fix: canvas image/video download - fix data URL parsing in cvDownloadPreview (v1.5.21, hot 1.5.23)

This commit is contained in:
mac
2026-07-20 08:48:56 +08:00
parent 42e23663ee
commit aadd2b04ae
28 changed files with 5881 additions and 329 deletions
+372 -2
View File
@@ -16,10 +16,14 @@ function dlog(...a) { try { console.log("[voice]", ...a); } catch {} }
const router = require("./router");
const providerSync = require("./provider_sync");
const { INJECT_SCRIPT } = require("./voice_inject");
const { TOKEN_INJECT_SCRIPT } = require("./token_inject");
const tokenUsage = require("./token_usage");
const DEFAULT_PORT = 19222; // Codex 调试端口(避开 Chrome 默认 9222)
const BINDING_NAME = "__voiceBridge"; // 页面 → 主进程 的 CDP binding
const CDP_HTTP_TIMEOUT = 3000; // GET /json 超时
const CDP_WS_TIMEOUT = 5000; // WebSocket 握手超时
const CDP_CMD_TIMEOUT = 5000; // 单条 CDP 命令超时
const READY_POLL_INTERVAL = 500; // 轮询就绪间隔
const READY_POLL_MAX_MS = 20000; // 轮询就绪最长等待
const RECORD_MAX_MS = 60000; // 单次录音上限(注入脚本侧也限)
@@ -35,7 +39,10 @@ let state = {
target: null, // { id, url, title }
};
let session = null; // 当前 CdpSession
let networkSessions = []; // worker 等附加 target 的 Network 监听 session
let networkAttachGeneration = 0;
let onStatusCb = null;
let tokenSyncTimer = null;
function snapshot() {
const { phase, port, codexPath, lastError, model, target } = state;
@@ -50,6 +57,29 @@ function setPhase(phase, extra = {}) {
}
}
async function syncTokenUsage() {
if (!session) return;
const usage = tokenUsage.readDaily();
if (!usage || !usage.available) return;
try {
await session.send("Runtime.evaluate", {
expression: "window.__tdTokenSet && window.__tdTokenSet(" + JSON.stringify(usage) + ")",
awaitPromise: false,
});
} catch {}
}
function startTokenSync() {
if (tokenSyncTimer) clearInterval(tokenSyncTimer);
tokenSyncTimer = setInterval(() => { syncTokenUsage().catch(() => {}); }, 2500);
syncTokenUsage().catch(() => {});
}
function stopTokenSync() {
if (tokenSyncTimer) clearInterval(tokenSyncTimer);
tokenSyncTimer = null;
}
// —— 探测 Codex 可执行路径 ——
function codexExecutable() {
const home = os.homedir();
@@ -139,6 +169,10 @@ function wsConnect(wsUrl) {
const socket = net.connect({ host, port: Number(port) });
const events = { open: [], message: [], close: [], error: [] };
const emit = (name, arg) => (events[name] || []).forEach((cb) => cb(arg));
socket.setTimeout(CDP_WS_TIMEOUT, () => {
if (!handshakeDone) emit("error", new Error("WS 连接超时"));
try { socket.destroy(); } catch {}
});
// —— 发送一帧文本(客户端必须掩码)——
function sendFrame(str) {
@@ -252,6 +286,7 @@ function wsConnect(wsUrl) {
return;
}
handshakeDone = true;
try { socket.setTimeout(0); } catch {}
emit("open");
parseFrames();
} else {
@@ -287,7 +322,20 @@ function openSession(wsUrl, onBindingCalled) {
return new Promise((resolve, reject) => {
if (closed) return reject(new Error("CDP 会话已关闭"));
const id = nextId++;
pending.set(id, { resolve, reject });
const timer = setTimeout(() => {
pending.delete(id);
reject(new Error("CDP 命令超时: " + method));
}, CDP_CMD_TIMEOUT);
pending.set(id, {
resolve: (value) => {
clearTimeout(timer);
resolve(value);
},
reject: (err) => {
clearTimeout(timer);
reject(err);
},
});
ws.send(JSON.stringify({ id, method, params }));
});
}
@@ -406,7 +454,7 @@ function connectedPhase() {
return session ? "connected" : "disconnected";
}
// —— 注入:addBinding + addScriptToEvaluateOnNewDocument + 立即 evaluate ——
// —— 注入:addBinding + addScriptToEvaluateOnNewDocument + 立即 evaluate + Network 监听 ——
async function inject(sessionObj) {
await sessionObj.send("Runtime.enable");
try {
@@ -414,8 +462,266 @@ async function inject(sessionObj) {
} catch {}
await sessionObj.send("Runtime.addBinding", { name: BINDING_NAME });
