@autorag/librarian 2.11.0 → 2.13.0

package/dist/datasource/connector-timeouts.d.ts

@@ -9,8 +9,12 @@
  * of making progress. Interactive reads (`search`, `doctor`, chunk lookups) keep
  * the short default.
  */
-/** Non-interactive budget for a connector `sync`/`index` step (30 minutes). */
-export declare const DEFAULT_CONNECTOR_SYNC_TIMEOUT_MS = 1800000;
+/**
+ * Non-interactive budget for a connector `sync`/`index` step (one hour). A no-op mailcrawl `sync` of a
+ * 1,000-message INBOX measured 16 to 27 minutes because mailcrawl re-reads every message on each run
+ * (NomaDamas/mailcrawl#48), so 30 minutes did not hold even when nothing had changed.
+ */
+export declare const DEFAULT_CONNECTOR_SYNC_TIMEOUT_MS = 3600000;
 /**
  * Operator-facing hint appended to a sync/index diagnostic when the connector
  * timeout killed the step. It names the override and explains why the next

package/dist/datasource/crawler-skill.d.ts

@@ -58,6 +58,7 @@ export declare class CrawlerDatasourceSkill implements DatasourceSkill {
     private syncSidecar;
     /** Release the sidecar's database handles. Safe to call repeatedly; later use reopens them. */
     close(): void;
+    /** One method: the sidecar's hybrid search (or BM25 without an embedder), with the CLI as the last fallback. */
     retrievalMethods(): readonly RetrievalMethod[];
     private cliMethod;
     describeSources(): readonly SourceDescription[];

package/dist/datasource/crawler-types.d.ts

@@ -45,7 +45,7 @@ export interface CrawlerCliOptions {
      */
     readonly timeoutMs?: number;
     /**
-     * Spawn timeout for `sync` in milliseconds. Default 1_800_000 (30 min): a first
+     * Spawn timeout for `sync` in milliseconds. Default 3_600_000 (1 hour): a first
      * full import over a real archive can take many minutes, and a timed-out sync
      * restarts from scratch on the next refresh.
      */

package/dist/datasource/index.d.ts

@@ -13,7 +13,7 @@ export { GitHubConnector, type GitHubConnectorOptions, GitHubSkill, type GitHubS
 export { HISTER_SKILL_TYPE, HisterClient, type HisterClientOptions, HisterSkill, type HisterSkillOptions, } from "./skills/hister/index.js";
 export { LarkClient, type LarkClientOptions, LarkSkill, type LarkSkillOptions } from "./skills/lark/index.js";
 export type { LazykatokFailureReason, LazykatokHit, LazykatokOptions, LazykatokSearchMode, LazykatokSearchOptions, LazykatokSearchResult, LazykatokSkillClient, LazykatokSkillOptions, } from "./skills/lazykatok/index.js";
-export { LazykatokBm25Method, LazykatokClient, LazykatokSemanticMethod, LazykatokSkill, } from "./skills/lazykatok/index.js";
+export { LazykatokBm25Method, LazykatokClient, LazykatokHybridMethod, LazykatokSkill, } from "./skills/lazykatok/index.js";
 export { MailExportConnector, type MailExportConnectorOptions, MailExportSkill, type MailExportSkillOptions, } from "./skills/mail-export/index.js";
 export { MAILCRAWL_MODES, MailcrawlClient, type MailcrawlFailure, type MailcrawlFailureReason, type MailcrawlIndexInfo, MailcrawlMethod, type MailcrawlOptions, type MailcrawlSearchHit, type MailcrawlSearchMode, type MailcrawlSearchResult, MailcrawlSkill, type MailcrawlSyncInfo, type MailcrawlSyncResult, } from "./skills/mailcrawl/index.js";
 export { NotcrawlClient, type NotcrawlOptions, NotionSkill, type NotionSkillOptions, } from "./skills/notion/index.js";

package/dist/datasource/sidecar/methods.d.ts

@@ -6,7 +6,7 @@
 import type { RetrievalMethod, RetrievalMethodDescriptor, RetrievalOptions, RetrievalResult } from "../../retrieval/types.js";
 import type { CrawlerDescriptor } from "./descriptors.js";
 import type { CrawlerSidecar } from "./index.js";
-export type SidecarMode = "bm25" | "semantic" | "hybrid";
+export type SidecarMode = "bm25" | "hybrid";
 export interface SidecarMethodOptions {
     readonly sidecar: CrawlerSidecar;
     readonly descriptor: CrawlerDescriptor;
@@ -25,3 +25,10 @@ export declare class SidecarRetrievalMethod implements RetrievalMethod {
     retrieve(query: string, options: RetrievalOptions): Promise<RetrievalResult[]>;
     private toResult;
 }
+/**
+ * The one method a sidecar-backed crawler registers. The sidecar's hybrid search leads when semantic
+ * search is enabled and its BM25 search leads otherwise. When the leader throws, the sidecar's own BM25
+ * answers next and the crawler's CLI search last, so the datasource still searches when the embedder or
+ * the sidecar index is down. A leader that answers with no hits is final.
+ */
+export declare function sidecarFirstMethod(options: Omit<SidecarMethodOptions, "mode">, cli: RetrievalMethod): RetrievalMethod;

