@having5548/dsh-downloader 0.0.0-stage → 0.5.3

package/lib/core/ai-domains.js

@@ -0,0 +1,130 @@
+/**
+ * Built-in DIRECT allow-list for Chinese AI platforms.
+ *
+ * Why this exists: a proxy node that flaps takes down every long-lived
+ * connection that goes through it. The model APIs this harness talks to are all
+ * domestic services that never needed a proxy in the first place, so the
+ * downloader pins them to DIRECT at the highest rule priority — ahead of the
+ * subscription's own rules and ahead of the user's `extraRules`.
+ *
+ * Two things consume this list:
+ * - `UpstreamManager` turns it into `DOMAIN-SUFFIX,<domain>,DIRECT` lines so no
+ *   download to an AI endpoint can ever be routed through a node;
+ * - `session-guard.js` turns it into a `NO_PROXY` value for `$DSH_HOME/.env`, so
+ *   the harness's *own* LLM traffic stays direct even when the launcher's global
+ *   proxy policy is active.
+ *
+ * Entries are matched as suffixes (`deepseek.com` also covers
+ * `api.deepseek.com`). A stale entry is harmless — it simply never matches.
+ * @module dsh-downloader/ai-domains
+ */
+
+/** Domestic AI platform domains that must always be reached directly. */
+export const AI_DIRECT_DOMAINS = [
+	// DeepSeek — this harness's own model endpoint
+	"deepseek.com",
+	// 智谱 AI / Z.ai / ChatGLM
+	"bigmodel.cn",
+	"z.ai",
+	"chatglm.cn",
+	// 月之暗面 Kimi
+	"moonshot.cn",
+	"moonshot.ai",
+	"kimi.com",
+	// 阿里通义千问 / 百炼 DashScope
+	"dashscope.aliyuncs.com",
+	"tongyi.aliyun.com",
+	// 字节跳动 豆包 / 火山方舟 Ark
+	"volces.com",
+	"volcengine.com",
+	"doubao.com",
+	// 百度文心一言 / 千帆
+	"baidubce.com",
+	"yiyan.baidu.com",
+	// 腾讯混元
+	"hunyuan.tencent.com",
+	"tencentcloudapi.com",
+	// 科大讯飞 星火
+	"xfyun.cn",
+	"xf-yun.com",
+	// MiniMax
+	"minimax.chat",
+	"minimaxi.com",
+	// 零一万物
+	"lingyiwanwu.com",
+	"01.ai",
+	// 阶跃星辰
+	"stepfun.com",
+	"stepfun.ai",
+	// 商汤 日日新
+	"sensetime.com",
+	"sensecore.io",
+	// 百川智能
+	"baichuan-ai.com",
+	// 硅基流动 SiliconFlow
+	"siliconflow.cn",
+	"siliconflow.com",
+	// 昆仑万维 天工
+	"tiangong.cn",
+	"skywork.ai",
+	// 华为云 盘古 / ModelArts
+	"huaweicloud.com",
+	"myhuaweicloud.com",
+	// 网易有道 子曰
+	"youdao.com",
+	// 澜舟科技
+	"langboat.com",
+	// 元象 XVERSE
+	"xverse.cn",
+	// 面壁智能
+	"modelbest.cn",
+	// 出门问问
+	"mobvoi.com",
+	// 国家超算互联网 / 智算平台
+	"scnet.cn",
+	"openi.cn"
+];
+
+/**
+ * Normalize one user-supplied domain into a bare lowercase suffix.
+ * Accepts `https://api.example.com/v1`, `*.example.com`, ` example.com `.
+ * @returns the bare host, or "" when nothing usable remains.
+ */
+export function normalizeDomain(value) {
+	if (typeof value !== "string") return "";
+	let text = value.trim().toLowerCase();
+	if (text.length === 0) return "";
+	if (text.includes("://")) {
+		try {
+			text = new URL(text).hostname;
+		} catch {
+			text = text.slice(text.indexOf("://") + 3);
+		}
+	}
+	text = text.split("/")[0].split(":")[0].replace(/^\*\./, "").replace(/^\.+/, "").replace(/\.+$/, "");
+	return /^[a-z0-9.-]+$/.test(text) && text.includes(".") ? text : "";
+}
+
+/**
+ * Compose the effective DIRECT domain list.
