smart-web-mcp 0.35.1 → 0.37.0

package/CHANGELOG.md

@@ -8,6 +8,7 @@ The format is based on Keep a Changelog[^1] and this project uses SemVer.
 
 ### Added
 
+- `kr_place_search` now provides Korean map-style local recommendations by merging Kakao Local, Google Places, and Naver Local candidates with distance/rating/review-count ranking for 맛집, 카페, 산책, and dog-walk use cases.
 - Removed hosted API-key provider lanes (Firecrawl, Exa, Tavily, Browserless) from the smart-web runtime/benchmark surface so the default product remains no-secret/local-first.
 - `smartfetch` now accepts an MCP `output_budget` preset (`compact`, `balanced`, or `full`) and `fetch.outputBudget.preset` settings, so token-sensitive hosts can request smaller structured payloads without hand-tuning every numeric cap.
 - Added MCP host-profile guidance for Hermes/Raon, Codex, OpenCode, and Claude Code, including budget-control precedence and compact/balanced/full follow-up retrieval guidance.
@@ -76,6 +77,8 @@ The format is based on Keep a Changelog[^1] and this project uses SemVer.
 - Added compact/balanced/full MCP budget controls for `smartsearch` and `smartcrawl` summary outputs so search/crawl structuredContent is token-bounded by default, not only the rendered text channel.
 - Added `smartfetch` `pipeline.acquire.attempt_map` so default structured output shows provider seed, direct/impit, Jina Reader, archive, and browser fallback phases with succeeded/failed/skipped status.
 - Added compact `smartfetch.research_handoff` metadata for blocked, partial, or low-confidence results so downstream research workers and editable report renderers can consume the source URL plus projected `structuredContent` evidence without a new smart-web mode.
+- Added an opt-in local adaptive scraper fallback for Scrapling-compatible `scrapling extract get` commands, recorded as `adaptive_scraper` in `pipeline.acquire.attempt_map` before the browser fallback.
+- Added provider-benchmark coverage for the optional adaptive scraper lane so no-secret local candidates such as Scrapling can be compared against current smartfetch baselines without API keys.
 
 ### Changed
 
@@ -156,6 +159,20 @@ The format is based on Keep a Changelog[^1] and this project uses SemVer.
 - Korean public API helpers now validate outbound URLs and redirects before fetching, blocking localhost, private, and reserved targets unless `network.allowPrivateHosts` is explicitly enabled.
 - CI now declares read-only default repository permissions, Dependabot auto-merge pins the reusable workflow without forwarding all caller secrets, and the workflow check enforces those guards.
 
+## [0.37.0](https://github.com/jojo-labs/smart-web/compare/v0.36.0...v0.37.0) (2026-06-27)
+
+
+### Features
+
+* add Korean place search MCP tool ([#401](https://github.com/jojo-labs/smart-web/issues/401)) ([6c8e3ec](https://github.com/jojo-labs/smart-web/commit/6c8e3ec8f9152a0146313efa3e5079456e785bbf))
+
+## [0.36.0](https://github.com/jojo-labs/smart-web/compare/v0.35.1...v0.36.0) (2026-06-25)
+
+
+### Features
+
+* add adaptive scraper fallback ([#399](https://github.com/jojo-labs/smart-web/issues/399)) ([2e46b7d](https://github.com/jojo-labs/smart-web/commit/2e46b7d29e950f9a7b81b2b6fa5858e76663bc01))
+
 ## [0.35.1](https://github.com/jojo-labs/smart-web/compare/v0.35.0...v0.35.1) (2026-06-24)
 
 

