@autorag/librarian 2.13.3 → 2.13.5

package/dist/retrieval/skip.d.ts

@@ -11,7 +11,7 @@
  * stderr, paths and all.
  */
 import type { RetrievalMethodDescriptor, RetrievalUnsearchedSurface } from "./types.js";
-/** Surface label shared by every MinSync-backed local method (vector, BM25, hybrid). */
+/** Surface label shared by every MinSync-backed local method (vector, sparse, hybrid). */
 export declare const MINSYNC_SURFACE = "minsync";
 /** One method that did not run, with the error that stopped it. */
 export interface RetrievalSkip {

package/dist/retrieval/types.d.ts

@@ -7,7 +7,7 @@ export interface RetrievalResult {
 }
 export interface RetrievalMethodDescriptor {
     name: string;
-    type: "posix" | "vector" | "bm25" | "hybrid" | "visual" | "remote";
+    type: "posix" | "vector" | "bm25" | "sparse" | "hybrid" | "visual" | "remote";
     description: string;
     status: "active" | "stub";
     capabilities: string[];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@autorag/librarian",
-  "version": "2.13.3",
+  "version": "2.13.5",
   "description": "Personal librarian agent for office work on the Pi framework, with pluggable retrieval methods",
   "license": "Apache-2.0",
   "repository": {
@@ -65,6 +65,7 @@
     "chardet": "^2.2.0",
     "discord.js": "14.27.0",
     "fast-xml-parser": "^5.11.2",
+    "gpt-tokenizer": "^4.1.0",
     "iconv-lite": "^0.7.3",
     "jev-use": "^0.8.0",
     "jszip": "^3.10.2",
@@ -77,7 +78,8 @@
     "slides-grab": "^1.5.2",
     "smol-toml": "1.9.1",
     "tesseract.js": "^7.0.0",
-    "typebox": "^1.3.37"
+    "typebox": "^1.3.37",
+    "xlsx": "https://cdn.sheetjs.com/xlsx-0.20.3/xlsx-0.20.3.tgz"
   },
   "devDependencies": {
     "@biomejs/biome": "2.6.4",
@@ -86,7 +88,6 @@
     "@types/ws": "^8.18.2",
     "typescript": "^7.0.2",
     "vitest": "^5.0.5",
-    "xlsx": "https://cdn.sheetjs.com/xlsx-0.20.3/xlsx-0.20.3.tgz",
     "yaml": "^2.9.1"
   }
 }

package/skills/autorag/SKILL.md

@@ -81,8 +81,9 @@ 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. Jikji
 indexes the source folders directly and is not refreshed here, so AutoRAG also
-drops excluded paths from the `jikji_find` answer pack and the baseline prefetch
-at retrieval time; the on-disk `.jikji_agent_map.md` stays complete, and direct
+drops excluded paths from the `jikji_find` answer pack and from the files Jikji
+hands to the verification step as starting points, at retrieval time; the
+on-disk `.jikji_agent_map.md` stays complete, and direct
 file reads remain available.
 
 `status` is model-free and path-opaque. `health` resolves the single model,
@@ -223,8 +224,10 @@ grant an agent more than its profile has. `retrieval` optionally adds
 Retrieval memory is never connected to an automation agent.
 
 Every agent run gets a loop guard appended to its system prompt, naming the
-rule's folders, file types and events, and forbidding any change that would
-fire the rule again, editing the rules file, and running `autorag automate`. At
+rule's folders, file types and events, and how the rule decides what fires it.
+The agent must make no change that fires the rule again (a rename or write whose
+events the rule does not pick up is fine), designs a non-firing change before it
+declines, and never edits the rules file or runs `autorag automate`. At
 most one agent run is in flight across all rules, and an agent rule is
 disabled after 6 runs within 60 seconds (command rules: 20). The run's final
 reply is the `stdout` of its log entry.