+ * @param options - `protectAiPlatforms` toggle and user `extraDirectDomains`.
+ * @returns a de-duplicated, sorted list of bare suffixes.
+ */
+export function composeDirectDomains({ protectAiPlatforms = true, extraDirectDomains = [] } = {}) {
+	const domains = new Set();
+	if (protectAiPlatforms !== false) for (const domain of AI_DIRECT_DOMAINS) domains.add(domain);
+	for (const raw of Array.isArray(extraDirectDomains) ? extraDirectDomains : []) {
+		const normalized = normalizeDomain(raw);
+		if (normalized.length > 0) domains.add(normalized);
+	}
+	return [...domains].sort();
+}
+
+/**
+ * Turn a domain list into rule-engine lines.
+ * @param domains - bare suffixes.
+ * @returns `DOMAIN-SUFFIX,<domain>,DIRECT` lines in list order.
+ */
+export function directRuleLines(domains) {
+	return domains.map((domain) => `DOMAIN-SUFFIX,${domain},DIRECT`);
+}

package/lib/core/profiles.js

@@ -0,0 +1,372 @@
+/**
+ * FlClash-style configuration store.
+ *
+ * FlClash keeps a *list* of profiles, each of which is either a subscription
+ * URL or an imported local file, with its own label, last-update time and
+ * auto-update interval; the newest label comes from the response's
+ * `Content-Disposition`, and traffic/expiry come from `subscription-userinfo`.
+ * This module mirrors that model so the plugin can do the same from the
+ * settings panel.
+ *
+ * Layout under `$DSH_HOME/dsh-downloader/`:
+ * - `profiles.json`   — the metadata list (this module owns it)
+ * - `profiles/<id>.yaml` — the body of each imported configuration
+ *
+ * A "file" profile never auto-updates, exactly like FlClash's `realAutoUpdate`.
+ * @module dsh-downloader/profiles
+ */
+import { existsSync, mkdirSync, readFileSync, readdirSync, renameSync, rmSync, writeFileSync } from "node:fs";
+import { join } from "node:path";
+import { parseSubscription } from "./subscription.js";
+import { filenameFromDisposition } from "../download/util.js";
+
+/** Default auto-update interval for a newly added profile (FlClash's default is comparable). */
+export const DEFAULT_AUTO_UPDATE_HOURS = 24;
+
+/** Tolerance for a "合理的" auto-update interval, in hours. */
+const MIN_AUTO_UPDATE_HOURS = 1;
+const MAX_AUTO_UPDATE_HOURS = 24 * 30;
+
+/** Short, sortable, collision-resistant id (no external dependency). */
+function newId() {
+	return `${Date.now().toString(36)}${Math.random().toString(36).slice(2, 8)}`;
+}
+
+/**
+ * Parse the `subscription-userinfo` response header.
+ * Format: `upload=0; download=0; total=0; expire=0` (bytes; expire is unix seconds).
+ * @returns `{ upload, download, total, expire }`, all numbers, 0 when absent.
+ */
+export function parseSubscriptionUserinfo(headerValue) {
+	const info = { upload: 0, download: 0, total: 0, expire: 0 };
+	if (typeof headerValue !== "string" || headerValue.trim().length === 0) return info;
+	for (const part of headerValue.split(";")) {
+		const eq = part.indexOf("=");
+		if (eq === -1) continue;
+		const key = part.slice(0, eq).trim();
+		const value = Number.parseInt(part.slice(eq + 1).trim(), 10);
+		if (!Number.isFinite(value)) continue;
+		if (key in info) info[key] = value;
+	}
+	return info;
+}
+
+/**
+ * Reduce any accepted import string to the raw subscription URL.
+ *
+ * Accepts a plain `http(s)://` URL and the `clash://` forms FlClash handles as
+ * deep links: `clash://install-config?url=<encoded>`, `clash://?url=<encoded>`,
+ * and a bare `clash://<encoded url>`.