package/README.md

@@ -30,6 +30,7 @@ If `smart-web` returns reproducible incorrect or misleading output, open a GitHu
 - docs export manifests are versioned and include per-document SHA-256 hashes plus resource-handle manifests for stable local Markdown mirrors; resource-handle manifests include both exported documents and metadata handles for the summary/manifest files, and successful export `structuredContent` embeds the same handle manifest for host follow-up calls
 - budgeted `smartfetch` projection records explicit truncation metadata instead of scattering silent hardcoded caps; MCP calls default to the compact preset for token-sensitive agent hosts
 - weak generic pages can recover through JSON-LD or Next.js payload extraction and same-origin RSS/Atom feed discovery before falling back to a handoff
+- optional local adaptive scraper fallback can call a Scrapling-compatible `scrapling` command for AI-targeted extraction on weak JavaScript shells before spending a full Playwright browser pass
 - paywalled article pages prefer lawful archive recovery and search-indexed reference previews before telling the host to escalate elsewhere
 - LinkedIn authwall handling prefers compliant public fallbacks such as archives and search-indexed reference metadata before giving up
 - legal paper fallback for academic URLs: DOI-aware OpenAlex, Unpaywall, Semantic Scholar, CORE discovery, and Europe PMC enrichment plus bioRxiv/medRxiv API fallback when a direct paper page is thin or blocked
@@ -280,6 +281,13 @@ npx -y smart-web-mcp --init-settings
     "enableArchiveFallback": true,
     "enableWayback": true,
     "enableArchiveMd": true,
+    // Optional local adaptive scraper fallback. This is disabled by default;
+    // install Scrapling separately and opt in only on deployments that want an
+    // extra no-secret local shell/bot-challenge extraction attempt before the
+    // built-in Playwright fallback.
+    "enableAdaptiveScraper": false,
+    "adaptiveScraperCommand": "scrapling",
+    "adaptiveScraperTimeoutMs": 20000,
     // Output projection budget for smartfetch responses. MCP calls use compact
     // by default unless the caller passes a budget control explicitly; this setting
     // controls core/CLI defaults and custom deployments.
@@ -363,6 +371,20 @@ This stays opt-in because it is a third-party relay path. When enabled, it only
 
 Relay eligibility is provider-policy based. Generic and document-like providers can opt in, while browser/auth/social/commerce providers stay off unless their provider declares support. `pipeline.policy.relay_style_allowed` and `pipeline.policy.relay_style_used` tell hosts whether a relay path was allowed and whether it was used.
 
