feat: enhance session sync and add voice input feature

- 增强会话同步:确保所有会话完全同步到新账号
  - 强制更新 model_provider 字段
  - 添加 session_index.jsonl 重建功能
  - 确保数据库 UPSERT 逻辑正确处理所有情况
- 新增语音输入功能(voice_cdp.js, voice_inject.js)
- 测试脚本和增强模块

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
mac
2026-07-16 19:16:45 +08:00
parent a6207fdefd
commit 02fb553c9e
19 changed files with 1884 additions and 95 deletions
+184 -10
View File
@@ -2,6 +2,12 @@
// 统一在主进程发起 image/video/tts 请求并把成品下载转 base64 data URL 返回渲染层。
// key 只在主进程流转;渲染层被 CSP 拦截的远程 URL 一律由这里下载转 base64。
const router = require("./router");
const fs = require("fs");
const http = require("http");
const https = require("https");
const os = require("os");
const path = require("path");
const { execFileSync } = require("child_process");
// —— 从 baseUrl 解析出主机名(失败则原样返回)——
function hostOf(baseUrl) {
@@ -31,12 +37,116 @@ function headers(key) {
};
}
function grokDebugLog(entry) {
try {
const dir = path.join(os.homedir(), "Library", "Application Support", "codex-market");
fs.mkdirSync(dir, { recursive: true });
fs.appendFileSync(
path.join(dir, "grok-video-debug.jsonl"),
JSON.stringify({ time: new Date().toISOString(), ...entry }) + "\n"
);
} catch {}
}
function systemProxyUrl() {
const envProxy = process.env.HTTPS_PROXY || process.env.https_proxy || process.env.HTTP_PROXY || process.env.http_proxy || "";
if (envProxy) return envProxy;
if (process.platform !== "darwin") return "";
try {
const out = execFileSync("scutil", ["--proxy"], { encoding: "utf8", timeout: 1000 });
const get = (name) => {
const m = out.match(new RegExp("^\\s*" + name + "\\s*:\\s*(.+)$", "m"));
return m ? m[1].trim() : "";
};
const httpsOn = get("HTTPSEnable") === "1";
const httpOn = get("HTTPEnable") === "1";
const host = httpsOn ? get("HTTPSProxy") : httpOn ? get("HTTPProxy") : "";
const port = httpsOn ? get("HTTPSPort") : httpOn ? get("HTTPPort") : "";
return host && port ? `http://${host}:${port}` : "";
} catch {
return "";
}
}
function proxyAgentFor(targetUrl, proxyUrl) {
try {
if (/^https:/i.test(targetUrl)) {
const { HttpsProxyAgent } = require("https-proxy-agent");
return new HttpsProxyAgent(proxyUrl);
}
const { HttpProxyAgent } = require("http-proxy-agent");
return new HttpProxyAgent(proxyUrl);
} catch {
return null;
}
}
function requestBuffer(url, agent) {
return new Promise((resolve, reject) => {
let u;
try {
u = new URL(url);
} catch (e) {
reject(e);
return;
}
const mod = u.protocol === "https:" ? https : http;
const req = mod.get(
url,
{
agent,
headers: { "User-Agent": "codex-market" },
},
(res) => {
const loc = res.headers.location;
if (res.statusCode >= 300 && res.statusCode < 400 && loc) {
res.resume();
requestBuffer(new URL(loc, url).toString(), agent).then(resolve, reject);
return;
}
if (res.statusCode < 200 || res.statusCode >= 300) {
res.resume();
reject(new Error("HTTP " + res.statusCode));
return;
}
const chunks = [];
res.on("data", (c) => chunks.push(c));
res.on("end", () => resolve(Buffer.concat(chunks)));
}
);
req.setTimeout(120000, () => req.destroy(new Error("下载超时")));
req.on("error", reject);
});
}
// —— 下载任意远程资源 → base64 字符串 ——
async function downloadAsBase64(url) {
const res = await fetch(url);
if (!res.ok) throw new Error("下载失败 HTTP " + res.status);
const buf = Buffer.from(await res.arrayBuffer());
return buf.toString("base64");
const proxyUrl = systemProxyUrl();
let proxyErr = "";
if (proxyUrl) {
const agent = proxyAgentFor(url, proxyUrl);
if (agent) {
try {
return (await requestBuffer(url, agent)).toString("base64");
