@klarkxy/dsh-web-search-manager 0.1.8 → 0.1.10

package/lib/index.js

@@ -4,8 +4,129 @@ import { WebError } from "@deepseek-ai/dsh-web";
 import { DEEPSEEK_DEFAULT_API_VERSION, DEEPSEEK_DEFAULT_BASE_URL, DEEPSEEK_DEFAULT_MAX_TOKENS, DEEPSEEK_DEFAULT_MAX_USES, DEEPSEEK_DEFAULT_MODEL, DeepSeekSearchProvider } from "@deepseek-ai/dsh-web-search-deepseek";
 import { ExaSearchProvider } from "@deepseek-ai/dsh-web-search-exa";
 import { HttpFetchProvider } from "@deepseek-ai/dsh-web-fetch-http";
+import { isIP } from "node:net";
 import { defineDomain, domainTable } from "@deepseek-ai/dsh-storage-domain";
 import { z } from "zod";
+//#region src/jina-fetch.ts
+/** Do not send obvious local addresses, credentials, or nested Reader URLs to a third party. */
+function readerUrl(value) {
+	try {
+		const url = new URL(value);
+		const host = url.hostname.replace(/^\[|\]$/g, "").replace(/\.$/, "").toLowerCase();
+		if (!["http:", "https:"].includes(url.protocol) || url.username || url.password || isIP(host) || !host.includes(".") || host === "r.jina.ai" || /\.(?:localhost|local|internal|lan|home|test|invalid)$/.test(host)) return void 0;
+		return `https://r.jina.ai/${url.href}`;
+	} catch {
+		return;
+	}
+}
+function object(value) {
+	return value !== null && typeof value === "object" && !Array.isArray(value) ? value : void 0;
+}
+/** Accept Reader's text envelope or negotiated JSON, but not error/empty responses. */
+function readerContent(result) {
+	if (result.statusCode < 200 || result.statusCode >= 300 || result.body.kind !== "text") return void 0;
+	const content = result.body.content;
+	if (content.trimStart().startsWith("{")) try {
+		const envelope = object(JSON.parse(content));
+		if (!envelope || envelope.code !== void 0 && envelope.code !== 200) return void 0;
+		const page = object(envelope.data) ?? envelope;
+		return typeof page.content === "string" && page.content.trim() && !page.warning ? page.content : void 0;
+	} catch {
+		return;
+	}
+	const marker = /^Markdown Content:[\t ]*\r?$/m.exec(content);
+	if (!marker || /^Warning:/im.test(content.slice(0, marker.index))) return void 0;
+	return content.slice(marker.index + marker[0].length).trim() ? content : void 0;
+}
+/** Bound the Reader stage even if a provider fails to honour its cooperative signal. */
+async function readWithTimeout(read, timeoutMs, parent) {
+	parent?.throwIfAborted();
+	const controller = new AbortController();
+	const signal = parent ? AbortSignal.any([parent, controller.signal]) : controller.signal;
+	const timer = setTimeout(() => controller.abort(), timeoutMs);
+	let off = () => {};
+	const aborted = new Promise((_, reject) => {
+		const onAbort = () => reject(signal.reason);
+		signal.addEventListener("abort", onAbort, { once: true });
+		off = () => signal.removeEventListener("abort", onAbort);
+	});
+	try {
+		return await Promise.race([Promise.resolve().then(() => {
+			signal.throwIfAborted();
+			return read(signal);
+		}), aborted]);
+	} finally {
+		clearTimeout(timer);
+		off();
+	}
+}
+/** One managed fetch route: anonymous Jina Reader first, existing safe HTTP transport second. */
+var JinaFirstFetchProvider = class {
+	id = "http";
+	options;
+	constructor(options) {
+		if (!Number.isInteger(options.readerTimeoutMs) || options.readerTimeoutMs < 1 || options.readerTimeoutMs > 2147483647) throw new Error("Invalid Reader timeout");
