@webcrawl/mcp 1.4.3 → 1.4.5

package/dist/core/engine/chrome.d.ts

@@ -1,14 +1,33 @@
-/**
- * Path to a real installed Chrome/Chromium, or null.
- * Priority: WEBCRAWL_CHROME_PATH, INTERACT_CHROME_PATH, CHROME_PATH, then per-OS
- * defaults (same locations `webcrawl doctor` checks).
- */
-export declare function installedChromePath(): string | null;
-export declare function isLocalChromeAvailable(): boolean;
+export { DEFAULT_CONNECT_TIMEOUT_MS, DEFAULT_TIMEOUT_MS, VIRTUAL_TIME_MS, discoverChromeBinaries, findPlaywrightHeadlessShell, installedChromePath, isLocalChromeAvailable, parseDevToolsActivePort, chromeCdpArgs, chromeDumpDomArgs, killProcessGroup, resolveChromeConnectTimeoutMs, resolveChromeTimeoutMs, type ChromeHeadlessMode, } from "./chrome-launch.js";
+export { ChromeConnectError } from "./chrome-cdp.js";
+export interface ChromeRenderAttempt {
+    ok: boolean;
+    method: string;
+    ms: number;
+    error?: string;
+}
+export interface ChromeRenderResult {
+    html: string | null;
+    method?: string;
+    error?: string;
+    attempts: ChromeRenderAttempt[];
+}
 /**
  * Render `url` in local Chrome and return the hydrated HTML, or null when
  * Chrome is missing, times out, or produces nothing.
  */
 export declare function renderViaLocalChrome(url: string, opts?: {
     timeoutMs?: number;
+    connectTimeoutMs?: number;
+    binaries?: string[];
 }): Promise<string | null>;
+/**
+ * Same as `renderViaLocalChrome` but with the attempt log and an actionable
+ * error when every driver fails. Callers that need to report honest JS-render
+ * failures should use this.
+ */
+export declare function renderViaLocalChromeDetailed(url: string, opts?: {
+    timeoutMs?: number;
+    connectTimeoutMs?: number;
+    binaries?: string[];
+}): Promise<ChromeRenderResult>;