} catch (e) {
proxyErr = String((e && e.message) || e);
// 代理下载失败时再直连兜底,保留原行为。
}
}
}
const ctrl = new AbortController();
const timer = setTimeout(() => ctrl.abort(), 120000);
try {
const res = await fetch(url, { headers: { "User-Agent": "codex-market" }, signal: ctrl.signal });
if (!res.ok) throw new Error("下载失败 HTTP " + res.status);
const buf = Buffer.from(await res.arrayBuffer());
return buf.toString("base64");
} catch (e) {
const directErr = String((e && e.message) || e);
throw new Error("下载失败:" + directErr + (proxyErr ? ";代理下载也失败:" + proxyErr : ""));
} finally {
clearTimeout(timer);
}
}
// —— 组装 data URL ——
@@ -290,7 +400,8 @@ async function runTts(base, llm, params) {
// —— 递归深挖视频地址(覆盖 result_url、data.content.video_url、data.data.content.video_url 等嵌套)——
// port 自 提示词提取/server.js
function findVideoUrl(o) {
if (!o || typeof o === "string") return "";
if (!o) return "";
if (typeof o === "string") return /^https?:\/\//.test(o) ? o : "";
if (Array.isArray(o)) {
for (const v of o) {
const u = findVideoUrl(v);
@@ -324,6 +435,27 @@ function joinUrl(base, p) {
return trimSlash(base) + (String(p || "").startsWith("/") ? p : "/" + p);
}
function pickTaskId(o) {
if (!o || typeof o !== "object") return "";
for (const k of ["request_id", "requestId", "task_id", "taskId", "id"]) {
if (typeof o[k] === "string" && o[k]) return o[k];
}
for (const k of ["data", "result", "task"]) {
const v = o[k];
if (v && typeof v === "object" && !Array.isArray(v)) {
const id = pickTaskId(v);
if (id) return id;
}
}
return "";
}
function pushRecentPoll(arr, value) {
if (!value) return;
arr.push(String(value).slice(0, 500));
if (arr.length > 3) arr.shift();
}
// —— Seedance/TokenDun 视频:异步任务制(建任务 → 轮询 → 下载成品转 base64)——
// mode: "text"(文生视频,只用 prompt) | "image"(图生视频)。model 须用户填写。
// 图生视频素材两条来路:params.imageUrl(手填公网URL,优先) 或 params.image(上游 base64)。
@@ -472,6 +604,21 @@ async function runGrokVideo(base, llm, params) {
}
const createUrl = joinUrl(base, params.endpointPath || "/v1/videos/generations");
const debugBase = {
base: hostOf(base),
create_url: createUrl,
mode,
request: {
model: body.model,
prompt: body.prompt,
resolution: body.resolution || "",
duration: body.duration || "",
aspect_ratio: body.aspect_ratio || "",
has_image: !!body.image,
reference_image_count: Array.isArray(body.reference_images) ? body.reference_images.length : 0,
},
};
grokDebugLog({ type: "grok_video_create", ...debugBase });
const cres = await fetch(createUrl, {
method: "POST",
headers: headers(llm.key),
@@ -479,15 +626,25 @@ async function runGrokVideo(base, llm, params) {
});
if (!cres.ok) {
const detail = await cres.text().catch(() => "");
return { ok: false, error: `Grok 视频建任务失败(${cres.status}):${detail.slice(0, 300)}` };
grokDebugLog({ type: "grok_video_create_failed", ...debugBase, status: cres.status, detail: detail.slice(0, 300) });
return { ok: false, error: `Grok 视频建任务失败(${cres.status}):POST ${createUrl} → ${detail.slice(0, 300)}` };
}
const data = await cres.json();
const requestId = data.request_id || data.id || (data.data && (data.data.request_id || data.data.id));
let data;
let rawCreate = "";
try {
rawCreate = await cres.text();
data = JSON.parse(rawCreate);
} catch {
return { ok: false, error: `Grok 视频建任务返回非 JSON:${rawCreate.slice(0, 300)}` };
}
const requestId = pickTaskId(data);
if (!requestId) return { ok: false, error: `Grok 视频建任务未返回 request_id:${JSON.stringify(data).slice(0, 300)}` };