await sessionObj.send("Page.enable");
// 语音注入脚本
await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: INJECT_SCRIPT });
await sessionObj.send("Runtime.evaluate", { expression: INJECT_SCRIPT, awaitPromise: false });
// Token 用量 UI 注入脚本(只负责 UI + localStorage,不管网络抓取)
await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: TOKEN_INJECT_SCRIPT });
await sessionObj.send("Runtime.evaluate", { expression: TOKEN_INJECT_SCRIPT, awaitPromise: false });
// 启用 Network 域,在 CDP 层面拦截 LLM 响应抓 token 用量
await sessionObj.send("Network.enable");
}
// ============================================================================
// Token 用量:CDP Network 域拦截 LLM 响应抓 usage(不受页面 CSP 限制)
// ============================================================================
// 待完成的请求:requestId → { url, mimeType }
const pendingNetworkReqs = new Map();
let nextNetworkSessionId = 1;
const tokenNet = {
matchedCount: 0,
usageCount: 0,
noUsageCount: 0,
lastMatchedUrl: "",
lastUsageUrl: "",
lastNoUsageUrl: "",
lastUsageTotal: 0,
lastError: "",
attachedWorkers: 0,
};
const LLM_URL_RE = /\/(chat\/completions|v1\/messages|messages|generateContent|streamGenerateContent|completions|responses|backend-api\/(?:codex|conversation|responses)|codex|conversation)(\/|$|\?)/i;
function networkReqKey(ownerSession, requestId) {
if (!ownerSession._tdTokenNetId) ownerSession._tdTokenNetId = nextNetworkSessionId++;
return ownerSession._tdTokenNetId + ":" + requestId;
}
function bindNetworkHandlers(ownerSession) {
if (!ownerSession) return;
if (!ownerSession._tdTokenNetId) ownerSession._tdTokenNetId = nextNetworkSessionId++;
ownerSession.on("Network.responseReceived", (params) => handleNetworkResponse(params, ownerSession));
ownerSession.on("Network.loadingFinished", (params) => handleNetworkLoadingFinished(params, ownerSession));
ownerSession.on("Network.loadingFailed", (params) => handleNetworkLoadingFailed(params, ownerSession));
}
function handleNetworkResponse(params, ownerSession) {
const { requestId, response } = params || {};
if (!requestId || !response || !ownerSession) return;
const url = response.url || "";
if (!LLM_URL_RE.test(url)) return;
// 只处理 JSON 响应
const mime = (response.mimeType || "").toLowerCase();
if (!mime.includes("json") && !mime.includes("text")) return;
tokenNet.matchedCount += 1;
tokenNet.lastMatchedUrl = url;
pendingNetworkReqs.set(networkReqKey(ownerSession, requestId), { url, mimeType: mime });
}
function handleNetworkLoadingFailed(params, ownerSession) {
const { requestId } = params || {};
if (requestId && ownerSession) pendingNetworkReqs.delete(networkReqKey(ownerSession, requestId));
}
async function handleNetworkLoadingFinished(params, ownerSession) {
const { requestId } = params || {};
if (!requestId || !ownerSession) return;
const key = networkReqKey(ownerSession, requestId);
if (!pendingNetworkReqs.has(key)) return;
const reqInfo = pendingNetworkReqs.get(key) || {};
pendingNetworkReqs.delete(key);
if (!session || !ownerSession) return;
try {
const r = await ownerSession.send("Network.getResponseBody", { requestId });
const body = (r && r.body) || "";
if (!body || body.length < 10) return;
// 可能被 base64 编码
let text = body;
if (r && r.base64Encoded) {
try { text = Buffer.from(body, "base64").toString("utf8"); } catch { return; }
}
// 解析 JSON 提取 usage
const usage = extractUsageFromText(text);
if (usage && usage.total > 0) {
tokenNet.usageCount += 1;
tokenNet.lastUsageUrl = reqInfo.url || "";
tokenNet.lastUsageTotal = usage.total || 0;
tokenNet.lastError = "";
dlog(
"token: +" + usage.total +
" input=" + (usage.input || 0) +
" output=" + (usage.output || 0) +
" cached=" + (usage.cached || 0) +
" from",
reqInfo.url || "?"