+ * @returns the URL, or null when nothing usable is found.
+ */
+export function parseImportLink(value) {
+	if (typeof value !== "string") return null;
+	const text = value.trim();
+	if (text.length === 0) return null;
+	if (/^https?:\/\//i.test(text)) return text;
+	if (!/^clash:\/\//i.test(text)) return null;
+	const query = text.slice(text.indexOf("?") + 1);
+	const params = new URLSearchParams(text.includes("?") ? query : "");
+	const fromQuery = params.get("url") ?? params.get("Url") ?? params.get("URL");
+	if (typeof fromQuery === "string" && /^https?:\/\//i.test(fromQuery)) return fromQuery;
+	// clash://install-config/<encoded> or clash://<encoded>
+	const tail = text.replace(/^clash:\/\//i, "").replace(/^(install-config|config)\/?/i, "");
+	if (tail.length === 0) return null;
+	try {
+		const decoded = decodeURIComponent(tail);
+		if (/^https?:\/\//i.test(decoded)) return decoded;
+	} catch {
+		// Not percent-encoded.
+	}
+	return /^https?:\/\//i.test(tail) ? tail : null;
+}
+
+/** A stable, human label for a URL whose response carried no filename. */
+export function labelFromUrl(url) {
+	try {
+		const parsed = new URL(url);
+		return parsed.hostname.replace(/^www\./, "");
+	} catch {
+		return "配置";
+	}
+}
+
+/** Reject a label that would escape the profiles directory. */
+function safeLabel(value, fallback) {
+	const text = String(value ?? "").trim().replace(/[\\/:*?"<>|]+/g, "_").slice(0, 60);
+	return text.length > 0 ? text : fallback;
+}
+
+/** Validate a configuration body; throws with a readable reason when unusable. */
+export function validateConfigText(raw) {
+	if (typeof raw !== "string" || raw.trim().length === 0) throw new Error("配置内容为空");
+	let parsed;
+	try {
+		parsed = parseSubscription(raw);
+	} catch (error) {
+		throw new Error(`无法解析配置：${error instanceof Error ? error.message : String(error)}`);
+	}
+	if (parsed.names.length === 0) throw new Error("配置里没有解析到任何节点");
+	return parsed;
+}
+
+export class ProfileStore {
+	#dataDir;
+	/** @type {{version: number, currentId: string|null, profiles: object[]}} */
+	#state = { version: 1, currentId: null, profiles: [] };
+	#loaded = false;
+
+	constructor({ dataDir }) {
+		this.#dataDir = dataDir;
+	}
+
+	get metaPath() {
+		return join(this.#dataDir, "profiles.json");
+	}
+
+	get bodyDir() {
+		return join(this.#dataDir, "profiles");
+	}
+
+	bodyPath(id) {
+		return join(this.bodyDir, `${id}.yaml`);
+	}
+
+	/** Read the metadata file (idempotent). */
+	load() {
+		if (this.#loaded) return this.#state;
+		this.#loaded = true;
+		try {
+			const parsed = JSON.parse(readFileSync(this.metaPath, "utf8"));
+			if (parsed !== null && typeof parsed === "object" && Array.isArray(parsed.profiles)) {
+				this.#state = {
+					version: 1,
+					currentId: typeof parsed.currentId === "string" ? parsed.currentId : null,
+					profiles: parsed.profiles.filter((item) => item !== null && typeof item === "object" && typeof item.id === "string")
+				};
+			}
+		} catch {
+			// First run, or unreadable metadata: start empty rather than fail.
