@autorag/librarian 2.6.1 → 2.6.3

package/dist/minsync/method.d.ts

@@ -14,6 +14,10 @@ export interface MinSyncVectorMethodOptions {
     readonly maxChunkSize?: number;
     readonly mode?: MinSyncQueryMode;
     readonly runtime?: MinSyncRuntime;
+    /** Default `topK` used when the caller omits one. Default 50. */
+    readonly defaultTopK?: number;
+    /** Fetch cap applied when a scope narrows the query. Default 100. */
+    readonly scopedTopK?: number;
 }
 export type MinSyncHybridMethodOptions = Omit<MinSyncVectorMethodOptions, "mode">;
 export declare class MinSyncVectorMethod implements RetrievalMethod {
@@ -26,6 +30,8 @@ export declare class MinSyncVectorMethod implements RetrievalMethod {
     private readonly maxChunkSize;
     private readonly mode;
     private readonly runtime;
+    private readonly defaultTopK;
+    private readonly scopedTopK;
     private installFailed;
     constructor(options: MinSyncVectorMethodOptions);
     describe(): RetrievalMethodDescriptor;

package/dist/mirror/index-store.d.ts

@@ -8,7 +8,7 @@ export interface ParsedMirrorEntry {
     readonly updatedAt: string;
 }
 /** Why refresh deliberately left a source out of the parsed mirror. */
-export type ParsedMirrorSkipReason = "duplicate-excluded" | "parser-skipped" | "parser-failed";
+export type ParsedMirrorSkipReason = "duplicate-excluded" | "user-excluded" | "parser-skipped" | "parser-failed";
 /**
  * A source the last refresh decided not to mirror, pinned to the exact version it
  * decided on. Staleness checks compare against this record instead of re-deriving

package/dist/mirror/index.d.ts

@@ -2,4 +2,4 @@ export type { ParsedMirrorEntry, ParsedMirrorIndex } from "./index-store.js";
 export { emptyMirrorIndex, loadMirrorIndex, saveMirrorIndex } from "./index-store.js";
 export { parsedMirrorIndexPath, parsedMirrorRoot, parsedOutputPath, refreshReadinessPath } from "./paths.js";
 export type { ParsedMirrorDiagnostic, ParsedMirrorDiagnosticCode, ParsedMirrorSyncOptions, ParsedMirrorSyncResult, } from "./sync.js";
-export { DEFAULT_MAX_SOURCE_BYTES, detectMirrorStaleness, syncParsedMirrors } from "./sync.js";
+export { DEFAULT_MAX_SOURCE_BYTES, detectMirrorStaleness, isPathExcluded, syncParsedMirrors } from "./sync.js";

package/dist/mirror/sync.d.ts

@@ -12,6 +12,8 @@ export interface ParsedMirrorSyncOptions {
     readonly maxSourceBytes?: number;
     /** Absolute source paths omitted from parsed mirrors (for example exact duplicate copies). */
     readonly excludeSourcePaths?: ReadonlySet<string>;
+    /** Absolute source paths or directories omitted by the user. */
+    readonly userExcludedSourcePaths?: ReadonlySet<string>;
 }
 export type ParsedMirrorDiagnosticCode = "unsupported-file" | "parser-skipped" | "parser-failed" | "duplicate-excluded" | "deleted-mirror" | "stale-index" | "pdf-extract-thin" | "parser-warning";
 /** Path-opaque refresh diagnostic. `source` is an opaque virtual path, never a real fs path. */
@@ -42,3 +44,4 @@ export declare function syncParsedMirrors(options: ParsedMirrorSyncOptions): Pro
  * getRefreshStatus).
  */
 export declare function detectMirrorStaleness(options: ParsedMirrorSyncOptions): Promise<ParsedMirrorDiagnostic[]>;
+export declare function isPathExcluded(sourcePath: string, excluded: ReadonlySet<string> | undefined): boolean;

