | null;
+  sourceAuthor: string | null;
+  sourceCreatedAt: Date | null;
+  sourceUpdatedAt: Date | null;
+  sourceVersion: string | null;
+  aclTags: string[];
+}, workspaceId: string): SearchRowBase {
   return {
     chunkId: row.chunkId,
     workspaceId,
@@ -521,12 +704,79 @@ export async function getDocumentChunk(db: Database, workspaceId: string, chunkI
     fileId: row.fileId,
     title: row.title,
     text: row.text,
-    score: 1,
     chunkIndex: row.chunkIndex,
     metadata: row.metadata,
+    sourceKind: normalizeKnowledgeSourceKind(row.sourceKind),
+    sourceUri: row.sourceUri,
+    sourceExternalId: row.sourceExternalId,
+    sourceTitle: row.sourceTitle,
+    sourceAuthor: row.sourceAuthor,
+    sourceCreatedAt: row.sourceCreatedAt?.toISOString() ?? null,
+    sourceUpdatedAt: row.sourceUpdatedAt?.toISOString() ?? null,
+    sourceVersion: row.sourceVersion,
+    aclTags: cleanStringArray(row.aclTags),
   };
 }
 
+function mergeDocumentSearchRows(rows: CombinedSearchRow[], mode: DocumentSearchMode): DocumentSearchResult[] {
+  const byChunk = new Map<string, CombinedSearchRow>();
+  for (const row of rows) {
+    const existing = byChunk.get(row.chunkId);
+    if (!existing) {
+      byChunk.set(row.chunkId, row);
+      continue;
+    }
+    byChunk.set(row.chunkId, {
+      ...existing,
+      vectorScore: Math.max(existing.vectorScore ?? 0, row.vectorScore ?? 0) || null,
+      keywordScore: Math.max(existing.keywordScore ?? 0, row.keywordScore ?? 0) || null,
+    });
+  }
+  return [...byChunk.values()].map((row) => {
+    const vectorScore = row.vectorScore;
+    const keywordScore = row.keywordScore;
+    const matchType: DocumentSearchMode = vectorScore !== null && keywordScore !== null
+      ? "hybrid"
+      : vectorScore !== null
+        ? "vector"
+        : "keyword";
+    return {
+      ...row,
+      score: combinedSearchScore(mode, vectorScore, keywordScore, matchType),
+      matchType,
+    };
+  }).sort((left, right) =>
+    right.score - left.score
+    || (right.vectorScore ?? 0) - (left.vectorScore ?? 0)
+    || (right.keywordScore ?? 0) - (left.keywordScore ?? 0)
+    || left.chunkIndex - right.chunkIndex
+  );
+}
+
+function combinedSearchScore(
+  mode: DocumentSearchMode,
+  vectorScore: number | null,
+  keywordScore: number | null,
+  matchType: DocumentSearchMode,
+): number {
+  const vector = vectorScore ?? 0;
+  const keyword = keywordScore ?? 0;
+  if (mode === "vector") return roundScore(vector);
+  if (mode === "keyword") return roundScore(keyword);
+  return roundScore(Math.min(1, (0.65 * vector) + (0.35 * keyword) + (matchType === "hybrid" ? 0.1 : 0)));
+}
+
+function normalizeKeywordScore(rank: number): number {
+  if (!Number.isFinite(rank) || rank <= 0) {
+    return 0;
+  }
+  return rank / (rank + 1);
+}
+
+function roundScore(value: number): number {
+  return Number(value.toFixed(6));
+}
+
 export async function parseDocumentBytes(bytes: Uint8Array, file: FileAsset, parser: DocumentParser = new LiteParseDocumentParser()): Promise<ParsedDocument> {
   return await parser.parse(bytes, file);
 }
@@ -635,6 +885,43 @@ function withOverlap(previous: string, overlapChars: number, next: string, maxCh
   return candidate.length <= maxChars ? candidate : next;
 }
 
+function cleanString(value: string | undefined | null): string | undefined {
+  const trimmed = value?.trim();
+  return trimmed ? trimmed : undefined;
+}
+
+function cleanStringArray(values: string[] | undefined | null): string[] {
+  return [...new Set((values ?? []).map((value) => value.trim()).filter(Boolean))];
+}
+
+function parseOptionalDate(value: string | undefined): Date | null {
+  const trimmed = cleanString(value);
+  if (!trimmed) {
+    return null;
+  }
+  const date = new Date(trimmed);
+  if (Number.isNaN(date.getTime())) {
+    throw new Error(`Invalid source timestamp: ${trimmed}`);
+  }
+  return date;
+}
+
+function normalizeKnowledgeSourceKind(value: string): KnowledgeSourceKind {
+  switch (value) {
+    case "manual_upload":
+    case "meeting_transcript":
+    case "repository":
+    case "email":
+    case "chat":
+    case "document":
+    case "web":
+    case "other":
+      return value;
+    default:
+      return "other";
+  }
+}
+
 function vectorLiteral(values: number[]): string {
   return `[${values.join(",")}]`;
 }
@@ -661,6 +948,15 @@ function mapDocument(row: typeof schema.documents.$inferSelect): Document {
     parser: row.parser,
     chunkCount: row.chunkCount,
     error: row.error,
+    sourceKind: normalizeKnowledgeSourceKind(row.sourceKind),
+    sourceUri: row.sourceUri,
+    sourceExternalId: row.sourceExternalId,
+    sourceTitle: row.sourceTitle,
+    sourceAuthor: row.sourceAuthor,
+    sourceCreatedAt: row.sourceCreatedAt?.toISOString() ?? null,
+    sourceUpdatedAt: row.sourceUpdatedAt?.toISOString() ?? null,
+    sourceVersion: row.sourceVersion,
+    aclTags: cleanStringArray(row.aclTags),
     createdAt: row.createdAt.toISOString(),
     updatedAt: row.updatedAt.toISOString(),
   };