import WebSocket from "ws"; import { rawDataToString } from "../infra/ws.js"; type CdpResponse = { id: number; result?: unknown; error?: { message?: string }; }; type Pending = { resolve: (value: unknown) => void; reject: (err: Error) => void; }; type CdpSendFn = ( method: string, params?: Record, ) => Promise; function createCdpSender(ws: WebSocket) { let nextId = 1; const pending = new Map(); const send: CdpSendFn = ( method: string, params?: Record, ) => { const id = nextId++; const msg = { id, method, params }; ws.send(JSON.stringify(msg)); return new Promise((resolve, reject) => { pending.set(id, { resolve, reject }); }); }; const closeWithError = (err: Error) => { for (const [, p] of pending) p.reject(err); pending.clear(); try { ws.close(); } catch { // ignore } }; ws.on("message", (data) => { try { const parsed = JSON.parse(rawDataToString(data)) as CdpResponse; if (typeof parsed.id !== "number") return; const p = pending.get(parsed.id); if (!p) return; pending.delete(parsed.id); if (parsed.error?.message) { p.reject(new Error(parsed.error.message)); return; } p.resolve(parsed.result); } catch { // ignore } }); ws.on("close", () => { closeWithError(new Error("CDP socket closed")); }); return { send, closeWithError }; } async function fetchJson(url: string, timeoutMs = 1500): Promise { const ctrl = new AbortController(); const t = setTimeout(() => ctrl.abort(), timeoutMs); try { const res = await fetch(url, { signal: ctrl.signal }); if (!res.ok) throw new Error(`HTTP ${res.status}`); return (await res.json()) as T; } finally { clearTimeout(t); } } async function withCdpSocket( wsUrl: string, fn: (send: CdpSendFn) => Promise, ): Promise { const ws = new WebSocket(wsUrl, { handshakeTimeout: 5000 }); const { send, closeWithError } = createCdpSender(ws); const openPromise = new Promise((resolve, reject) => { ws.once("open", () => resolve()); ws.once("error", (err) => reject(err)); }); await openPromise; try { return await fn(send); } catch (err) { closeWithError(err instanceof Error ? err : new Error(String(err))); throw err; } finally { try { ws.close(); } catch { // ignore } } } export async function captureScreenshotPng(opts: { wsUrl: string; fullPage?: boolean; }): Promise { return await captureScreenshot({ wsUrl: opts.wsUrl, fullPage: opts.fullPage, format: "png", }); } export async function captureScreenshot(opts: { wsUrl: string; fullPage?: boolean; format?: "png" | "jpeg"; quality?: number; // jpeg only (0..100) }): Promise { return await withCdpSocket(opts.wsUrl, async (send) => { await send("Page.enable"); let clip: | { x: number; y: number; width: number; height: number; scale: number } | undefined; if (opts.fullPage) { const metrics = (await send("Page.getLayoutMetrics")) as { cssContentSize?: { width?: number; height?: number }; contentSize?: { width?: number; height?: number }; }; const size = metrics?.cssContentSize ?? metrics?.contentSize; const width = Number(size?.width ?? 0); const height = Number(size?.height ?? 0); if (width > 0 && height > 0) { clip = { x: 0, y: 0, width, height, scale: 1 }; } } const format = opts.format ?? "png"; const quality = format === "jpeg" ? Math.max(0, Math.min(100, Math.round(opts.quality ?? 85))) : undefined; const result = (await send("Page.captureScreenshot", { format, ...(quality !== undefined ? { quality } : {}), fromSurface: true, captureBeyondViewport: true, ...(clip ? { clip } : {}), })) as { data?: string }; const base64 = result?.data; if (!base64) throw new Error("Screenshot failed: missing data"); return Buffer.from(base64, "base64"); }); } export async function createTargetViaCdp(opts: { cdpPort: number; url: string; }): Promise<{ targetId: string }> { const version = await fetchJson<{ webSocketDebuggerUrl?: string }>( `http://127.0.0.1:${opts.cdpPort}/json/version`, 1500, ); const wsUrl = String(version?.webSocketDebuggerUrl ?? "").trim(); if (!wsUrl) throw new Error("CDP /json/version missing webSocketDebuggerUrl"); return await withCdpSocket(wsUrl, async (send) => { const created = (await send("Target.createTarget", { url: opts.url })) as { targetId?: string; }; const targetId = String(created?.targetId ?? "").trim(); if (!targetId) throw new Error("CDP Target.createTarget returned no targetId"); return { targetId }; }); } export type CdpRemoteObject = { type: string; subtype?: string; value?: unknown; description?: string; unserializableValue?: string; preview?: unknown; }; export type CdpExceptionDetails = { text?: string; lineNumber?: number; columnNumber?: number; exception?: CdpRemoteObject; stackTrace?: unknown; }; export async function evaluateJavaScript(opts: { wsUrl: string; expression: string; awaitPromise?: boolean; returnByValue?: boolean; }): Promise<{ result: CdpRemoteObject; exceptionDetails?: CdpExceptionDetails; }> { return await withCdpSocket(opts.wsUrl, async (send) => { await send("Runtime.enable").catch(() => {}); const evaluated = (await send("Runtime.evaluate", { expression: opts.expression, awaitPromise: Boolean(opts.awaitPromise), returnByValue: opts.returnByValue ?? true, userGesture: true, includeCommandLineAPI: true, })) as { result?: CdpRemoteObject; exceptionDetails?: CdpExceptionDetails; }; const result = evaluated?.result; if (!result) throw new Error("CDP Runtime.evaluate returned no result"); return { result, exceptionDetails: evaluated.exceptionDetails }; }); } export type AriaSnapshotNode = { ref: string; role: string; name: string; value?: string; description?: string; backendDOMNodeId?: number; depth: number; }; type RawAXNode = { nodeId?: string; role?: { value?: string }; name?: { value?: string }; value?: { value?: string }; description?: { value?: string }; childIds?: string[]; backendDOMNodeId?: number; }; function axValue(v: unknown): string { if (!v || typeof v !== "object") return ""; const value = (v as { value?: unknown }).value; if (typeof value === "string") return value; if (typeof value === "number" || typeof value === "boolean") { return String(value); } return ""; } function formatAriaSnapshot( nodes: RawAXNode[], limit: number, ): AriaSnapshotNode[] { const byId = new Map(); for (const n of nodes) { if (n.nodeId) byId.set(n.nodeId, n); } // Heuristic: pick a root-ish node (one that is not referenced as a child), else first. const referenced = new Set(); for (const n of nodes) { for (const c of n.childIds ?? []) referenced.add(c); } const root = nodes.find((n) => n.nodeId && !referenced.has(n.nodeId)) ?? nodes[0]; if (!root?.nodeId) return []; const out: AriaSnapshotNode[] = []; const stack: Array<{ id: string; depth: number }> = [ { id: root.nodeId, depth: 0 }, ]; while (stack.length && out.length < limit) { const popped = stack.pop(); if (!popped) break; const { id, depth } = popped; const n = byId.get(id); if (!n) continue; const role = axValue(n.role); const name = axValue(n.name); const value = axValue(n.value); const description = axValue(n.description); const ref = `ax${out.length + 1}`; out.push({ ref, role: role || "unknown", name: name || "", ...(value ? { value } : {}), ...(description ? { description } : {}), ...(typeof n.backendDOMNodeId === "number" ? { backendDOMNodeId: n.backendDOMNodeId } : {}), depth, }); const children = (n.childIds ?? []).filter((c) => byId.has(c)); for (let i = children.length - 1; i >= 0; i--) { const child = children[i]; if (child) stack.push({ id: child, depth: depth + 1 }); } } return out; } export async function snapshotAria(opts: { wsUrl: string; limit?: number; }): Promise<{ nodes: AriaSnapshotNode[] }> { const limit = Math.max(1, Math.min(2000, Math.floor(opts.limit ?? 500))); return await withCdpSocket(opts.wsUrl, async (send) => { await send("Accessibility.enable").catch(() => {}); const res = (await send("Accessibility.getFullAXTree")) as { nodes?: RawAXNode[]; }; const nodes = Array.isArray(res?.nodes) ? res.nodes : []; return { nodes: formatAriaSnapshot(nodes, limit) }; }); } export async function snapshotDom(opts: { wsUrl: string; limit?: number; maxTextChars?: number; }): Promise<{ nodes: DomSnapshotNode[]; }> { const limit = Math.max(1, Math.min(5000, Math.floor(opts.limit ?? 800))); const maxTextChars = Math.max( 0, Math.min(5000, Math.floor(opts.maxTextChars ?? 220)), ); const expression = `(() => { const maxNodes = ${JSON.stringify(limit)}; const maxText = ${JSON.stringify(maxTextChars)}; const nodes = []; const root = document.documentElement; if (!root) return { nodes }; const stack = [{ el: root, depth: 0, parentRef: null }]; while (stack.length && nodes.length < maxNodes) { const cur = stack.pop(); const el = cur.el; if (!el || el.nodeType !== 1) continue; const ref = "n" + String(nodes.length + 1); const tag = (el.tagName || "").toLowerCase(); const id = el.id ? String(el.id) : undefined; const className = el.className ? String(el.className).slice(0, 300) : undefined; const role = el.getAttribute && el.getAttribute("role") ? String(el.getAttribute("role")) : undefined; const name = el.getAttribute && el.getAttribute("aria-label") ? String(el.getAttribute("aria-label")) : undefined; let text = ""; try { text = String(el.innerText || "").trim(); } catch {} if (maxText && text.length > maxText) text = text.slice(0, maxText) + "…"; const href = (el.href !== undefined && el.href !