package/dist/retrieval/engine.d.ts

@@ -15,6 +15,7 @@
  */
 import { DatasourceAccessContext, type DatasourceAccessContextOptions } from "../datasource/access-context.js";
 import { RetrievalMethodRegistry } from "./registry.js";
+import type { Reranker } from "./rerank.js";
 import type { RetrievalDiagnostic, RetrievalMethod, RetrievalOptions, RetrievalResult, RetrievalUnsearchedSurface } from "./types.js";
 /** Options for constructing a standalone {@link RetrievalEngine}. */
 export interface RetrievalEngineOptions {
@@ -40,6 +41,13 @@ export interface RetrievalEngineOptions {
      * preservation — matching the existing {@link AutoRAGAgent} behavior.
      */
     readonly isMinSyncBinaryMissing?: () => boolean;
+    /**
+     * Optional post-merge reranker. When set, merged results are reordered by
+     * the reranker after dedup and score normalization. A reranker failure is
+     * reported as a `rerank-failed` diagnostic and the unranked order is kept —
+     * the engine never silently drops evidence because a reranker was down.
+     */
+    readonly reranker?: Reranker;
 }
 /**
  * Model-free retrieval pipeline.
@@ -71,6 +79,7 @@ export declare class RetrievalEngine {
     private readonly defaultTopK;
     private readonly defaultDedup;
     private readonly isMinSyncBinaryMissing;
+    private readonly reranker;
     constructor(options?: RetrievalEngineOptions);
     /** The underlying method registry. Intentionally public for tooling. */
     getMethodRegistry(): RetrievalMethodRegistry;
@@ -100,6 +109,11 @@ export declare class RetrievalEngine {
         diagnostics: RetrievalDiagnostic[];
         unsearched: RetrievalUnsearchedSurface[];
     }>;
+    /**
+     * Reorder merged results with the configured reranker. On failure the
+     * unranked order is returned and a `rerank-failed` diagnostic is appended.
+     */
+    private applyRerank;
     /**
      * Run retrieval and return results keyed per method (unmerged) with
      * diagnostics. Unlike {@link retrieve}, this does NOT merge or sort —

package/dist/retrieval/index.d.ts

@@ -2,6 +2,7 @@ export { RetrievalEngine, type RetrievalEngineOptions } from "./engine.js";
 export type { MergeOptions } from "./merger.js";
 export { ParallelRetriever, ResultMerger } from "./merger.js";
 export { RetrievalMethodRegistry } from "./registry.js";
+export { type CreateRerankerOptions, createReranker, DEFAULT_RERANK_API_KEY_ENV, DEFAULT_RERANK_MODEL, DEFAULT_RERANK_PROVIDER, OpenRouterReranker, type OpenRouterRerankerOptions, type Reranker, type RerankerDescriptor, type RerankOptions, } from "./rerank.js";
 export { matchesVirtualPathScope, normalizeVirtualPath, normalizeVirtualPathScope, virtualPathScopeToRegExp, } from "./scope.js";
 export { describeRetrievalError, groupUnsearchedSurfaces, MINSYNC_SURFACE, type RetrievalSkip, retrievalSurfaceFor, } from "./skip.js";
 export type { CuratedResult, NumberedResult, RetrievalDiagnostic, RetrievalDiagnosticCode, RetrievalMethod, RetrievalMethodDescriptor, RetrievalOptions, RetrievalResult, RetrievalUnsearchedSurface, RetrievalWithDiagnostics, } from "./types.js";