+		this.options = options;
+	}
+	available() {
+		return this.options.direct.available();
+	}
+	async fetch(request, signal) {
+		signal?.throwIfAborted();
+		const url = readerUrl(request.url);
+		if (url && this.options.reader.available()) try {
+			const result = await readWithTimeout((combined) => this.options.reader.fetch({ url }, combined), this.options.readerTimeoutMs, signal);
+			signal?.throwIfAborted();
+			const content = readerContent(result);
+			if (content !== void 0) return {
+				...result,
+				url: request.url,
+				body: {
+					kind: "text",
+					content
+				}
+			};
+		} catch {
+			signal?.throwIfAborted();
+		}
+		signal?.throwIfAborted();
+		return this.options.direct.fetch(request, signal);
+	}
+};
+//#endregion
+//#region src/provider-error.ts
+/** Preserve diagnostics, removing known credential values and common authentication fields only. */
+function redactDescription(text, secrets = []) {
+	let result = text;
+	for (const secret of secrets.filter(Boolean).sort((a, b) => b.length - a.length)) for (const value of /* @__PURE__ */ new Set([secret, encodeURIComponent(secret)])) result = result.split(value).join("[REDACTED]");
+	result = result.replace(/(https?:\/\/)[^\s/@]+@/gi, "$1[REDACTED]@").replace(/\b(Bearer|Basic)\s+[A-Za-z0-9._~+/=-]+/gi, "$1 [REDACTED]").replace(/([?&](?:api[_-]?key|access[_-]?token|refresh[_-]?token|token|password|secret|key)=)[^&#\s]*/gi, "$1[REDACTED]").replace(/(["']?(?:api[_-]?key|access[_-]?token|refresh[_-]?token|password|secret|authorization|cookie|set-cookie)["']?\s*[:=]\s*)(?:"[^"]*"|'[^']*'|[^\s,;}]+)/gi, "$1[REDACTED]");
+	return result.slice(0, 1e3);
+}
+var HttpStatusError = class extends WebError {
+	status;
+	constructor(status) {
+		const safe = Number.isInteger(status) && status >= 100 && status <= 599 ? status : 0;
+		super(safe ? `供应商请求失败（HTTP ${safe}）。` : "供应商请求失败。", "WEB_PROVIDER_ERROR");
+		this.status = safe;
+	}
+};
+/** Keep upstream code and descriptions, not the error object, stack, response or headers. */
+function providerError(error, secrets = []) {
+	const seen = /* @__PURE__ */ new Set();
+	const descriptions = [];
+	let code;
+	let current = error;
+	for (let depth = 0; depth < 3 && current && typeof current === "object" && !seen.has(current); depth++) {
+		seen.add(current);
+		const item = current;
+		if (!code && typeof item.code === "string" && item.code) code = item.code;
+		const description = typeof item.description === "string" && item.description.trim() ? item.description : item.message;
+		if (typeof description === "string" && description.trim() && !descriptions.includes(description)) descriptions.push(description);
+		current = item.cause;
+	}
+	if (typeof error === "string" && error.trim()) descriptions.push(error);
+	return new WebError(redactDescription(descriptions.join("；") || "供应商未提供错误描述。", secrets), redactDescription(code || "WEB_PROVIDER_ERROR", secrets));
+}
 const SEARCH_ENGINE_PAGE = "https://lite.duckduckgo.com/lite/";
 const MAX_RESPONSE_BYTES$2 = 2e6;
 const USER_AGENT = "Mozilla/5.0 (compatible; dsh-editor-search/1.0)";
@@ -117,7 +238,7 @@ var SearchEngineProvider = class {
 			});
 			if (!response.ok) {
 				await response.body?.cancel().catch(() => {});
-				throw new WebError(`搜索页请求失败（HTTP ${response.status}）。`, "WEB_PROVIDER_ERROR");