package/dist/datasource/skills/clawgallery/skill.d.ts

@@ -29,6 +29,10 @@ export declare class ClawGallerySkill implements DatasourceSkill {
     describe(): DatasourceSkillDescriptor;
     polling(): PollingMetadata;
     index(options?: DatasourceIndexOptions): Promise<DatasourceIndexResult>;
+    /**
+     * Hybrid RRF-fuses every channel the index has, so it runs alone and keyword search stands in only
+     * when hybrid fails. A mode the operator pinned explicitly runs by itself with no fallback.
+     */
     retrievalMethods(): readonly RetrievalMethod[];
     describeSources(): readonly SourceDescription[];
     skillManifest(): DatasourceSkillManifest;

package/dist/datasource/skills/discrawl/skill.d.ts

@@ -48,6 +48,11 @@ export declare class DiscrawlSkill implements DatasourceSkill {
     describe(): DatasourceSkillDescriptor;
     polling(): PollingMetadata;
     index(options?: DatasourceIndexOptions): Promise<DatasourceIndexResult>;
+    /**
+     * One method per query. Hybrid is the default and already degrades to FTS inside the single
+     * `discrawl search` call when embeddings are unavailable, so FTS and semantic are not registered
+     * beside it; an operator who pins `defaultMode` to `fts` or `semantic` gets exactly that mode.
+     */
     retrievalMethods(): readonly RetrievalMethod[];
     describeSources(): readonly SourceDescription[];
     skillManifest(): DatasourceSkillManifest;

package/dist/datasource/skills/factory.d.ts

@@ -50,7 +50,11 @@ export interface DatasourceEmbeddingOptions {
     /** MinSync binary, embedder and chunking shared with the local-file index; vault indexes run on it too. */
     readonly minSync?: ObsidianMinSyncOptions;
 }
-/** Skill names this factory can build. */
+/** Whether the skill's backend exists on `platform`. Names with no platform restriction are always supported. */
+export declare function datasourceSkillSupported(name: string, platform?: NodeJS.Platform): boolean;
+/** Skill names this factory can build on `platform`. */
+export declare function builtinDatasourceSkillNames(platform?: NodeJS.Platform): readonly string[];
+/** Skill names this factory can build on the running platform. */
 export declare const BUILTIN_DATASOURCE_SKILL_NAMES: readonly string[];
 /**
  * Build datasource skills from the trusted config section. `false` or

package/dist/datasource/skills/lazykatok/index.d.ts

@@ -6,7 +6,7 @@
  * the form `/kakao/<instance-id>/chunks/<chunk-id>`.
  */
 export { LazykatokClient } from "./client.js";
-export { LazykatokBm25Method, type LazykatokMethodOptions, type LazykatokSearchClient, LazykatokSemanticMethod, } from "./methods.js";
+export { LazykatokBm25Method, LazykatokHybridMethod, type LazykatokMethodOptions, type LazykatokSearchClient, } from "./methods.js";
 export { LazykatokSkill, type LazykatokSkillClient, type LazykatokSkillOptions } from "./skill.js";
 export type { LazykatokChunkResult, LazykatokDoctorResult, LazykatokFailureReason, LazykatokHit, LazykatokIndexResult, LazykatokOptions, LazykatokSearchMode, LazykatokSearchOptions, LazykatokSearchResult, LazykatokSyncResult, } from "./types.js";
 export { DEFAULT_LAZYKATOK_INDEX_TIMEOUT_MS, DEFAULT_LAZYKATOK_OPTIONS } from "./types.js";

package/dist/datasource/skills/lazykatok/methods.d.ts

@@ -27,12 +27,12 @@ export declare class LazykatokBm25Method implements RetrievalMethod {
     retrieve(query: string, options: RetrievalOptions): Promise<RetrievalResult[]>;
 }
 /**
- * Semantic (vector) retrieval over a KakaoTalk datasource via the external
- * `lazykatok` CLI in semantic mode. Lazykatok does not expose source scopes; access is
- * controlled at the datasource/tag level. All client failures collapse to an
- * empty result set; retrieval never throws.
+ * Hybrid retrieval over a KakaoTalk datasource via the external `lazykatok` CLI in hybrid mode
+ * (BM25 and semantic rankings fused by reciprocal rank fusion inside the CLI). Lazykatok does not
+ * expose source scopes; access is controlled at the datasource/tag level. Client failures reach the
+ * caller verbatim so the pipeline can fall back to {@link LazykatokBm25Method}.
  */
-export declare class LazykatokSemanticMethod implements RetrievalMethod {
+export declare class LazykatokHybridMethod implements RetrievalMethod {
     private readonly client;
     private readonly instanceId;
     private readonly tags;