package/dist/retrieval/rerank.d.ts

@@ -0,0 +1,114 @@
+/**
+ * Reranker seam — a post-retrieval reordering stage over merged evidence.
+ *
+ * The retrieval pipeline (registry → ParallelRetriever → DatasourceResultFilter
+ * → ResultMerger) produces a scored candidate list. A reranker reorders that
+ * list with a dedicated relevance model. The default implementation talks to
+ * OpenRouter's rerank router through the OpenRouter TypeScript SDK.
+ *
+ * Local-first is no longer a contract: a remote reranker is opt-in through
+ * trusted config, and a local reranker can implement this same interface
+ * without touching the pipeline. Nothing here silently falls back between
+ * providers — a configured reranker that cannot run is reported, not replaced.
+ */
+import type { RetrievalResult } from "./types.js";
+/** Default rerank model routed through OpenRouter. */
+export declare const DEFAULT_RERANK_MODEL = "voyageai/rerank-3-lite";
+/** Default rerank provider id. */
+export declare const DEFAULT_RERANK_PROVIDER = "openrouter";
+/** Default environment variable holding the OpenRouter API key. */
+export declare const DEFAULT_RERANK_API_KEY_ENV = "OPENROUTER_API_KEY";
+/** Static description of a reranker, used for diagnostics and inspection. */
+export interface RerankerDescriptor {
+    /** Stable label (e.g. `openrouter`). */
+    readonly name: string;
+    /** Provider id (e.g. `openrouter`). */
+    readonly provider: string;
+    /** Wire model id (e.g. `cohere/rerank-4-fast`). */
+    readonly model: string;
+    /** Whether the reranker has everything it needs to run (credentials, config). */
+    readonly available: boolean;
+    /** Why the reranker is unavailable, when `available` is false. */
+    readonly reason?: string;
+}
+/** Per-call rerank options. */
+export interface RerankOptions {
+    /** Return only the top N results. Omitted ⇒ every distinct result is returned, reordered. */
+    readonly topN?: number;
+    /** Abort the in-flight rerank request. */
+    readonly signal?: AbortSignal;
+}
+/** A post-retrieval reordering stage. */
+export interface Reranker {
+    describe(): RerankerDescriptor;
+    /**
+     * Reorder `results` by relevance to `query`. Returns a new array; the input
+     * is not mutated. Implementations must preserve each result's `source`/`id`
+     * identity so evidence stays traceable.
+     */
+    rerank(query: string, results: readonly RetrievalResult[], options?: RerankOptions): Promise<RetrievalResult[]>;
+}
+/** Minimal rerank client shape, so tests can inject a fake without the SDK. */
+export interface RerankClient {
+    rerank(request: {
+        requestBody: {
+            model: string;
+            query: string;
+            documents: string[];
+            topN?: number;
+        };
+    }): Promise<RerankResponse>;
+}
+/** Response body returned by OpenRouter's rerank router. */
+export interface RerankResponse {
+    readonly results: ReadonlyArray<{
+        readonly index: number;
+        readonly relevanceScore: number;
+    }>;
+    readonly model?: string;
+    readonly provider?: string;
+}
+/** Options for {@link OpenRouterReranker}. */
+export interface OpenRouterRerankerOptions {
+    /** Wire model id. @default DEFAULT_RERANK_MODEL */
+    readonly model?: string;
+    /** API key. When omitted, read from `apiKeyEnv` in the environment. */
+    readonly apiKey?: string;
+    /** Environment variable holding the API key. @default DEFAULT_RERANK_API_KEY_ENV */
+    readonly apiKeyEnv?: string;
+    /** Override the OpenRouter base URL (e.g. a self-hosted or gateway endpoint). */
+    readonly baseUrl?: string;
+    /** Per-request timeout in milliseconds. */
+    readonly timeoutMs?: number;
+    /** Environment to read the API key from. Defaults to `process.env`; a test seam. */
+    readonly env?: NodeJS.ProcessEnv;
+    /** Pre-built client, bypassing SDK construction. A test seam. */
+    readonly client?: RerankClient;
+}
+/** Options for {@link createReranker}: provider plus the OpenRouter reranker options. */
+export interface CreateRerankerOptions extends OpenRouterRerankerOptions {
+    /** Provider id. @default DEFAULT_RERANK_PROVIDER */
+    readonly provider?: string;
+}
+/**
+ * Build the configured reranker. Returns `undefined` when reranking is disabled
+ * or absent, or when the provider is unsupported. Provider-agnostic by design:
+ * a local reranker implements {@link Reranker} and plugs in here without
+ * touching the retrieval pipeline.
+ */
+export declare function createReranker(config: CreateRerankerOptions | false | undefined): Reranker | undefined;
+/**
+ * Reranker backed by OpenRouter's rerank router.
+ *
+ * Documents are sent as their `content`; the original `RetrievalResult` for each
+ * returned index is preserved and re-scored with the provider's relevance score.
+ */
+export declare class OpenRouterReranker implements Reranker {
+    private readonly model;
+    private readonly provider;
+    private readonly client;
+    private readonly unavailableReason;
+    constructor(options?: OpenRouterRerankerOptions);
+    describe(): RerankerDescriptor;
+    rerank(query: string, results: readonly RetrievalResult[], options?: RerankOptions): Promise<RetrievalResult[]>;
+}

package/dist/retrieval/types.d.ts

@@ -26,7 +26,7 @@ export interface RetrievalMethod {
     describe(): RetrievalMethodDescriptor;
     retrieve(query: string, options: RetrievalOptions): Promise<RetrievalResult[]>;
 }
-export type RetrievalDiagnosticCode = "retrieval-method-failed" | "minsync-unavailable";
+export type RetrievalDiagnosticCode = "retrieval-method-failed" | "minsync-unavailable" | "rerank-failed";
 /**
  * A retrieval surface (local MinSync files or one datasource) that was not
  * searched for this query. Reported next to partial results so a caller that

package/licenses/es-MIT.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 voidtools
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/licenses/everything-MIT-and-PCRE-BSD.txt