+				throw new HttpStatusError(response.status);
 			}
 			const sources = parseSearchEngineHtml(await readText(response, signal), response.url || "https://lite.duckduckgo.com/lite/");
 			return {
@@ -126,8 +247,7 @@ var SearchEngineProvider = class {
 			};
 		} catch (error) {
 			if (signal?.aborted) throw new WebError("搜索已取消。", "WEB_ABORTED");
-			if (error instanceof WebError) throw error;
-			throw new WebError("无法读取公开搜索页。", "WEB_PROVIDER_ERROR");
+			throw providerError(error);
 		}
 	}
 };
@@ -228,7 +348,7 @@ function createRestSearch(spec) {
 				});
 				if (!response.ok) {
 					await response.body?.cancel().catch(() => {});
-					throw new WebError(`搜索请求失败（HTTP ${response.status}）。`, "WEB_PROVIDER_ERROR");
+					throw new HttpStatusError(response.status);
 				}
 				const sources = spec.parse(await readJSON$1(response, signal));
 				return {
@@ -237,8 +357,7 @@ function createRestSearch(spec) {
 				};
 			} catch (error) {
 				if (signal?.aborted) throw new WebError("搜索已取消。", "WEB_ABORTED");
-				if (error instanceof WebError) throw error;
-				throw new WebError("无法完成搜索请求。", "WEB_PROVIDER_ERROR");
+				throw providerError(error, [this.options.apiKey]);
 			}
 		}
 	};
@@ -395,7 +514,7 @@ var TavilySearchProvider = class {
 			});
 			if (!response.ok) {
 				await response.body?.cancel().catch(() => {});
-				throw new WebError(`Tavily 请求失败（HTTP ${response.status}）。`, "WEB_PROVIDER_ERROR");
+				throw new HttpStatusError(response.status);
 			}
 			const payload = record(await readJSON(response, signal));
 			if (!Array.isArray(payload?.results)) throw new Error("Invalid response shape");
@@ -423,8 +542,7 @@ var TavilySearchProvider = class {
 			};
 		} catch (error) {
 			if (signal?.aborted) throw new WebError("Tavily 搜索已取消。", "WEB_ABORTED");
-			if (error instanceof WebError) throw error;
-			throw new WebError("Tavily 返回无效响应或网络连接失败。", "WEB_PROVIDER_ERROR");
+			throw providerError(error, [this.options.apiKey]);
 		}
 	}
 };
@@ -440,7 +558,7 @@ function registerBuiltins(manager) {
 	const offDeepSeek = manager.registerSearchProvider({
 		id: "deepseek-official",
 		label: "DeepSeek 搜索",
-		description: "使用已配置的 DeepSeek API Key。可能产生额外搜索费用。",
+		description: "使用已配置的 DeepSeek API Key，可能产生额外搜索费用。",
 		defaultBaseURL: DEEPSEEK_DEFAULT_BASE_URL,
 		credentialRef: "DEEPSEEK_API_KEY",
 		credentialShared: true,
@@ -531,16 +649,28 @@ function registerBuiltins(manager) {
 	}));
 	const offHttp = manager.registerFetchProvider({
 		id: "http",
-		label: "HTTP 读取",
+		label: "网页读取",
 		billing: "none",
-		description: "本机直连公开网页，无需 Key，不产生搜索费用。"
-	}, (options) => new HttpFetchProvider({
-		maxResponseBytes: 5e6,
-		maxBodyChars: options.maxFetchChars,
-		timeoutMs: options.timeoutMs,
-		maxRedirects: 5,
-		userAgent: "dsh-editor/managed-web"
-	}));
+		description: "优先匿名 Jina Reader 读取，失败后本机 HTTP 直连；无需 Key。"
+	}, (options) => {
+		const limits = {
+			maxResponseBytes: 5e6,
+			maxBodyChars: options.maxFetchChars,
+			timeoutMs: options.timeoutMs,
+			maxRedirects: 5,
+			userAgent: "dsh-editor/managed-web"
+		};
+		const readerTimeoutMs = Math.min(15e3, Math.floor(options.timeoutMs / 2));
+		return new JinaFirstFetchProvider({
+			reader: new HttpFetchProvider({
+				...limits,
+				maxBodyChars: limits.maxResponseBytes,
+				timeoutMs: readerTimeoutMs
+			}),