== null) ? String(el.href) : undefined; const type = (el.type !== undefined && el.type !== null) ? String(el.type) : undefined; const value = (el.value !== undefined && el.value !== null) ? String(el.value).slice(0, 500) : undefined; nodes.push({ ref, parentRef: cur.parentRef, depth: cur.depth, tag, ...(id ? { id } : {}), ...(className ? { className } : {}), ...(role ? { role } : {}), ...(name ? { name } : {}), ...(text ? { text } : {}), ...(href ? { href } : {}), ...(type ? { type } : {}), ...(value ? { value } : {}), }); const children = el.children ? Array.from(el.children) : []; for (let i = children.length - 1; i >= 0; i--) { stack.push({ el: children[i], depth: cur.depth + 1, parentRef: ref }); } } return { nodes }; })()`; const evaluated = await evaluateJavaScript({ wsUrl: opts.wsUrl, expression, awaitPromise: true, returnByValue: true, }); const value = evaluated.result?.value as unknown; if (!value || typeof value !== "object") return { nodes: [] }; const nodes = (value as { nodes?: unknown }).nodes; return { nodes: Array.isArray(nodes) ? (nodes as DomSnapshotNode[]) : [] }; } export type DomSnapshotNode = { ref: string; parentRef: string | null; depth: number; tag: string; id?: string; className?: string; role?: string; name?: string; text?: string; href?: string; type?: string; value?: string; }; export async function getDomText(opts: { wsUrl: string; format: "html" | "text"; maxChars?: number; selector?: string; }): Promise<{ text: string }> { const maxChars = Math.max( 0, Math.min(5_000_000, Math.floor(opts.maxChars ?? 200_000)), ); const selectorExpr = opts.selector ? JSON.stringify(opts.selector) : "null"; const expression = `(() => { const fmt = ${JSON.stringify(opts.format)}; const max = ${JSON.stringify(maxChars)}; const sel = ${selectorExpr}; const pick = sel ? document.querySelector(sel) : null; let out = ""; if (fmt === "text") { const el = pick || document.body || document.documentElement; try { out = String(el && el.innerText ? el.innerText : ""); } catch { out = ""; } } else { const el = pick || document.documentElement; try { out = String(el && el.outerHTML ? el.outerHTML : ""); } catch { out = ""; } } if (max && out.length > max) out = out.slice(0, max) + "\\n"; return out; })()`; const evaluated = await evaluateJavaScript({ wsUrl: opts.wsUrl, expression, awaitPromise: true, returnByValue: true, }); const textValue = (evaluated.result?.value ?? "") as unknown; const text = typeof textValue === "string" ? textValue : typeof textValue === "number" || typeof textValue === "boolean" ? String(textValue) : ""; return { text }; } export async function querySelector(opts: { wsUrl: string; selector: string; limit?: number; maxTextChars?: number; maxHtmlChars?: number; }): Promise<{ matches: QueryMatch[]; }> { const limit = Math.max(1, Math.min(200, Math.floor(opts.limit ?? 20))); const maxText = Math.max( 0, Math.min(5000, Math.floor(opts.maxTextChars ?? 500)), ); const maxHtml = Math.max( 0, Math.min(20000, Math.floor(opts.maxHtmlChars ?? 1500)), ); const expression = `(() => { const sel = ${JSON.stringify(opts.selector)}; const lim = ${JSON.stringify(limit)}; const maxText = ${JSON.stringify(maxText)}; const maxHtml = ${JSON.stringify(maxHtml)}; const els = Array.from(document.querySelectorAll(sel)).slice(0, lim); return els.map((el, i) => { const tag = (el.tagName || "").toLowerCase(); const id = el.id ? String(el.id) : undefined; const className = el.className ? String(el.className).slice(0, 300) : undefined; let text = ""; try { text = String(el.innerText || "").trim(); } catch {} if (maxText && text.length > maxText) text = text.slice(0, maxText) + "…"; const value = (el.value !== undefined && el.value !== null) ? String(el.value).slice(0, 500) : undefined; const href = (el.href !== undefined && el.href !== null) ? String(el.href) : undefined; let outerHTML = ""; try { outerHTML = String(el.outerHTML || ""); } catch {} if (maxHtml && outerHTML.length > maxHtml) outerHTML = outerHTML.slice(0, maxHtml) + "…"; return { index: i + 1, tag, ...(id ? { id } : {}), ...(className ? { className } : {}), ...(text ? { text } : {}), ...(value ? { value } : {}), ...(href ? { href } : {}), ...(outerHTML ? { outerHTML } : {}), }; }); })()`; const evaluated = await evaluateJavaScript({ wsUrl: opts.wsUrl, expression, awaitPromise: true, returnByValue: true, }); const matches = evaluated.result?.value; return { matches: Array.isArray(matches) ? (matches as QueryMatch[]) : [] }; } export type QueryMatch = { index: number; tag: string; id?: string; className?: string; text?: string; value?: string; href?: string; outerHTML?: string; };