package/dist/core/engine/chrome.js

@@ -1,131 +1,208 @@
 /**
- * Local Chrome dump-dom renderer (JS pages the native fetch ladder cannot hydrate).
+ * Local Chrome renderer (JS pages the native fetch ladder cannot hydrate).
  *
- * Uses a user-installed Chrome/Chromium binary — the same one `webcrawl doctor`
- * reports — with `--dump-dom` + virtual time so `document.write` / client-rendered
- * content appears in the HTML. Playwright is not required; this is the zero-key
- * local JS tier.
+ * Primary driver is Chrome DevTools Protocol:
+ * launch headless Chrome, navigate, wait until the DOM is content-ready
+ * (load + 200ms mutation-idle), then read `documentElement.outerHTML`.
+ *
+ * dump-dom is a last-resort fallback: Chrome 156 hangs on dump-dom with a
+ * fresh profile, which used to look like a successful scrape of the
+ * unrendered nav shell. Every attempt has a hard timeout and a process-group
+ * kill. Playwright's chrome-headless-shell is discovered automatically when
+ * present.
  */
 import { spawn } from "node:child_process";
-import { existsSync, mkdtempSync, rmSync } from "node:fs";
+import { mkdtempSync, rmSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { join } from "node:path";
-const DUMP_MAX_BYTES = 2 * 1024 * 1024;
-const DEFAULT_TIMEOUT_MS = 20_000;
-const VIRTUAL_TIME_MS = 10_000;
-/**
- * Path to a real installed Chrome/Chromium, or null.
- * Priority: WEBCRAWL_CHROME_PATH, INTERACT_CHROME_PATH, CHROME_PATH, then per-OS
- * defaults (same locations `webcrawl doctor` checks).
- */
-export function installedChromePath() {
-    const fromEnv = process.env.WEBCRAWL_CHROME_PATH?.trim() ||
-        process.env.INTERACT_CHROME_PATH?.trim() ||
-        process.env.CHROME_PATH?.trim();
-    if (fromEnv)
-        return fromEnv;
-    const candidates = [];
-    if (process.platform === "darwin") {
-        candidates.push("/Applications/Google Chrome.app/Contents/MacOS/Google Chrome", `${process.env.HOME || ""}/Applications/Google Chrome.app/Contents/MacOS/Google Chrome`);
-    }
-    else if (process.platform === "win32") {
-        candidates.push(`${process.env["ProgramFiles"] || "C:\\Program Files"}\\Google\\Chrome\\Application\\chrome.exe`, `${process.env["ProgramFiles(x86)"] || "C:\\Program Files (x86)"}\\Google\\Chrome\\Application\\chrome.exe`);
+import { ChromeConnectError, renderUrlViaCdp } from "./chrome-cdp.js";
+import { DUMP_MAX_BYTES, chromeDumpDomArgs, discoverChromeBinaries, killProcessGroup, resolveChromeConnectTimeoutMs, resolveChromeTimeoutMs, } from "./chrome-launch.js";
+export { DEFAULT_CONNECT_TIMEOUT_MS, DEFAULT_TIMEOUT_MS, VIRTUAL_TIME_MS, discoverChromeBinaries, findPlaywrightHeadlessShell, installedChromePath, isLocalChromeAvailable, parseDevToolsActivePort, chromeCdpArgs, chromeDumpDomArgs, killProcessGroup, resolveChromeConnectTimeoutMs, resolveChromeTimeoutMs, } from "./chrome-launch.js";
+export { ChromeConnectError } from "./chrome-cdp.js";
+function attemptPlan(binary) {
+    const isShell = /headless[-_]?shell/i.test(binary);
+    if (isShell) {
+        return [
+            { driver: "cdp", headless: "shell" },
+            { driver: "dump-dom", headless: "shell" },
+        ];
     }
-    else {
-        candidates.push("/usr/bin/google-chrome", "/usr/bin/google-chrome-stable", "/usr/bin/chromium-browser", "/usr/bin/chromium", "/snap/bin/chromium");
-    }
-    for (const p of candidates) {
-        try {
-            if (p && existsSync(p))
-                return p;
-        }
-        catch {
-            /* not installed here */
-        }
-    }
-    return null;
+    return [
+        { driver: "cdp", headless: "new" },
+        { driver: "cdp", headless: "old" },
+        { driver: "dump-dom", headless: "old" },
+        { driver: "dump-dom", headless: "new" },
+    ];
 }
-export function isLocalChromeAvailable() {
-    return installedChromePath() !== null;
+function methodName(binary, kind) {
+    const base = binary.split(/[/\\]/).pop() || binary;
+    return `${kind.driver}/${kind.headless} (${base})`;
 }
 /**
  * Render `url` in local Chrome and return the hydrated HTML, or null when
  * Chrome is missing, times out, or produces nothing.
  */
 export async function renderViaLocalChrome(url, opts = {}) {
-    const chrome = installedChromePath();
-    if (!chrome)
-        return null;
-    const timeoutMs = opts.timeoutMs ?? DEFAULT_TIMEOUT_MS;
+    const result = await renderViaLocalChromeDetailed(url, opts);
+    return result.html;
+}
+/**
+ * Same as `renderViaLocalChrome` but with the attempt log and an actionable
+ * error when every driver fails. Callers that need to report honest JS-render
+ * failures should use this.
+ */
+export async function renderViaLocalChromeDetailed(url, opts = {}) {
+    const binaries = opts.binaries?.length ? opts.binaries : discoverChromeBinaries();
+    if (binaries.length === 0) {
+        return {
+            html: null,
+            error: "No local Chrome/Chromium binary found. Install Chrome, set WEBCRAWL_CHROME_PATH, " +
+                "or run `npx playwright install chromium-headless-shell`. See `webcrawl doctor`.",
+            attempts: [],
+        };
+    }
+    const budgetMs = resolveChromeTimeoutMs(opts.timeoutMs);
+    const connectTimeoutMs = resolveChromeConnectTimeoutMs(opts.connectTimeoutMs);
+    const started = Date.now();
+    const attempts = [];
+    let lastError = "local Chrome produced no content";
+    for (const binary of binaries) {
+        let hungConnect = false;
+        for (const kind of attemptPlan(binary)) {
+            const remaining = budgetMs - (Date.now() - started);
+            if (remaining < 1_200)
+                break;
+            // A process that launched but never answered CDP will hang dump-dom too.
+            if (kind.driver === "dump-dom" && hungConnect)
+                continue;
+            const method = methodName(binary, kind);
+            const attemptBudget = kind.driver === "cdp"
+                ? Math.min(remaining, kind.headless === "new" ? 12_000 : 8_000)