package/skills/autorag-doctor/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: autorag-doctor
-description: Diagnose and repair a broken or half-working AutoRAG install so every configured source is both indexed and searchable. Checks AutoRAG, MinSync, Jikji, the embedding backend, and every CLI-backed datasource (lazykatok, discrawl, slacrawl, wacrawl, telecrawl, notcrawl, mailcrawl, rclone, Spotlight), then fixes orphan locks, orphan processes, embedding dimension or identity mismatches, stale indexes, and missing setup. Use when search returns nothing or too little, refresh hangs or fails, a datasource disappeared from results, indexes look stale, the embedder is unavailable, or the user asks to check, diagnose, verify, or repair AutoRAG.
+description: Diagnose and repair a broken or half-working AutoRAG install so every configured source is both indexed and searchable. Checks AutoRAG, MinSync, Jikji, the embedding backend, and every CLI-backed datasource (lazykatok, discrawl, slacrawl, wacrawl, telecrawl, notcrawl, mailcrawl), then fixes orphan locks, orphan processes, embedding dimension or identity mismatches, stale indexes, and missing setup. Use when search returns nothing or too little, refresh hangs or fails, a datasource disappeared from results, indexes look stale, the embedder is unavailable, or the user asks to check, diagnose, verify, or repair AutoRAG.
 license: Apache-2.0
 ---
 
@@ -72,8 +72,7 @@ Each CLI owns its archive, so ask the CLI, not AutoRAG. A datasource is only
 | Notion | `notcrawl doctor`, `notcrawl status` | `notcrawl sync --source desktop` |
 | Obsidian / notes | `autorag status` (AutoRAG-owned index) | `autorag refresh --method datasources`; if corrupt, delete `.autorag/datasources/obsidian/<instance>` first |
 | Mail | `mailcrawl doctor`, `mailcrawl status` | `mailcrawl sync && mailcrawl index` |
-| Cloud drive | `rclone listremotes` | `autorag refresh --method datasources` |
-| Spotlight (macOS) | `mdutil -s /` | indexed by the OS; no AutoRAG sync |
+| Coding-agent sessions | `wombat doctor --json` | `wombat sync --index` |
 | Lark / Feishu | `lark-cli auth status --format json` | no local sync; search is remote |
 
 Rules:
@@ -230,6 +229,10 @@ to retry those files once without re-indexing healthy ones.
   (`cargo install minsync`) or leave `minSync.autoInstall` on, then retry.
 - `jikji-unavailable`: the binary is missing and auto-install failed. It
   installs through cargo; verify the Rust toolchain, then `autorag refresh --method jikji` to retry.
+- `jikji-unsupported-version`: the `jikji` AutoRAG found is older than 0.2.0, so it was never run.
+  The message names the detected version and the binary path. `autorag refresh --method jikji`
+  installs the pinned build; with `jikji.autoInstall: false` or `jikji.binaryPath` upgrade it yourself
+  (`cargo install jikji-cli --version 0.2.1 --locked --force`).
 - `datasource-auth-error` / `datasource-rate-limited`: datasource credentials or quota. Report the
   missing environment-variable **name** and let the user supply it.
 - A configured datasource that returns nothing is a native store, connector, or
@@ -245,8 +248,8 @@ to retry those files once without re-indexing healthy ones.
 | `minsync-sync-failed` | MinSync indexing failed; the message carries MinSync's own reason | Fix the reported cause, then `autorag refresh --method minsync` |
 | `minsync-file-skipped` | MinSync skipped one file (unreadable, unparsable, or rejected by the embedder) and kept its older chunks; retried on the next sync | Fix that file's permissions/content, then `autorag refresh --method minsync` |
 | `minsync-lock-held` | Another `minsync` run holds `.minsync/lock`, often an orphan left by a killed host; retryable | Wait for it to finish or stop that `minsync` process, then `autorag refresh --method minsync` |
-| `lexical-tokenizer-unavailable` | Warning: the BM25 tokenizer needs the Kiwi library and model, which were not found, so MinSync tokenizes on whitespace and `인증서를` does not match `인증서` | Install Kiwi (set `KIWI_LIBRARY_PATH` and `KIWI_MODEL_PATH`, or install under `~/.local/kiwi`), then `autorag refresh --method minsync` |
 | `jikji-unavailable` | Jikji binary missing or install failed | Check cargo, retry refresh |
+| `jikji-unsupported-version` | Jikji older than 0.2.0 was found and not run; message names the version and path | `autorag refresh --method jikji` installs the pinned build, or upgrade it yourself |
 | `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` | No remote endpoint answered, no daemon could start, and the in-process addon is missing or failed | Run `minsync serve` by hand to read its error, install MinSync (`cargo install minsync`) or `@nomadamas/minsync-native`, or configure a remote `minSync.embedder` |
@@ -254,6 +257,7 @@ to retry those files once without re-indexing healthy ones.
 | `datasource-empty` | Store has no matching content | Re-sync with the CLI, or accept as empty |
 | `unknown-datasource-skill` | Config name is not a known template | Fix the name or add `"type"` |
 | `retrieval-method-failed` | One method errored during the query | Read `--debug` diagnostics |
+| `rerank-disabled` | Warning: the `rerank` block names no model, so reranking is off and each source contributes its own top 10 to the first answer | Set `rerank.model` to turn it on, or remove the block; nothing else is wrong |
 | `datasource-auth-error` | Datasource credentials missing or rejected | Report the env var name |
 | `query-route-fallback` | Jev routing unavailable (often `OPENROUTER_API_KEY` unset); searched local with the original question | `test -n "$OPENROUTER_API_KEY"`; report the env var name, never its value |
 | `query-decomposition-failed` | Decomposition model call failed; searched the original question | Check `queryDecomposition.model` resolves (`autorag models list --provider openrouter`) |