+#### Opt in to a local adaptive scraper command
+
+```json
+{
+  "fetch": {
+    "enableAdaptiveScraper": true,
+    "adaptiveScraperCommand": "scrapling",
+    "adaptiveScraperTimeoutMs": 20000
+  }
+}
+```
+
+This lane is for local, no-secret deployments that install Scrapling or a compatible CLI themselves. It runs only after direct retrieval still looks weak, blocked, or shell-like and before the built-in browser fallback. The attempt is recorded as `adaptive_scraper` in `pipeline.acquire.attempt_map`; if the command is missing, times out, returns empty text, or does not improve the active result, smartfetch keeps the existing result and continues through the normal fallback chain.
+
 #### Tune smartfetch output budgets
 
 ```json

package/dist/korea/index.d.ts

@@ -13,4 +13,5 @@ export { runCoupang, renderCoupang } from "./coupang.js";
 export { runSpellCheck, renderSpellCheck } from "./spellcheck.js";
 export { runZipcode, renderZipcode } from "./zipcode.js";
 export { runLaw, renderLaw } from "./law.js";
+export { runPlaceSearch, renderPlaceSearch } from "./place-search.js";
 export { runPublicRestroom, renderPublicRestroom } from "./public-restroom.js";

package/dist/korea/index.js

@@ -13,5 +13,6 @@ export { runCoupang, renderCoupang } from "./coupang.js";
 export { runSpellCheck, renderSpellCheck } from "./spellcheck.js";
 export { runZipcode, renderZipcode } from "./zipcode.js";
 export { runLaw, renderLaw } from "./law.js";
+export { runPlaceSearch, renderPlaceSearch } from "./place-search.js";
 export { runPublicRestroom, renderPublicRestroom } from "./public-restroom.js";
 //# sourceMappingURL=index.js.map

package/dist/korea/index.js.map

@@ -1 +1 @@
-{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/korea/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,cAAc,EAAE,MAAM,eAAe,CAAA;AAC3D,OAAO,EAAE,SAAS,EAAE,YAAY,EAAE,MAAM,aAAa,CAAA;AACrD,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,cAAc,CAAA;AACxD,OAAO,EAAE,WAAW,EAAE,cAAc,EAAE,MAAM,gBAAgB,CAAA;AAC5D,OAAO,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,YAAY,CAAA;AAClD,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,MAAM,WAAW,CAAA;AAC/C,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAA;AAClE,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAA;AAClE,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAA;AAClE,OAAO,EAAE,YAAY,EAAE,eAAe,EAAE,MAAM,iBAAiB,CAAA;AAC/D,OAAO,EAAE,gBAAgB,EAAE,mBAAmB,EAAE,MAAM,qBAAqB,CAAA;AAC3E,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,cAAc,CAAA;AACxD,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAA;AACjE,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,cAAc,CAAA;AACxD,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,UAAU,CAAA;AAC5C,OAAO,EAAE,iBAAiB,EAAE,oBAAoB,EAAE,MAAM,sBAAsB,CAAA"}
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../src/korea/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,cAAc,EAAE,MAAM,eAAe,CAAA;AAC3D,OAAO,EAAE,SAAS,EAAE,YAAY,EAAE,MAAM,aAAa,CAAA;AACrD,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,cAAc,CAAA;AACxD,OAAO,EAAE,WAAW,EAAE,cAAc,EAAE,MAAM,gBAAgB,CAAA;AAC5D,OAAO,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,YAAY,CAAA;AAClD,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,MAAM,WAAW,CAAA;AAC/C,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAA;AAClE,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAA;AAClE,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAA;AAClE,OAAO,EAAE,YAAY,EAAE,eAAe,EAAE,MAAM,iBAAiB,CAAA;AAC/D,OAAO,EAAE,gBAAgB,EAAE,mBAAmB,EAAE,MAAM,qBAAqB,CAAA;AAC3E,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,cAAc,CAAA;AACxD,OAAO,EAAE,aAAa,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAA;AACjE,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,MAAM,cAAc,CAAA;AACxD,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,UAAU,CAAA;AAC5C,OAAO,EAAE,cAAc,EAAE,iBAAiB,EAAE,MAAM,mBAAmB,CAAA;AACrE,OAAO,EAAE,iBAAiB,EAAE,oBAAoB,EAAE,MAAM,sBAAsB,CAAA"}

package/dist/korea/place-search.d.ts

@@ -0,0 +1,67 @@
+export type PlaceIntent = "auto" | "food" | "cafe" | "walk" | "dog_walk" | "errand" | "date" | "work";
+export type PlaceSort = "blended" | "relevance" | "distance" | "rating" | "review_count";
+export type PlaceProvider = "kakao" | "google" | "naver";
+export type PlaceSearchOptions = {
+    query?: string;
+    location?: string;
+    lat?: number;
+    lon?: number;
+    radiusM?: number;
+    limit?: number;
+    intent?: PlaceIntent;
+    sort?: PlaceSort;
+    providers?: PlaceProvider[];
+    timeoutMs?: number;
+};
+export type PlaceSearchAnchor = {
+    label: string;
+    lat: number;
+    lon: number;
+    provider: "input" | "kakao" | "google";
+};
+export type ProviderError = {
+    provider: PlaceProvider | "anchor";
+    code: string;
+    status: number;
+    message: string;
+};
+export type PlaceItem = {
+    name: string;
+    category?: string;
+    address?: string;
+    road_address?: string;
+    lat?: number;
+    lon?: number;
+    distance_m?: number;
+    phone?: string;
+    rating?: number;
+    review_count?: number;
+    blog_review_count?: number;
+    open_now?: boolean;
+    business_status?: string;
+    google_place_id?: string;
+    kakao_place_id?: string;
+    naver_mapx?: string;
+    naver_mapy?: string;
+    links: Partial<Record<PlaceProvider, string>>;
+    providers: PlaceProvider[];
+    score: number;
+    score_reasons: string[];
+};
+export type PlaceSearchResult = {
+    source: "kr-place-search";
+    query: string;
+    intent: PlaceIntent;
+    sort: PlaceSort;
+    location_hint?: string;
+    anchor?: PlaceSearchAnchor;
+    radius_m: number;
+    providers_requested: PlaceProvider[];
+    providers_used: PlaceProvider[];
+    provider_errors: ProviderError[];
+    total: number;
+    items: PlaceItem[];
+    warning?: string;
+};
+export declare function runPlaceSearch(options: PlaceSearchOptions): Promise<PlaceSearchResult>;
+export declare function renderPlaceSearch(result: PlaceSearchResult): string;