@@ -0,0 +1,51 @@
+Everything
+
+	Copyright (C) 2018 David Carpenter
+
+	Permission is hereby granted, free of charge, to any person obtaining a
+	copy of this software and associated documentation files (the "Software"),
+	to deal in the Software without restriction, including without limitation
+	the rights to use, copy, modify, merge, publish, distribute, sublicense,
+	and/or sell copies of the Software, and to permit persons to whom the
+	Software is furnished to do so, subject to the following conditions:
+
+	The above copyright notice and this permission notice shall be included in
+	all copies or substantial portions of the Software.
+
+	THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+	IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+	FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+	AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+	LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+	FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+	DEALINGS IN THE SOFTWARE.
+
+Perl-Compatible Regular Expressions
+
+	Copyright (c) 1997-2012 University of Cambridge
+
+	Redistribution and use in source and binary forms, with or without
+	modification, are permitted provided that the following conditions are met:
+
+	* Redistributions of source code must retain the above copyright notice,
+	  this list of conditions and the following disclaimer.
+
+	* Redistributions in binary form must reproduce the above copyright
+	  notice, this list of conditions and the following disclaimer in the
+	  documentation and/or other materials provided with the distribution.
+
+	* Neither the name of the University of Cambridge nor the names of its
+	  contributors may be used to endorse or promote products derived from
+	  this software without specific prior written permission.
+
+	THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+	AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+	IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+	ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+	LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+	CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+	SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+	INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+	CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+	ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+	POSSIBILITY OF SUCH DAMAGE.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@autorag/librarian",
-  "version": "2.6.1",
+  "version": "2.6.3",
   "description": "Self-evolving librarian agent on Pi framework with pluggable retrieval methods",
   "license": "MIT",
   "repository": {
@@ -28,7 +28,8 @@
     "skills",
     "README.md",
     "NOTICE",
-    "licenses"
+    "licenses",
+    "vendor"
   ],
   "publishConfig": {
     "access": "public"
@@ -53,6 +54,7 @@
     "@earendil-works/pi-ai": "^0.86.1",
     "@earendil-works/pi-coding-agent": "0.86.1",
     "@earendil-works/pi-tui": "0.86.1",
+    "@openrouter/sdk": "^1.4.21",
     "chardet": "^2.2.0",
     "fast-xml-parser": "^5.11.1",
     "iconv-lite": "^0.7.3",

package/skills/autorag/SKILL.md

@@ -41,6 +41,14 @@ near/contains families require human review. Exact duplicate exclusion during
 refresh is enabled by default and can be disabled with
 `"excludeExactDuplicates": false` in `config.json`.
 