);
// 通过 Runtime.evaluate 调注入脚本更新 UI
try {
await session.send("Runtime.evaluate", {
expression: "window.__tdTokenAdd && window.__tdTokenAdd(" + JSON.stringify(usage) + ", 'network')",
awaitPromise: false,
});
} catch {}
} else {
tokenNet.noUsageCount += 1;
tokenNet.lastNoUsageUrl = reqInfo.url || "";
}
} catch (e) {
tokenNet.lastError = String((e && e.message) || e || "getResponseBody failed").slice(0, 160);
// getResponseBody 可能失败(响应体已被清掉或太大),静默跳过
}
}
function extractUsageFromText(text) {
let best = null;
const take = (obj) => {
best = betterUsage(best, extractUsageFromObject(obj));
};
try {
take(JSON.parse(text));
} catch {}
if (usageTotal(best) > 0) return best;
// Stream/SSE: data: {"usage":...}\n\n or data: [DONE]
const lines = String(text || "").split(/\r?\n/);
for (const line of lines) {
const s = line.trim();
const payload = s.startsWith("data:") ? s.slice(5).trim() : s;
if (!payload || payload === "[DONE]") continue;
try { take(JSON.parse(payload)); } catch {}
}
return best;
}
function cleanTokenNumber(n) {
n = Number(n);
return Number.isFinite(n) && n > 0 ? Math.round(n) : 0;
}
function usageTotal(u) {
return u ? cleanTokenNumber(u.total) : 0;
}
function betterUsage(a, b) {
return usageTotal(b) > usageTotal(a) ? b : a;
}
function hasOwn(obj, key) {
return Object.prototype.hasOwnProperty.call(obj, key);
}
function looksLikeUsage(usage) {
if (!usage || typeof usage !== "object") return false;
const keys = [
"input_tokens", "prompt_tokens", "output_tokens", "completion_tokens", "total_tokens",
"promptTokenCount", "prompt_token_count", "inputTokenCount", "outputTokenCount",
"candidatesTokenCount", "candidates_token_count", "totalTokenCount", "total_token_count",
"token_count", "cache_creation_input_tokens", "cache_read_input_tokens",
"cachedContentTokenCount", "cached_content_token_count", "reasoning_tokens",
"thoughtsTokenCount", "thoughts_token_count",
];
if (keys.some((key) => hasOwn(usage, key))) return true;
return hasOwn(usage, "total") && (hasOwn(usage, "input") || hasOwn(usage, "output") || hasOwn(usage, "cached"));
}
function extractUsageFromRawUsage(usage) {
if (!usage || typeof usage !== "object") return null;
if (!looksLikeUsage(usage)) return null;
const input = cleanTokenNumber(
usage.input_tokens ||
usage.prompt_tokens ||
usage.promptTokenCount ||
usage.prompt_token_count ||
usage.inputTokenCount
);
const output = cleanTokenNumber(
usage.output_tokens ||
usage.completion_tokens ||
usage.candidatesTokenCount ||
usage.candidates_token_count ||
usage.outputTokenCount
);
const reasoning = cleanTokenNumber(usage.reasoning_tokens || usage.thoughtsTokenCount || usage.thoughts_token_count);
const details = usage.input_tokens_details || usage.prompt_tokens_details || usage.inputTokenDetails || usage.promptTokenDetails || {};
const cached =
cleanTokenNumber(details.cached_tokens) +
cleanTokenNumber(usage.cache_creation_input_tokens) +
cleanTokenNumber(usage.cache_read_input_tokens) +
cleanTokenNumber(usage.cachedContentTokenCount) +
cleanTokenNumber(usage.cached_content_token_count);
const total = cleanTokenNumber(
usage.total_tokens ||
usage.totalTokenCount ||
usage.total_token_count ||
usage.token_count ||
usage.total
) || input + output + reasoning + cached;
if (!total) return null;
return { total, input, output: output + reasoning, cached };
}
function extractUsageFromObject(obj) {
if (!obj || typeof obj !== "object") return null;
const direct =
extractUsageFromRawUsage(obj.usage) ||
extractUsageFromRawUsage(obj.usage_metadata) ||
extractUsageFromRawUsage(obj.usageMetadata) ||
extractUsageFromRawUsage(obj);
if (direct) return direct;
let best = null;
for (const val of Object.values(obj)) {
if (!val || typeof val !== "object") continue;
if (Array.isArray(val)) {
for (const item of val) best = betterUsage(best, extractUsageFromObject(item));
} else {
best = betterUsage(best, extractUsageFromObject(val));
}
}
return best;
}
async function attachNetworkSession(target, generation) {
if (!target || !target.webSocketDebuggerUrl) return;