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

@@ -56,7 +56,7 @@ explicit path (`--config` or `AUTORAG_CONFIG`); `--force` refuses to replace the
 implicit `~/.autorag/config.json`. Explicit user paths win;
 otherwise propose one to three document-dense roots and get approval before
 indexing. Supported parsed formats are `md`, `markdown`, `txt`, `text`, `pdf`,
-`docx`, `pptx`, `xlsx`, `xls`, `hwp`, `hwpx`, and `eml`. Legacy `.doc` is not
+`docx`, `pptx`, `xlsx`, `xlsm`, `xlsb`, `xls`, `ods`, `csv`, `tsv`, `hwp`, `hwpx`, and `eml`. Legacy `.doc` is not
 a supported parsed format.
 
 Config resolution follows the usual order: `--config`, `AUTORAG_CONFIG`,
@@ -113,7 +113,7 @@ setup. Configure datasources directly in trusted config, wizard-style:
 
 1. Probe every datasource for setup feasibility before asking the user
    anything: the backing CLI exists (`lazykatok`, `discrawl`, `slacrawl`,
-   `wacrawl`, `telecrawl`, `notcrawl`, `mailcrawl`, `rclone`) and its
+   `wacrawl`, `telecrawl`, `notcrawl`, `mailcrawl`, `wombat`) and its
    local store or archive is present. CLI-backed datasources own their own
    archive, index, and authentication, so environment credentials (such as bot
    tokens) are never required or checked for them. Non-CLI connectors
@@ -136,8 +136,8 @@ setup. Configure datasources directly in trusted config, wizard-style:
 Datasource skills belong in trusted config.
 
 Config keys may be builtin template names (`kakao`, `whatsapp`, `telegram`,
-`slack`, `discord`, `clawgallery`, `notion`, `github`, `cloud-drive`,
-`mailcrawl`, `obsidian`, `rss`, `spotlight`, `lark`, `hister`) or connection
+`slack`, `discord`, `clawgallery`, `notion`, `github`,
+`mailcrawl`, `obsidian`, `rss`, `lark`, `hister`, `wombat`) or connection
 aliases with `"type": "<template>"`. Unknown names are skipped with an
 `unknown-datasource-skill` warning; they do not fail config resolution.
 `scope` narrows a query to a sub-path as ordinary filtering. Tags are

package/skills/autorag-setup/SKILL.md

@@ -94,7 +94,7 @@ approval before indexing:
 | Windows | Documents, Downloads, Desktop shell folders | OneDrive document roots, user-named project docs |
 
 Supported parsed formats are `md`, `markdown`, `txt`, `text`, `pdf`, `docx`,
-`pptx`, `xlsx`, `xls`, `hwp`, `hwpx`, and `eml`. OCR for `jpg`, `jpeg`, `png`, `bmp`,
+`pptx`, `xlsx`, `xlsm`, `xlsb`, `xls`, `ods`, `csv`, `tsv`, `hwp`, `hwpx`, and `eml`. OCR for `jpg`, `jpeg`, `png`, `bmp`,
 and `tiff` is optional (`parserOptions.ocr.enabled`). Do not present legacy
 `.doc` as a supported parsed format.
 