package/dist/datasource/skills/lazykatok/skill.d.ts

@@ -38,6 +38,7 @@ export declare class LazykatokSkill implements DatasourceSkill {
     index(options?: DatasourceIndexOptions): Promise<DatasourceIndexResult>;
     /** Release the sidecar's database handles. Safe to call repeatedly; later use reopens them. */
     close(): void;
+    /** One method: sidecar hybrid when a sidecar is configured, else the CLI's hybrid; BM25 is the fallback either way. */
     retrievalMethods(): readonly RetrievalMethod[];
     private syncSidecar;
     describeSources(): readonly SourceDescription[];

package/dist/datasource/skills/lazykatok/types.d.ts

@@ -23,7 +23,7 @@ export interface LazykatokOptions {
      */
     readonly timeoutMs?: number;
     /**
-     * Spawn timeout for `sync`/`index` in milliseconds. Default 1_800_000 (30 min):
+     * Spawn timeout for `sync`/`index` in milliseconds. Default 3_600_000 (1 hour):
      * the first semantic index over a large archive was measured at ~11 minutes, and
      * a timed-out build restarts from scratch on the next refresh.
      */
@@ -59,7 +59,7 @@ export declare const DEFAULT_LAZYKATOK_TIMEOUT_MS = 60000;
  * the 60-second interactive default: a first semantic index over a large archive
  * was measured at ~11 minutes.
  */
-export declare const DEFAULT_LAZYKATOK_INDEX_TIMEOUT_MS = 1800000;
+export declare const DEFAULT_LAZYKATOK_INDEX_TIMEOUT_MS = 3600000;
 /** The live macOS adapter `sync` names by default on darwin. */
 export declare const DEFAULT_LAZYKATOK_SOURCE = "macos";
 export declare const DEFAULT_LAZYKATOK_MAX_BUFFER_BYTES: number;

package/dist/datasource/skills/mailcrawl/skill.d.ts

@@ -27,6 +27,7 @@ export declare class MailcrawlSkill implements DatasourceSkill {
     describe(): DatasourceSkillDescriptor;
     polling(): PollingMetadata;
     index(options?: DatasourceIndexOptions): Promise<DatasourceIndexResult>;
+    /** Hybrid search first; BM25 only when mailcrawl cannot run hybrid (for example, no usable semantic index). */
     retrievalMethods(): readonly RetrievalMethod[];
     describeSources(): readonly SourceDescription[];
     skillManifest(): DatasourceSkillManifest;

package/dist/datasource/skills/mailcrawl/types.d.ts

@@ -19,7 +19,7 @@ export interface MailcrawlOptions {
      * `indexTimeoutMs`, because a 0.2.0 native index run is not interactive.
      */
     readonly timeoutMs?: number;
-    /** Non-interactive budget for `sync`/`index`; defaults to 30 minutes. */
+    /** Non-interactive budget for `sync`/`index`; defaults to one hour. */
     readonly indexTimeoutMs?: number;
     readonly maxBufferBytes?: number;
     readonly env?: Readonly<Record<string, string | undefined>>;

package/dist/dupey/cli.d.ts

@@ -38,3 +38,21 @@ export declare class DupeyCliError extends Error {
     constructor(message: string);
 }
 export declare function scanWithDupey(dir: string, options?: DupeyCliOptions): Promise<DupeyScanResult>;
+export type DupeyRelation = "exact" | "near" | "contains" | "different";
+export type DupeyCompareResult = {
+    readonly relation: DupeyRelation;
+    readonly nearScore: number;
+    readonly jaccard: number;
+    readonly containmentAInB?: number;
+    readonly containmentBInA?: number;
+}
+/** dupey cannot extract text from one of the files (for example HTML or an image), so no family can form. */
+ | {
+    readonly relation: "unsupported";
+};
+/**
+ * Would `dupey scan` put these two files in one family? Classifies with the
+ * same default thresholds the scan uses, so the answer matches what a later
+ * scan of a directory holding both will report.
+ */
+export declare function compareWithDupey(a: string, b: string, options?: DupeyCliOptions): Promise<DupeyCompareResult>;

package/dist/index.d.ts

@@ -2,6 +2,7 @@ export { AutoRAGAgent, type AutoRAGAgentOptions, type AutoRAGRefreshComponentSta
 export { type AutoRAGResultsDetails, reportSchema, type SearchDocumentDiagnostic, type SearchDocumentDiagnosticCode, type SearchDocumentDiagnosticSeverity, type SearchDocumentsResponse, type SearchDocumentsStreamEvent, type SearchDocumentWarning, } from "./agent/index.js";
 export { buildSystemPrompt, type SystemPromptConfig } from "./agent/system-prompt.js";
 export * from "./cloud/index.js";
+export * from "./convert/index.js";
 export { AutoRAGLite, type AutoRAGLiteOptions, buildAgentOptions, type CliConfig, createAutoRAGLite, type ResolveConfigInput, resolveConfig, resolveConfigReadOnly, writeConfigObject, writeDefaultConfig, } from "./core.js";
 export * from "./datasource/index.js";
 export * from "./dupey/index.js";