+                : Math.min(remaining, 6_000);
+            const t0 = Date.now();
+            try {
+                const html = kind.driver === "cdp"
+                    ? (await renderUrlViaCdp(binary, url, {
+                        timeoutMs: attemptBudget,
+                        connectTimeoutMs: Math.min(connectTimeoutMs, remaining),
+                        headless: kind.headless,
+                    })).html
+                    : await dumpDom(binary, url, kind.headless, attemptBudget);
+                const ms = Date.now() - t0;
+                if (html && html.length > 100) {
+                    attempts.push({ ok: true, method, ms });
+                    return { html, method, attempts };
+                }
+                lastError = `${method} returned empty HTML`;
+                attempts.push({ ok: false, method, ms, error: lastError });
+            }
+            catch (err) {
+                const ms = Date.now() - t0;
+                lastError = `${method}: ${err instanceof Error ? err.message : String(err)}`;
+                attempts.push({ ok: false, method, ms, error: lastError });
+                if (err instanceof ChromeConnectError && err.hung)
+                    hungConnect = true;
+            }
+        }
+    }
+    return {
+        html: null,
+        error: `JavaScript rendering failed (${lastError}). ` +
+            "Chrome hung, timed out, or produced no content. Install Chrome, point WEBCRAWL_CHROME_PATH " +
+            "at chrome-headless-shell (`npx playwright install chromium-headless-shell`), " +
+            "run `webcrawl login` and retry with `--mode cloud`, or set WEBCRAWL_RENDER_URL.",
+        attempts,
+    };
+}
+async function dumpDom(chrome, url, headless, timeoutMs) {
     let profile = "";
     try {
         profile = mkdtempSync(join(tmpdir(), "wc-chrome-"));
     }
     catch {
-        return null;
+        throw new Error("could not create Chrome profile directory");
     }
-    const args = [
-        "--headless=new",
-        "--no-sandbox",
-        "--disable-gpu",
-        "--disable-dev-shm-usage",
-        "--no-first-run",
-        "--no-default-browser-check",
-        "--disable-crash-reporter",
-        "--disable-breakpad",
-        `--user-data-dir=${profile}`,
-        `--virtual-time-budget=${VIRTUAL_TIME_MS}`,
-        `--timeout=${timeoutMs}`,
-        "--dump-dom",
+    const args = chromeDumpDomArgs({
+        profile,
+        headless,
         url,
-    ];
+        timeoutMs,
+        binaryPath: chrome,
+    });
     try {
-        return await new Promise((resolve) => {
+        return await new Promise((resolve, reject) => {
             let settled = false;
-            const done = (value) => {
+            const done = (err, value) => {
                 if (settled)
                     return;
                 settled = true;
-                resolve(value);
+                if (err)
+                    reject(err);
+                else
+                    resolve(value ?? "");
             };
             let child;
             try {
-                child = spawn(chrome, args, { stdio: ["ignore", "pipe", "pipe"] });
+                child = spawn(chrome, args, {
+                    stdio: ["ignore", "pipe", "pipe"],
+                    detached: true,
+                });
             }
-            catch {
-                done(null);
+            catch (err) {
+                done(err instanceof Error ? err : new Error(String(err)));
                 return;
             }
             const chunks = [];
             let bytes = 0;
-            const timer = setTimeout(() => {
-                try {
-                    child.kill("SIGKILL");
-                }
-                catch {
-                    /* already gone */
+            let idle = null;
+            const finish = (err) => {
+                clearTimeout(timer);
+                if (idle)
+                    clearTimeout(idle);
+                killProcessGroup(child.pid, child);
+                if (err) {
+                    done(err);
+                    return;
                 }
-                done(null);
-            }, timeoutMs + 2_000);
+                const text = Buffer.concat(chunks).toString("utf8");
+                done(null, text.trim());
+            };
+            const timer = setTimeout(() => {
+                if (bytes > 100)
+                    finish(null);
+                else
+                    finish(new Error(`dump-dom timed out after ${timeoutMs}ms`));
+            }, timeoutMs);
             child.stdout?.on("data", (c) => {
                 if (bytes >= DUMP_MAX_BYTES)
                     return;
                 const take = c.subarray(0, DUMP_MAX_BYTES - bytes);
                 chunks.push(take);
                 bytes += take.length;
+                // Chrome 156 often prints the DOM then never exits (`--dump-dom` hang).
+                if (idle)
+                    clearTimeout(idle);
+                idle = setTimeout(() => {
+                    if (bytes > 40)
+                        finish(null);
+                }, 400);
             });
-            child.on("error", () => {
+            child.on("error", (err) => {
                 clearTimeout(timer);
-                done(null);
+                if (idle)
+                    clearTimeout(idle);
+                done(err);
             });
             child.on("close", () => {
                 clearTimeout(timer);
+                if (idle)
+                    clearTimeout(idle);
                 if (bytes === 0) {
-                    done(null);
+                    done(new Error("dump-dom produced no stdout"));
                     return;
                 }
                 const text = Buffer.concat(chunks).toString("utf8");
-                done(text.trim().length > 0 ? text : null);
+                done(null, text.trim().length > 0 ? text : "");
             });
         });
     }