@@ -162,8 +162,9 @@ workspace still answers, but without MinSync/Jikji evidence.
 
 Refresh (never a query) auto-installs the binaries: MinSync installs a verified
 GitHub release into `<workspace>/.autorag/bin` (`minSync.autoInstall` defaults
-to true), and Jikji installs `jikji-cli` through cargo (`jikji.autoInstall`
-defaults to true; requires the Rust toolchain). Set `"autoInstall": false` only
+to true), and Jikji installs the pinned `jikji-cli` 0.2.1 through cargo (`jikji.autoInstall`
+defaults to true; requires the Rust toolchain). A `jikji` older than 0.2.0 on PATH
+or in the cache is skipped and replaced by that build on refresh. Set `"autoInstall": false` only
 when managing the binary yourself. Refresh is incremental: MinSync syncs only
 changed parsed mirrors, and `jikji prepare` reuses unchanged documents. Jikji
 roots prepare one at a time (they share one central SQLite file that rejects
@@ -178,8 +179,11 @@ with the `JIKJI_DATA_DIR` environment variable). The only file `refresh` adds
 inside an approved document root is the `.jikji_agent_map.md` guide. Tell the
 user that before the first `refresh`, and never delete or edit Jikji's central
 store by hand; removing it only forces a full Jikji re-prepare on the next
-refresh. Older Jikji versions wrote a `<root>/.jikji` directory instead; such a
-leftover is now unused and safe for the user to delete.
+refresh. Jikji 0.1.x wrote a `<root>/.jikji` directory instead; such a
+leftover is now unused and safe for the user to delete. Jikji 0.2.0 and newer
+refuse to index a root nested inside, or containing, an already indexed root
+(`prepare root overlaps existing indexed root`), which is why overlapping search
+roots are collapsed to the outermost one before Jikji runs.
 
 Exact duplicate exclusion is enabled by default. AutoRAG invokes the external
 `dupey` CLI before parsed-mirror indexing, keeps the newest filesystem copy for
@@ -231,10 +235,11 @@ endpoint POSTs `{baseUrl}/embed`; `openai:` with no `baseUrl` means
 value. `dimension` and `batchSize` must be positive integers, and `dimension`
 must match the endpoint (the local default is 1024).
 
-### Jev routing and question decomposition (on by default)
+### Jev decision model (required) and question decomposition (on by default)
 
-Leave both enabled. They are the recommended setup: they make simple questions
-fast and multi-part questions thorough.
+Jev is required: AutoRAG cannot answer without it. Decomposition is on by
+default; leave it enabled. Together they make simple questions fast and
+multi-part questions thorough.
 
 - **Jev** (`jev`, default `{ "backend": "dazzi" }` when signed in to Dazzi
   Models, else `{ "backend": "openrouter" }`; model `jev-1.13` on Dazzi,
@@ -263,10 +268,13 @@ fast and multi-part questions thorough.
 