+### Excluding local files from the index
+
+`"excludePaths"` in `config.json` lists files or folders (absolute, or relative
+to `workspacePath`) that refresh keeps out of the parsed mirror and MinSync.
+A folder entry excludes everything under it. Removing an entry and running
+`autorag refresh --method minsync` indexes the file again. Excluded sources are
+recorded as `user-excluded` skips, so they are not reported as stale.
+
 `status` is model-free and path-opaque. `health` resolves the single model,
 checks credential presence, and normally probes one live completion. If the
 model, authentication, configuration, or indexes are unhealthy, use
@@ -116,9 +124,10 @@ autorag serve --force
 autorag p2p policy list --json
 ```
 
-Prefer a full refresh so parsed mirrors, MinSync, Jikji, and configured
-datasources stay aligned. `--method` accepts
-`parsed,minsync,datasources,jikji,all`. BM25 is a MinSync retrieval mode, not
+Prefer a full refresh so parsed mirrors, MinSync, Jikji, configured
+datasources, and (on Windows) the bundled Everything file-name index stay
+aligned. `--method` accepts
+`parsed,minsync,datasources,jikji,everything,all`. BM25 is a MinSync retrieval mode, not
 a `--method` name. Use `--method` only for deliberate narrowing. Scheduled
 maintenance should use non-daemon `autorag watch --once`, typically every
 1 hour, with the same config used by search and no overlapping runs.

package/skills/autorag-doctor/SKILL.md

@@ -58,6 +58,7 @@ Each CLI owns its archive, so ask the CLI, not AutoRAG. A datasource is only
 |---|---|---|
 | MinSync (local docs) | `minsync status`, `minsync check`, `minsync verify` | `autorag refresh --method minsync` |
 | Jikji (discovery) | `jikji doctor` | `autorag refresh --method jikji` |
+| Everything (Windows file names) | `autorag status --json` → `components.everything` | `autorag refresh --method everything --json` |
 | KakaoTalk | `lazykatok doctor` | `lazykatok sync && lazykatok index` |
 | Discord | `discrawl --json metadata` | `discrawl sync` |
 | Slack | `slacrawl --json doctor` | `slacrawl sync` |
@@ -207,6 +208,7 @@ timer, or Task Scheduler).
 | `index-not-ready` | Index missing or never built | `autorag refresh --json` |
 | `minsync-unavailable` | MinSync binary missing or install failed | Check cargo, retry refresh |
 | `jikji-unavailable` | Jikji binary missing or install failed | Check cargo, retry refresh |
+| `everything-index-failed` | Windows Everything instance could not start or index; message carries ES exit code and stderr | Fix the reported cause, `autorag refresh --method everything --json` |
 | `embedding-identity-mismatch` | Indexed vectors use a different embedder | `autorag index rebuild --method minsync` |
 | `embedder-unavailable` | Embedding gateway or runtime down | `autorag gateway status --format json` |
 | `lock-conflict` | Another runtime holds the lock | `autorag gateway stop` |

package/skills/autorag-lite-setup/SKILL.md

@@ -110,8 +110,9 @@ autorag lite refresh --force --json
   incremental refresh is not enough, such as after config changes to roots or
   index settings.
 - `--method <csv>` deliberately narrows the refresh. Valid values are
-  `parsed`, `minsync`, `datasources`, `jikji`, and `all`. Omit the flag to run
-  all methods. Unknown values are rejected.
+  `parsed`, `minsync`, `datasources`, `jikji`, `everything` (Windows only;
+  a no-op elsewhere), `fsearch` (macOS/Linux only; a no-op elsewhere), and
+  `all`. Omit the flag to run all methods. Unknown values are rejected.
 - MinSync and Jikji auto-install on first use by default. If they are missing
   or broken, run a full refresh or return to setup rather than silently
   degrading to lexical-only search.

package/skills/autorag-setup/SKILL.md

@@ -40,6 +40,7 @@ Check `--config`, `AUTORAG_CONFIG`, `$AUTORAG_HOME/config.json`, or
 - `searchPaths`, `workspacePath`, and `memoryPath`
 - `model.provider`, `model.id`, `model.api`, `model.baseUrl`, `model.apiKeyEnv`
 - `bm25`, `minSync`, and `jikji`
+- `limits` (retrieval, baseline-prefetch, and model-facing candidate caps)
 - `datasources`, `datasourceAccess`, and `ui`
 
 Preserve explicit user choices and a working config unless the user asks to
@@ -122,6 +123,13 @@ flags may be omitted. For a custom endpoint, add `api`, `baseUrl`, and
 }
 ```
 
+When `provider`/`id` names a pi-ai catalog model, the catalog entry stays the
+base: `baseUrl`, `api`, and any declared `reasoning`, `input`, `contextWindow`,
+or `maxTokens` override only those fields, and the catalog's reasoning,
+thinking, and compat settings are kept. Only an id outside the catalog (private
+proxy, Ollama, LiteLLM) gets a generic text model with a 128k context window
+unless those fields are declared.
+
 Use `--force` only when intentionally replacing an existing config. Legacy cwd
 `autorag.config.json` is a migration source only and is never deleted by init.
 
@@ -178,6 +186,48 @@ Only store the environment-variable name, never its value. Dimension and batch
 size must be positive integers, and the dimension must match the embedder
 (default Qwen3 is 1024; legacy EmbeddingGemma is 768; text-embedding-3-small is 1536).
 