package/dist/core/extract/index.d.ts

@@ -6,4 +6,4 @@ export { dedupeHotelListings, extractHotelListingFromSerpResult, extractHotelLis
 export { dedupeBusinessListings, extractBusinessListingFromSerpResult, extractBusinessListingsFromHtml, type BusinessListing, } from "./business.js";
 export { cvss3BaseScore, cveSeverityLabel, extractCveIdsFromText, extractGhsaIdsFromText, looksLikeVulnId, mergeCveRecords, normalizeCveId, nvdUrl, parseNvdCve, parseNvdResponse, parseOsvVuln, severityFromScore, sortCveRecords, type CveAffected, type CveRecord, type CveSeverity, } from "./cve.js";
 export { extractBusinessListingsWithScaleDown, extractFlightOffersWithScaleDown, extractHotelListingsWithScaleDown, extractPropertyListingsWithScaleDown, } from "./scaledown.js";
-export { extractByPromptAndSchema, extractQuoteAuthorPairs, hasCustomExtractRequest, shapeQuoteAuthors, wantsQuoteAuthorExtract, type QuoteAuthor, } from "./structured.js";
+export { coerceExtractSchema, extractByPromptAndSchema, extractQuoteAuthorPairs, hasCustomExtractRequest, isEmptyExtractResult, isJsTemplateJunk, nestedQuoteAuthorFields, normalizeExtractSchema, shapeQuoteAuthors, stripNonContentHtml, wantsQuoteAuthorExtract, type QuoteAuthor, } from "./structured.js";

package/dist/core/extract/index.js

@@ -6,4 +6,4 @@ export { dedupeHotelListings, extractHotelListingFromSerpResult, extractHotelLis
 export { dedupeBusinessListings, extractBusinessListingFromSerpResult, extractBusinessListingsFromHtml, } from "./business.js";
 export { cvss3BaseScore, cveSeverityLabel, extractCveIdsFromText, extractGhsaIdsFromText, looksLikeVulnId, mergeCveRecords, normalizeCveId, nvdUrl, parseNvdCve, parseNvdResponse, parseOsvVuln, severityFromScore, sortCveRecords, } from "./cve.js";
 export { extractBusinessListingsWithScaleDown, extractFlightOffersWithScaleDown, extractHotelListingsWithScaleDown, extractPropertyListingsWithScaleDown, } from "./scaledown.js";
-export { extractByPromptAndSchema, extractQuoteAuthorPairs, hasCustomExtractRequest, shapeQuoteAuthors, wantsQuoteAuthorExtract, } from "./structured.js";
+export { coerceExtractSchema, extractByPromptAndSchema, extractQuoteAuthorPairs, hasCustomExtractRequest, isEmptyExtractResult, isJsTemplateJunk, nestedQuoteAuthorFields, normalizeExtractSchema, shapeQuoteAuthors, stripNonContentHtml, wantsQuoteAuthorExtract, } from "./structured.js";

package/dist/core/extract/structured.d.ts

@@ -2,8 +2,27 @@ export interface QuoteAuthor {
     text: string;
     author: string;
 }
+/** A bare property map or shorthand array becomes a JSON Schema. */
+export declare function coerceExtractSchema(node: unknown): Record<string, unknown>;
+/**
+ * Root-array wrap: a top-level array becomes
+ * `{ type: object, properties: { items: <array> } }`.
+ */
+export declare function normalizeExtractSchema(schema?: Record<string, unknown>): Record<string, unknown> | undefined;
+export declare function schemaProperties(schema?: Record<string, unknown>): Record<string, unknown>;
 export declare function hasCustomExtractRequest(prompt?: string, schema?: Record<string, unknown>): boolean;
+/** Nested object/array field names that look like quote+author records. */
+export declare function nestedQuoteAuthorFields(schema?: Record<string, unknown>): {
+    arrayPath: string[];
+    textKey: string;
+    authorKey: string;
+} | null;
 export declare function wantsQuoteAuthorExtract(prompt?: string, schema?: Record<string, unknown>): boolean;
+/** Strip script/style/template (including unclosed tags) before reading visible DOM. */
+export declare function stripNonContentHtml(html: string): string;
+/** Inline `document.write("... " + d['text'] + ...)` leftovers, not real quotes. */
+export declare function isJsTemplateJunk(text: string): boolean;
+export declare function isEmptyExtractResult(data: Record<string, unknown> | null | undefined): boolean;
 /** Pull quote/author pairs from HTML (cards, microdata, `var data = [...]`) and markdown. */
 export declare function extractQuoteAuthorPairs(html: string, markdown?: string): QuoteAuthor[];
 export declare function shapeQuoteAuthors(pairs: QuoteAuthor[], schema?: Record<string, unknown>): Record<string, unknown>;