 Both need a credential for the chosen backend: a Dazzi Models `/login` (check with
 `autorag models list --provider dazzi --available`), or the user's
-`OPENROUTER_API_KEY` (`test -n "$OPENROUTER_API_KEY"`, never print it). Tell the
-user Jev routing is on. Without a credential, routing falls back to a single
-local search and the run always verifies, so searches still work. A
-`query-route-fallback` diagnostic (`autorag search --debug`) shows that state.
+`OPENROUTER_API_KEY` (`test -n "$OPENROUTER_API_KEY"`, never print it). The first
+question of every session checks that Jev answers; without a working credential
+the search stops with `[decision-model-unavailable]` (`autorag search` exits 3,
+and `--json` carries `code`, `reason`, `backend`, `model`). Read `reason`:
+`missing-credential` means set the key or sign in, `unauthorized` means replace
+the key, `model-not-found` means fix `jev.model`, `payment-required` means the
+backend account is out of credit, `unreachable` and `rate-limited` mean retry.
 
 ```json
 {
@@ -277,14 +285,24 @@ local search and the run always verifies, so searches still work. A
 
 `autorag init` writes these defaults into new configs. To change them:
 
-- Jev backend: `"backend": "typesafe"` (`TYPESAFE_API_KEY`) or `"vercel"`
-  (`AI_GATEWAY_API_KEY`).
+- Jev backend and model: any provider with decision models, e.g.
+  `{ "backend": "openai", "model": "gpt-6-luna" }` (`OPENAI_API_KEY`),
+  `{ "backend": "openrouter", "model": "openai/gpt-6-luna-decisions" }`,
+  `"typesafe"` (`TYPESAFE_API_KEY`), `"vercel-ai-gateway"` (`AI_GATEWAY_API_KEY`),
+  `"cloudflare-workers-ai"`, or `"opencode"`. List the choices with
+  `autorag models list --decision --available`; never invent a model id.
+- sLLM fallback: when Jev is unavailable, the session runs on a small chat model
+  (default: the decomposition model) and warns `decision-model-fallback`. Set the
+  model with `"jev": { "sllm": { "model": { "provider": …, "id": … } } }`, or
+  `"fallback": false` to fail instead.
+- sLLM as the decision model: `"jev": { "backend": "sllm" }` decides with the
+  small chat model alone and never calls Jev (no Jev credential needed).
 - Decomposition model: any catalog `provider`/`id`, with the same fields as
   the top-level `model`.
 - `"queryDecomposition": false` decomposes with the search model itself.
-- `"jev": false` turns routing off entirely. Do this only when the user
-  explicitly opts out, for example because questions must never leave the
-  machine (Jev and decomposition send the question text to OpenRouter).
+- Jev cannot be turned off: `"jev": false` and `jev.enabled` are config errors.
+  When questions must never leave the machine, AutoRAG is not usable; say so
+  instead of editing the config (Jev sends the question text to its backend).
 
 When a user asks the running agent itself to change its settings (switch the
 default model, add a provider, check that a provider works), Jev's `config`
@@ -305,12 +323,13 @@ 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` |
+| `singleDatasourceTopK` | 20 | `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.jikjiTopK` | 10 (30 with a rerank model) | Jikji find candidate count |
+| `prefetch.minSyncTopK` | 10 (100 with a rerank model) | MinSync retrieve candidate count |
+| `prefetch.datasourceTopK` | 10 (20 with a rerank model) | Hits requested from each selected datasource |
 | `prefetch.jikjiPathLimit` | 100 | Max Jikji answer paths rendered into the baseline |
 | `prefetch.sectionLimit` | 100 | Max results rendered per baseline section |
 
@@ -326,15 +345,19 @@ resolution.
 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.
+`prefetch.datasourceTopK` / `prefetch.jikjiPathLimit`, not with a per-result
+content truncation.
+
+The `prefetch.*TopK` defaults depend on reranking. Reranking is opt-in and runs
+only when `rerank.model` names a model; then the pool is wide (Jikji 30, MinSync
+100, each datasource 20) and the reranker picks the best. Without it (the
+default) each source contributes its own top 10 to the first answer. An explicit
+value always wins.
 
 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.
+`maxItemsPerFeed`, and `maxContentChars` (RSS). `maxContentChars` defaults to
+20000 for RSS.
 