+			direct: new HttpFetchProvider(limits),
+			readerTimeoutMs
+		});
+	});
 	return () => {
 		offHttp();
 		offTavily();
@@ -859,16 +989,17 @@ var WebSearchManager = class {
 		}));
 	}
 	async execute(entry, signal, run) {
-		if (!this.selected(entry)) fail("WEB_DISABLED", "请在设置 → 网络搜索中配置凭据并主动启用。");
+		if (!this.selected(entry)) fail("WEB_DISABLED", "请在设置 → 网络搜索中配置凭据并启用。");
 		const controller = new AbortController();
 		const timeout = setTimeout(() => controller.abort(), this.settings.timeoutMs);
 		const combined = signal ? AbortSignal.any([signal, controller.signal]) : controller.signal;
 		const epoch = this.epoch;
 		entry.active.add(controller);
 		let started = false;
+		let apiKey;
 		try {
 			const ref = entry.descriptor.credentialRef;
-			const apiKey = ref ? await abortable(() => this.options.resolveCredential(ref), combined) : void 0;
+			apiKey = ref ? await abortable(() => this.options.resolveCredential(ref), combined) : void 0;
 			if (ref && !apiKey?.trim()) {
 				entry.configured = false;
 				this.abort(entry);
@@ -889,11 +1020,9 @@ var WebSearchManager = class {
 			return result;
 		} catch (error) {
 			if (started) entry.failures += 1;
-			const code = error instanceof WebError ? error.code : void 0;
-			if (code === "WEB_CREDENTIAL_MISSING") throw new WebError("搜索凭据不可用，请重新配置。", "WEB_CREDENTIAL_MISSING");
-			if (code === "WEB_DISABLED") throw new WebError("网络搜索未启用或设置已改变。", "WEB_DISABLED");
-			if (combined.aborted || code === "WEB_ABORTED") throw new WebError("网络请求已取消或超时。", "WEB_ABORTED");
-			throw new WebError("网络供应商请求失败，请检查凭据、额度和连接。未切换到其他供应商。", "WEB_PROVIDER_ERROR");
+			if (combined.aborted && !(error instanceof WebError && error.code === "WEB_CREDENTIAL_MISSING")) throw new WebError("网络请求已取消或超时。", "WEB_ABORTED");
+			const safe = providerError(error, apiKey ? [apiKey] : []);
+			throw new WebError(`[${entry.kind}:${entry.descriptor.id}] ${safe.message}`, safe.code);
 		} finally {
 			clearTimeout(timeout);
 			entry.active.delete(controller);
@@ -923,7 +1052,7 @@ var WebSearchManager = class {
 			await this.refresh();
 			proposed.searchOrder = this.searchOrder(proposed);
 			proposed.searchProvider = this.activeSearchId(proposed);
-			if (proposed.searchEnabled && !proposed.searchProvider) fail("WEB_CREDENTIAL_MISSING", "请先选择已安装的供应商并填写有效凭据。");
+			if (proposed.searchEnabled && !proposed.searchProvider) fail("WEB_CREDENTIAL_MISSING", "需选择已安装的供应商并填写有效凭据。");
 			if (proposed.searchEnabled && this.entries.get(providerKey("fetch", "http"))) {
 				proposed.fetchEnabled = true;
 				proposed.fetchProvider = "http";
@@ -932,7 +1061,7 @@ var WebSearchManager = class {
 			if (proposed.fetchEnabled) {
 				const entry = this.entries.get(providerKey("fetch", proposed.fetchProvider));
 				if (!entry) proposed.fetchEnabled = false;
-				else if (!entry.configured) fail("WEB_CREDENTIAL_MISSING", "请先选择已安装的供应商并填写有效凭据。");
+				else if (!entry.configured) fail("WEB_CREDENTIAL_MISSING", "需选择已安装的供应商并填写有效凭据。");
 			}
 			this.suspended = true;
 			this.epoch += 1;
@@ -943,7 +1072,7 @@ var WebSearchManager = class {
 			} catch {
 				this.storageFailed = true;
 				this.notify();
-				fail("WEB_CONFIG_SAVE_FAILED", "设置保存失败。为避免意外联网，本次运行已暂停网络访问。");
+				fail("WEB_CONFIG_SAVE_FAILED", "保存失败，本次运行已暂停网络访问以防意外联网。");
 			}
 			this.settings = proposed;
 			this.storageFailed = false;