let netSession = null;
try {
netSession = await openSession(target.webSocketDebuggerUrl, null);
bindNetworkHandlers(netSession);
await netSession.send("Network.enable");
if (!session || generation !== networkAttachGeneration) {
try { netSession.close(); } catch {}
return false;
}
networkSessions.push(netSession);
dlog("token network attached:", target.type || "target", target.id || "", target.url || "");
return true;
} catch (e) {
if (netSession) {
try { netSession.close(); } catch {}
}
tokenNet.lastError = String((e && e.message) || e || "attach network failed").slice(0, 160);
return false;
}
}
async function attachWorkerNetworkTargets(port, parentId) {
const targets = await listTargets(port);
const workers = targets.filter((t) =>
t &&
t.webSocketDebuggerUrl &&
(t.type === "worker" || t.type === "service_worker" || t.type === "shared_worker") &&
(!parentId || t.parentId === parentId)
);
const generation = networkAttachGeneration;
const results = await Promise.all(workers.map((worker) => attachNetworkSession(worker, generation)));
tokenNet.attachedWorkers = results.filter(Boolean).length;
}
function closeNetworkSessions() {
networkAttachGeneration += 1;
for (const s of networkSessions) {
try { s.close(); } catch {}
}
networkSessions = [];
pendingNetworkReqs.clear();
}
// —— 连接到已带端口的 Codex ——
@@ -423,17 +729,26 @@ async function connectToTarget(port) {
const targets = await listTargets(port);
const tgt = pickTarget(targets);
if (!tgt) return false;
closeNetworkSessions();
state.target = { id: tgt.id, url: tgt.url, title: tgt.title };
setPhase("connecting");
session = await openSession(tgt.webSocketDebuggerUrl, handleBinding);
session.on("Runtime.bindingCalled", handleBinding);
// Token 用量:Network 层拦截 LLM 响应
bindNetworkHandlers(session);
session.ws.on("close", () => {
stopTokenSync();
closeNetworkSessions();
session = null;
setPhase("disconnected");
});
setPhase("injecting");
await inject(session);
startTokenSync();
setPhase("connected", { target: state.target });
attachWorkerNetworkTargets(port, tgt.id).catch((e) => {
tokenNet.lastError = String((e && e.message) || e || "attach workers failed").slice(0, 160);
});
return true;
}
@@ -545,6 +860,8 @@ async function connect(cb, opts = {}) {
}
function disconnect() {
stopTokenSync();
closeNetworkSessions();
if (session) {
try {
session.close();
@@ -601,6 +918,56 @@ async function probeStt() {
return out;
}
// —— Token 用量读取 / 重置(通过 CDP Runtime.evaluate 调注入脚本的 window API)——
async function tokenStatus() {
if (!session) return { date: "", total: 0 };
try {
const r = await session.send("Runtime.evaluate", {
expression: "window.__tdTokenGet ? window.__tdTokenGet() : JSON.stringify({date:'',total:0})",
awaitPromise: false,
});
const raw = (r && r.result && r.result.value) || "";
const parsed = (() => { try { return JSON.parse(raw); } catch { return { date: "", total: 0 }; } })();
return {
date: parsed.date || "",
total: Number(parsed.total) || 0,
input: Number(parsed.input) || 0,
output: Number(parsed.output) || 0,
cached: Number(parsed.cached) || 0,
estimate: Number(parsed.estimate || parsed.estimated) || 0,
unknown: Number(parsed.unknown) || 0,
lastSource: parsed.lastSource || parsed.source || "",
network: { ...tokenNet },
};
} catch {
return { date: "", total: 0 };
}
}
async function tokenSet(usage) {
if (!session) return false;
try {
await session.send("Runtime.evaluate", {
expression: "window.__tdTokenSet && window.__tdTokenSet(" + JSON.stringify(usage || {}) + ")",
awaitPromise: false,
});
return true;
} catch {
return false;
}
}
async function tokenReset() {
if (!session) return false;
try {
await session.send("Runtime.evaluate", {
expression: "window.__tdTokenReset ? window.__tdTokenReset() : 'no'",
awaitPromise: false,
});
return true;
} catch {
return false;
}
}
module.exports = {
status,
connect,
@@ -608,6 +975,9 @@ module.exports = {
onStatus,
sttInfo,
probeStt,
tokenStatus,
tokenSet,
tokenReset,
// 供临时实测脚本复用
_listTargets: listTargets,
_pickTarget: pickTarget,