 ### Refresh safety limits
 
@@ -366,7 +389,7 @@ setup. Configure datasources directly in trusted config, wizard-style:
 
 1. Probe every datasource for setup feasibility before asking the user
    anything: the backing CLI exists (`lazykatok`, `discrawl`, `slacrawl`,
-   `wacrawl`, `telecrawl`, `notcrawl`, `mailcrawl`, `rclone`, `lark-cli`) and its
+   `wacrawl`, `telecrawl`, `notcrawl`, `mailcrawl`, `wombat`, `lark-cli`) and its
    local store or archive is present. CLI-backed datasources own their own
    archive, index, and authentication, so environment credentials (such as bot
    tokens) are never required or checked for them. Non-CLI connectors
@@ -388,8 +411,8 @@ setup. Configure datasources directly in trusted config, wizard-style:
 
 Datasource skills belong in trusted config. Builtin
 template names are `kakao`, `whatsapp`, `telegram`, `slack`, `discord`,
-`clawgallery`, `notion`, `github`, `github-gist`, `cloud-drive`,
-`mailcrawl`, `obsidian`, `rss`, `spotlight`, `lark`, and `hister`. Config keys may be connection
+`clawgallery`, `notion`, `github`, `github-gist`,
+`mailcrawl`, `obsidian`, `rss`, `lark`, `hister`, and `wombat`. Config keys may be connection
 aliases with `"type": "<template>"`. Unknown names are skipped with an
 `unknown-datasource-skill` warning; they do not fail config resolution.
 `scope` narrows a query to a sub-path as ordinary filtering. Tags are descriptive metadata only, not search filters. MCP `datasourceIds` selects configured connections before retrieval; discover their IDs with `autorag.datasources.list`.
@@ -399,9 +422,8 @@ aliases with `"type": "<template>"`. Unknown names are skipped with an
   "datasources": {
     "github": { "connector": { "repos": ["owner/repo"], "tokenEnv": "GITHUB_TOKEN" } },
     "github-gist": { "connector": { "tokenEnv": "GITHUB_TOKEN" } },
-    "google-drive": { "type": "cloud-drive", "connector": { "provider": "google-drive", "remote": "gdrive:" } },
-    "archive-drive": { "type": "cloud-drive", "connector": { "remote": "archive:" } },
     "mailcrawl": { "instanceId": "personal", "connector": { "account": "personal", "mailbox": "INBOX", "binaryPath": "mailcrawl" } },
+    "wombat": { "connector": { "binaryPath": "wombat", "agents": ["codex", "claude-code"] } },
     "obsidian": { "connector": { "vaultPath": "/path/to/vault" } },
     "rss": { "connector": { "feeds": [{ "url": "https://example.com/feed.xml" }] } }
   }

package/dist/datasource/skills/cloud-drive/index.d.ts

@@ -1,2 +0,0 @@
-export { RcloneConnector, type RcloneConnectorOptions, type RcloneRunner, type RcloneRunResult, } from "./rclone-connector.js";
-export { CLOUD_DRIVE_SKILL_DEFINITION, CloudDriveSkill, type CloudDriveSkillOptions, } from "./skill.js";

package/dist/datasource/skills/cloud-drive/rclone-connector.d.ts

@@ -1,49 +0,0 @@
-/**
- * Provider-neutral incremental rclone mirror.
- *
- * `lsjson` inventories a remote, a workspace manifest computes its diff, and
- * only added/changed indexable files are copied into a managed mirror. The
- * manifest is published only after every changed file is available, so search
- * can continue using the previous completed snapshot after an interrupted run.
- */
-import type { DefaultParserRegistryOptions } from "../../../parser/defaults.js";
-import type { ConnectorFetchResult, DatasourceConnector } from "../../connector.js";
-export interface RcloneRunResult {
-    readonly ok: boolean;
-    readonly stdout: string;
-    readonly stderr: string;
-    readonly code: number | null;
-}
-export type RcloneRunner = (args: readonly string[], timeoutMs: number) => Promise<RcloneRunResult>;
-export interface RcloneConnectorOptions {
-    readonly binaryPath?: string;
-    readonly remote?: string;
-    readonly provider?: string;
-    readonly workspaceRoot?: string;
-    readonly instanceId?: string;
-    readonly skillName?: string;
-    readonly extensions?: readonly string[];
-    readonly include?: readonly string[];
-    readonly exclude?: readonly string[];
-    readonly exportFormats?: string;
-    readonly maxDocuments?: number;
-    readonly maxBytesPerFile?: number;
-    /** Characters kept per document body. Default 100000. */
-    readonly maxContentChars?: number;
-    readonly concurrency?: number;
-    readonly bandwidthLimit?: string;
-    readonly dryRun?: boolean;