+### Retrieval and ingest caps
+
+`limits` bounds retrieval, baseline prefetch, and the candidate lists handed to
+the model. Every field is optional — an omitted field keeps the shipped default,
+so add the section only to tighten or widen a specific cap. Values must be
+positive integers; unknown keys (and unknown `prefetch` keys) fail config
+resolution.
+
+| Field | Default | Controls |
+|---|---|---|
+| `mergedEvidenceCeiling` | 500 | `search_all_documents` / model-free merge ceiling when the model omits `topK` |
+| `singleDatasourceTopK` | 50 | `search_datasource_*` merge default when the model omits `topK` |
+| `minSyncTopK` | 50 | MinSync semantic default `topK` |
+| `minSyncScopedQueryTopK` | 100 | MinSync fetch cap applied when a scope narrows the query |
+| `toolDescriptionInstanceScopes` | 8 | Instance scopes listed in one datasource tool description |
+| `prefetch.jikjiTopK` | 30 | Jikji find candidate count |
+| `prefetch.minSyncTopK` | 100 | MinSync retrieve candidate count |
+| `prefetch.jikjiPathLimit` | 100 | Max Jikji answer paths rendered into the baseline |
+| `prefetch.sectionLimit` | 100 | Max results rendered per baseline section |
+
+```json
+{
+  "limits": {
+    "mergedEvidenceCeiling": 1000,
+    "prefetch": { "jikjiTopK": 12, "sectionLimit": 30 }
+  }
+}
+```
+
+The baseline renders each prefetched result's full chunk content (not a
+per-result excerpt) so the model can see where the hit came from. Bound baseline
+size with `prefetch.sectionLimit` / `prefetch.minSyncTopK` /
+`prefetch.jikjiPathLimit`, not with a per-result content truncation.
+
+Ingest caps are trusted connector options under `datasources.<name>.connector`
+and are never settable from model/tool arguments: `maxDocuments`,
+`maxItemsPerFeed`, and `maxContentChars` (RSS); `maxDocuments`,
+`maxContentChars`, `maxResultsPerQuery`, and `maxBytesPerFile` (Spotlight);
+`maxDocuments`, `maxContentChars`, `maxBytesPerFile`, `concurrency`,
+`bandwidthLimit`, and `dryRun` (cloud-drive/rclone). `maxContentChars` defaults
+to 20000 for RSS and 100000 for Spotlight and cloud-drive.
+
 ## Probe and configure datasource skills (setup wizard)
 
 Use `autorag setup` to probe the local runtime, model profile, and known
@@ -269,9 +319,9 @@ autorag search "summarize the collection" --top-k 3 --json --debug
   performs one live completion probe.
 - `health --skip-probes` is only for intentionally offline validation and does
   not prove live provider access.
-- `refresh` syncs parsed mirrors, MinSync, Jikji, and authorized
-  datasources. `--method <csv>` may deliberately narrow it
-  (`parsed,minsync,datasources,jikji,all`).
+- `refresh` syncs parsed mirrors, MinSync, Jikji, authorized datasources, and
+  on Windows the bundled Everything file-name index. `--method <csv>` may
+  deliberately narrow it (`parsed,minsync,datasources,jikji,everything,all`).
 - Use `refresh --force` for a full resync only when incremental refresh is not
   enough. Keep destructive reset/rebuild operations scoped to workspace
   `.autorag` indexes, never source documents.

package/vendor/everything/manifest.json

@@ -0,0 +1,30 @@
+{
+  "everythingVersion": "1.4.1.1032",
+  "esVersion": "1.1.0.38",
+  "licenseFiles": [
+    "licenses/everything-MIT-and-PCRE-BSD.txt",
+    "licenses/es-MIT.txt"
+  ],
+  "architectures": {
+    "x64": {
+      "everythingArchive": "Everything-1.4.1.1032.x64.en-US.zip",
+      "everythingArchiveSha256": "705e0b46c6cbf74bd6082a6ebf77afac526868db9d7578f5dc1447f459484611",
+      "everythingMember": "everything.exe",
+      "everythingSha256": "f191f756996a14a11e5445fa7103d302efd510cf2fbf920e6c0c8ed51d512e36",
+      "esArchive": "ES-1.1.0.38.x64.zip",
+      "esArchiveSha256": "5e0c70cbf4f694080c34aa7c6c745e606c16fe76a4b5423b93ebf9dc34274c99",
+      "esMember": "es.exe",
+      "esSha256": "f7378761cf6e01f51c4123a485e628d70e3fea147d341f473ce2820844e5cee5"
+    },
+    "arm64": {
+      "everythingArchive": "Everything-1.4.1.1032.ARM64.en-US.zip",
+      "everythingArchiveSha256": "5a7b0cd0372f5e076738e88db20da219fb784c832e61256b3d63f95cd9caeba1",
+      "everythingMember": "EverythingARM64.exe",
+      "everythingSha256": "462caf17906e56be11ce3d701a6ea51bb1733afae8a4610880f3a6985ee9ca5e",
+      "esArchive": "ES-1.1.0.38.ARM64.zip",
+      "esArchiveSha256": "86eddd4ae8925833e33be1d09e7faa2442110d00daf0e48a8e963cdeb7a141fe",
+      "esMember": "es.exe",
+      "esSha256": "7fc406610559f686a95a9bfc4472acffdebcc143b485ae9371b5ef4590e7ec31"
+    }
+  }
+}