+		}
+		if (this.#state.currentId === null && this.#state.profiles.length > 0) this.#state.currentId = this.#state.profiles[0].id;
+		return this.#state;
+	}
+
+	#persist() {
+		mkdirSync(this.#dataDir, { recursive: true });
+		const temp = `${this.metaPath}.tmp`;
+		writeFileSync(temp, JSON.stringify(this.#state, null, 2), "utf8");
+		renameSync(temp, this.metaPath);
+	}
+
+	#writeBody(id, raw) {
+		mkdirSync(this.bodyDir, { recursive: true });
+		const target = this.bodyPath(id);
+		const temp = `${target}.tmp`;
+		writeFileSync(temp, raw, "utf8");
+		renameSync(temp, target);
+	}
+
+	/** Every profile, with the derived flags the UI needs. */
+	list() {
+		const state = this.load();
+		return state.profiles.map((profile) => ({
+			...profile,
+			current: profile.id === state.currentId,
+			dueAt: profile.source === "url" && profile.autoUpdate === true && typeof profile.lastUpdateAt === "number"
+				? profile.lastUpdateAt + profile.autoUpdateHours * 3600 * 1000
+				: null
+		}));
+	}
+
+	/** The selected profile's metadata, or null. */
+	current() {
+		const state = this.load();
+		return state.profiles.find((profile) => profile.id === state.currentId) ?? null;
+	}
+
+	get(id) {
+		return this.load().profiles.find((profile) => profile.id === id) ?? null;
+	}
+
+	/** The configuration body of one profile, or null when the file is gone. */
+	body(id) {
+		const path = this.bodyPath(id);
+		try {
+			return readFileSync(path, "utf8");
+		} catch {
+			return null;
+		}
+	}
+
+	/** The selected profile's body, or null. */
+	currentBody() {
+		const profile = this.current();
+		return profile === null ? null : this.body(profile.id);
+	}
+
+	/** Profiles whose auto-update interval has elapsed. */
+	dueProfiles(now = Date.now()) {
+		return this.list().filter((profile) => profile.dueAt !== null && profile.dueAt <= now);
+	}
+
+	/**
+	 * Add a subscription URL.
+	 * @param options - `label`, `autoUpdateHours`, and the `fetcher` used to download it.
+	 */
+	async addFromUrl(url, { label, autoUpdateHours = DEFAULT_AUTO_UPDATE_HOURS, fetcher, fetchProxyUrl = "" } = {}) {
+		if (typeof fetcher !== "function") throw new Error("内部错误：缺少抓取函数");
+		const response = await fetcher(url, fetchProxyUrl);
+		const raw = typeof response === "string" ? response : response.text;
+		const headers = typeof response === "string" ? {} : response.headers ?? {};
+		validateConfigText(raw);
+		const id = newId();
+		this.#writeBody(id, raw);
+		const profile = {
+			id,
+			label: safeLabel(label ?? filenameFromDisposition(headers["content-disposition"]) ?? "", labelFromUrl(url)),
+			source: "url",
+			url,
+			autoUpdate: true,
+			autoUpdateHours: clampHours(autoUpdateHours),
+			lastUpdateAt: Date.now(),
+			addedAt: Date.now(),
+			subscriptionInfo: parseSubscriptionUserinfo(headers["subscription-userinfo"])
+		};
+		const state = this.load();
+		state.profiles.push(profile);
+		if (state.currentId === null) state.currentId = id;
+		this.#persist();
+		return profile;
+	}
+
+	/** Add a configuration the user supplied as a file (never auto-updates). */
+	addFromFile(name, raw, { label } = {}) {
+		validateConfigText(raw);
+		const id = newId();
+		this.#writeBody(id, raw);
+		const fallback = String(name ?? "").replace(/\.[^.]+$/, "") || "本地配置";
+		const profile = {
+			id,
+			label: safeLabel(label ?? fallback, "本地配置"),
+			source: "file",
+			url: "",
+			autoUpdate: false,
+			autoUpdateHours: DEFAULT_AUTO_UPDATE_HOURS,
+			lastUpdateAt: Date.now(),
+			addedAt: Date.now(),
+			subscriptionInfo: null,
+			fileName: String(name ?? "")
+		};
+		const state = this.load();
+		state.profiles.push(profile);
+		if (state.currentId === null) state.currentId = id;
+		this.#persist();
+		return profile;
+	}
+
+	/** Re-download a URL profile. File profiles are refused, like FlClash. */
+	async update(id, { fetcher, fetchProxyUrl = "" } = {}) {
+		const profile = this.get(id);
+		if (profile === null) throw new Error(`找不到配置：${id}`);
+		if (profile.source !== "url" || profile.url.length === 0) throw new Error("本地文件配置不能从网络更新");
+		const response = await fetcher(profile.url, fetchProxyUrl);
+		const raw = typeof response === "string" ? response : response.text;
+		const headers = typeof response === "string" ? {} : response.headers ?? {};
+		validateConfigText(raw);
+		this.#writeBody(id, raw);
+		const fromHeader = filenameFromDisposition(headers["content-disposition"]);
+		profile.label = safeLabel(fromHeader ?? profile.label, profile.label);
+		profile.lastUpdateAt = Date.now();
+		profile.subscriptionInfo = parseSubscriptionUserinfo(headers["subscription-userinfo"]);
+		this.#persist();
+		return this.get(id);
+	}
+
+	remove(id) {
+		const state = this.load();
+		const index = state.profiles.findIndex((profile) => profile.id === id);
+		if (index === -1) throw new Error(`找不到配置：${id}`);
+		state.profiles.splice(index, 1);
+		try {
+			rmSync(this.bodyPath(id), { force: true });
+		} catch {
+			// The metadata is the source of truth; a stray body is harmless.