-    readonly timeoutMs?: number;
-    readonly runner?: RcloneRunner;
-    /** Explicit operator-owned rclone.conf path, passed read-only. */
-    readonly configPath?: string;
-    readonly parserOptions?: DefaultParserRegistryOptions;
-}
-export declare class RcloneConnector implements DatasourceConnector {
-    private readonly options;
-    private readonly runner;
-    constructor(options?: RcloneConnectorOptions);
-    fetch(): Promise<ConnectorFetchResult>;
-    private fetchIncremental;
-    private fetchWithoutMirror;
-}

package/dist/datasource/skills/cloud-drive/skill.d.ts

@@ -1,14 +0,0 @@
-import type { DatasourceConnector } from "../../connector.js";
-import { ConnectorDatasourceSkill, type ConnectorSkillDefinition, type ConnectorSkillOptions } from "../../connector-skill.js";
-import { type RcloneConnectorOptions } from "./rclone-connector.js";
-export declare const CLOUD_DRIVE_SKILL_DEFINITION: ConnectorSkillDefinition;
-export interface CloudDriveSkillOptions extends Omit<ConnectorSkillOptions, "connector"> {
-    /** Unique model-visible datasource name for this configured connection. */
-    readonly skillName?: string;
-    readonly provider?: "google-drive" | "onedrive" | "icloud" | string;
-    readonly connector?: DatasourceConnector;
-    readonly connectorOptions?: RcloneConnectorOptions;
-}
-export declare class CloudDriveSkill extends ConnectorDatasourceSkill {
-    constructor(options?: CloudDriveSkillOptions);
-}

package/dist/datasource/skills/spotlight/connector.d.ts

@@ -1,53 +0,0 @@
-import type { ConnectorFetchFail, ConnectorFetchOk } from "../../connector.js";
-export interface SpotlightConnectorFailure extends ConnectorFetchFail {
-    readonly stderr?: string;
-    readonly stdout?: string;
-    readonly code?: number | null;
-}
-export type SpotlightConnectorResult = ConnectorFetchOk | SpotlightConnectorFailure;
-export interface ProcessRunResult {
-    readonly ok: boolean;
-    readonly stdout: string;
-    readonly stderr: string;
-    readonly code: number | null;
-    /** Spawn error code, e.g. `"ENOENT"` when the binary is missing. */
-    readonly error?: string;
-}
-/** Process execution seam; injectable for tests. Never throws. */
-export type ProcessRunner = (command: string, args: readonly string[], timeoutMs: number) => Promise<ProcessRunResult>;
-export type FileReader = (path: string) => Promise<string>;
-export interface SpotlightConnectorOptions {
-    /** Trusted Spotlight filters AND-ed with each user query. Required. */
-    readonly queries?: readonly string[];
-    /** Trusted directory restricting every query (`mdfind -onlyin`). */
-    readonly onlyIn?: string;
-    /** Hits kept per query after dedupe. Default 100. */
-    readonly maxResultsPerQuery?: number;
-    /** Maximum documents hydrated by one live search. Default 300. */
-    readonly maxDocuments?: number;
-    /** Characters kept per document body. Default 100000. */
-    readonly maxContentChars?: number;
-    /** Files larger than this are skipped. Default 256 KiB. */
-    readonly maxBytesPerFile?: number;
-    /** Per-process timeout in ms. Default 15s. */
-    readonly timeoutMs?: number;
-    /** Injectable platform; defaults to `process.platform`. */
-    readonly platform?: NodeJS.Platform;
-    /** Injectable process runner; defaults to a bounded spawn wrapper. */
-    readonly run?: ProcessRunner;
-    /** Injectable file reader; defaults to utf8 `fs.readFile`. */
-    readonly readFile?: FileReader;
-}
-/** Executes live Spotlight queries. macOS owns indexing; no local snapshot is collected. */
-export declare class SpotlightConnector {
-    private readonly options;
-    private readonly run;
-    private readonly readFile;
-    private readonly platform;
-    constructor(options?: SpotlightConnectorOptions);
-    /** Refresh-time health check only: never runs `mdfind` or hydrates files. */
-    probe(): Promise<SpotlightConnectorResult>;
-    /** Search the current macOS Spotlight index and hydrate only the requested hits. */
-    search(query: string, limit?: number, acceptDocument?: (docId: string) => boolean): Promise<SpotlightConnectorResult>;
-    private readinessFailure;
-}

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

@@ -1,2 +0,0 @@
-export { type FileReader, type ProcessRunner, type ProcessRunResult, SpotlightConnector, type SpotlightConnectorFailure, type SpotlightConnectorOptions, type SpotlightConnectorResult, } from "./connector.js";
-export { SpotlightSkill, type SpotlightSkillOptions } from "./skill.js";

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

@@ -1,33 +0,0 @@
-import type { RetrievalMethod, RetrievalOptions, RetrievalResult } from "../../../retrieval/types.js";
-import type { DatasourceIndexResult, DatasourceSkill, DatasourceSkillDescriptor, DatasourceSkillManifest, PollingMetadata, SourceDescription } from "../../types.js";
-import { SpotlightConnector, type SpotlightConnectorOptions } from "./connector.js";
-export interface SpotlightSkillOptions {
-    /** Config alias used as the independent datasource/skill identity. */
-    readonly skillName?: string;
-    readonly instanceId?: string;
-    readonly instances?: readonly string[];
-    readonly tags?: readonly string[];
-    /** Trusted connector configuration; a pre-built connector wins. */
-    readonly connector?: SpotlightConnector;
-    readonly connectorOptions?: SpotlightConnectorOptions;
-}
-/** Live macOS Spotlight datasource. Refresh probes readiness; retrieval runs `mdfind`. */
-export declare class SpotlightSkill implements DatasourceSkill {
-    private readonly connector;
-    private readonly skillName;
-    private readonly instanceId;
-    private readonly instances;
-    private readonly tags;
-    private lastIndexedAt;
-    private lastError;
-    constructor(options?: SpotlightSkillOptions);
-    describe(): DatasourceSkillDescriptor;
-    polling(): PollingMetadata;
-    index(): Promise<DatasourceIndexResult>;
-    retrievalMethods(): readonly RetrievalMethod[];
-    describeSources(): readonly SourceDescription[];
-    skillManifest(): DatasourceSkillManifest;
-    search(query: string, options: RetrievalOptions): Promise<RetrievalResult[]>;
-    private fail;
-    private note;
-}

package/dist/minsync/kiwi.d.ts

@@ -1,17 +0,0 @@
-/** What the probe reads, injectable so every platform's layout can be checked from any host. */
-export interface KiwiProbeEnvironment {
-    readonly env: NodeJS.ProcessEnv;
-    readonly platform: NodeJS.Platform;
-    readonly home: string | undefined;
-    readonly exists: (path: string) => boolean;
-}
-/**
- * MinSync's `ko` tokenizer loads the Kiwi native library and its base model at run time (through `kiwi-rs`); neither
- * ships with MinSync. Without them every sync that needs Korean tokenization fails, so AutoRAG checks first.
- *
- * This mirrors the filesystem part of `kiwi-rs`'s discovery (`KIWI_LIBRARY_PATH` / `KIWI_MODEL_PATH`, then the
- * per-user `~/.local/kiwi` install and the well-known system directories). Names left to the OS loader search path
- * are not probed, so a library found only that way reads as missing; point `KIWI_LIBRARY_PATH` at it to fix that.
- * kiwi-rs treats an explicit path as authoritative, and so does this probe.
- */
-export declare function kiwiAvailable(environment?: KiwiProbeEnvironment): boolean;

package/dist/minsync/lexical-language.d.ts

@@ -1,18 +0,0 @@
-/**
- * The BM25 tokenizer presets AutoRAG sets in MinSync's `[lexical].language`. MinSync also has `ko`, `ja`, `zh` and
- * `ar`, which AutoRAG never selects.
- */
-export type MinSyncLexicalLanguage = "simple" | "multilingual";
-/**
- * The tokenizer AutoRAG always asks MinSync for, whatever the global `languages` setting says.
- *
- * `simple` splits on whitespace only, so a Korean word with a particle (인증서를) never matches its stem (인증서).
- * `ko` fixes that by indexing only Kiwi's morphemes, but then a compound (연구목표) exists only as 연구 and 목표 and an
- * exact-term query loses to documents that merely contain both pieces. `multilingual` indexes the original
- * whitespace-separated words as well as every analyzer's tokens, so exact terms and inflected forms both match. On
- * a sample of real Korean documents (60 stem queries with inflected forms, 40 exact-term queries, BM25 only) it had the
- * best recall on both and kept most of `simple`'s top-5 precision on exact terms, where `ko` lost most of it.
- *
- * The preset needs the Kiwi native library; see `kiwi.ts`.
- */
-export declare const MINSYNC_LEXICAL_LANGUAGE: MinSyncLexicalLanguage;