+		}
+		if (state.currentId === id) state.currentId = state.profiles.length > 0 ? state.profiles[0].id : null;
+		this.#persist();
+		return this.current();
+	}
+
+	/** Rename a profile (FlClash lets the label be edited). */
+	rename(id, label) {
+		const profile = this.get(id);
+		if (profile === null) throw new Error(`找不到配置：${id}`);
+		profile.label = safeLabel(label, profile.label);
+		this.#persist();
+		return profile;
+	}
+
+	select(id) {
+		const state = this.load();
+		if (!state.profiles.some((profile) => profile.id === id)) throw new Error(`找不到配置：${id}`);
+		state.currentId = id;
+		this.#persist();
+		return this.current();
+	}
+
+	/** Reorder by an explicit id list; unknown ids are ignored, missing ones appended. */
+	reorder(ids) {
+		const state = this.load();
+		const wanted = Array.isArray(ids) ? ids : [];
+		const ordered = [];
+		for (const id of wanted) {
+			const found = state.profiles.find((profile) => profile.id === id);
+			if (found !== undefined && !ordered.includes(found)) ordered.push(found);
+		}
+		for (const profile of state.profiles) if (!ordered.includes(profile)) ordered.push(profile);
+		state.profiles = ordered;
+		this.#persist();
+		return this.list();
+	}
+
+	/** Toggle/replace the auto-update settings of one profile. */
+	setAutoUpdate(id, enabled, hours) {
+		const profile = this.get(id);
+		if (profile === null) throw new Error(`找不到配置：${id}`);
+		if (profile.source !== "url") throw new Error("本地文件配置没有自动更新");
+		profile.autoUpdate = enabled !== false;
+		if (hours !== undefined) profile.autoUpdateHours = clampHours(hours);
+		this.#persist();
+		return profile;
+	}
+
+	/** Removed bodies left behind by a crash; keeps the directory tidy. */
+	pruneOrphanBodies() {
+		const state = this.load();
+		const known = new Set(state.profiles.map((profile) => `${profile.id}.yaml`));
+		let removed = 0;
+		try {
+			for (const name of readdirSync(this.bodyDir)) {
+				if (known.has(name) || name.endsWith(".tmp")) continue;
+				rmSync(join(this.bodyDir, name), { force: true });
+				removed += 1;
+			}
+		} catch {
+			// No directory yet.
+		}
+		return removed;
+	}
+
+	/** True when a metadata file already exists (used for the legacy migration path). */
+	get exists() {
+		return existsSync(this.metaPath);
+	}
+}
+
+/** Keep an interval inside a sane range. */
+function clampHours(value) {
+	const hours = Number(value);
+	if (!Number.isFinite(hours)) return DEFAULT_AUTO_UPDATE_HOURS;
+	return Math.min(MAX_AUTO_UPDATE_HOURS, Math.max(MIN_AUTO_UPDATE_HOURS, Math.round(hours)));
+}