grokDebugLog({ type: "grok_video_accepted", ...debugBase, request_id: requestId });
const pollUrl = joinUrl(base, "/v1/videos/" + encodeURIComponent(requestId));
const pollHeaders = { Authorization: "Bearer " + (llm.key || ""), Accept: "application/json", "User-Agent": "codex-market" };
let lastPoll = "";
const recentPolls = [];
for (let i = 0; i < 120; i++) {
await sleep(5000);
let pd;
@@ -502,21 +659,38 @@ async function runGrokVideo(base, llm, params) {
lastPoll = "请求异常:" + String((e && e.message) || e);
continue;
}
pushRecentPoll(recentPolls, JSON.stringify(pd));
const d = pd && pd.data && typeof pd.data === "object" && !Array.isArray(pd.data) ? pd.data : pd;
const st = normalizeVideoStatus((d && (d.status || d.state)) || (pd && pd.status));
lastPoll = "status=" + st;
if (st === "succeeded") {
const url = findVideoUrl(pd);
if (!url) return { ok: false, error: `Grok 视频已完成但未找到视频地址:${JSON.stringify(pd).slice(0, 500)}` };
if (!url) {
lastPoll = "status=succeeded,但暂未返回视频地址";
continue;
}
const b64 = await downloadAsBase64(url);
grokDebugLog({ type: "grok_video_done", ...debugBase, request_id: requestId, status: st, url });
return { ok: true, mime: "video/mp4", dataUrl: toDataUrl("video/mp4", b64), url, _model: model };
}
if (st === "failed") {
const err = (d && d.error && (d.error.message || d.error)) || (d && (d.fail_reason || d.message || d.reason)) || (pd && pd.message) || st;
grokDebugLog({ type: "grok_video_failed", ...debugBase, request_id: requestId, status: st, error: typeof err === "string" ? err : JSON.stringify(err) });
return { ok: false, error: "Grok 视频任务失败:" + (typeof err === "string" ? err : JSON.stringify(err)) };
}
}
return { ok: false, error: "Grok 视频轮询 10 分钟未完成(request_id=" + requestId + ",最近一次轮询:" + (lastPoll || "无") + ")" };
return {
ok: false,
error:
"Grok 视频轮询 10 分钟未完成(GET " +
pollUrl +
",request_id=" +
requestId +
",最近一次轮询:" +
(lastPoll || "无") +
(recentPolls.length ? ",最近原始返回:" + recentPolls.join(" | ") : "") +
")",
};
} catch (e) {
return { ok: false, error: "Grok 视频生成异常:" + String((e && e.message) || e) };
}
+72 -2
View File
@@ -234,8 +234,12 @@ function buildThreadUpsertSql(records, target, cols) {
const updateSet = [];
for (const c of desired) {
if (c === "id") continue;
if (c === "has_user_event") updateSet.push(`${c}=CASE WHEN excluded.${c}=1 THEN 1 ELSE threads.${c} END`);
else if (["cwd", "first_user_message", "preview", "thread_source"].includes(c)) {
// 【增强】model_provider 强制更新为目标 provider(切账号核心逻辑)
if (c === "model_provider") {
updateSet.push(`${c}=${sql(target)}`);
} else if (c === "has_user_event") {
updateSet.push(`${c}=CASE WHEN excluded.${c}=1 THEN 1 ELSE threads.${c} END`);
} else if (["cwd", "first_user_message", "preview", "thread_source"].includes(c)) {
updateSet.push(`${c}=CASE WHEN COALESCE(excluded.${c},'')<>'' THEN excluded.${c} ELSE threads.${c} END`);
} else if (c === "title") {
updateSet.push(`${c}=CASE WHEN COALESCE(threads.${c},'')='' OR threads.${c}='(无标题对话)' THEN excluded.${c} ELSE threads.${c} END`);
@@ -252,6 +256,7 @@ function buildThreadUpsertSql(records, target, cols) {
`ON CONFLICT(id) DO UPDATE SET ${updateSet.join(",")};`
);
}
// 【增强】确保数据库中所有会话的 provider 都更新为目标(兜底逻辑)
lines.push(`UPDATE threads SET model_provider=${sql(target)} WHERE COALESCE(model_provider,'')<>${sql(target)};`);
lines.push("COMMIT;");
return lines.join("\n") + "\n";
@@ -288,6 +293,59 @@ function ts() {
return `${d.getFullYear()}${p(d.getMonth() + 1)}${p(d.getDate())}-${p(d.getHours())}${p(d.getMinutes())}${p(d.getSeconds())}`;
}
// 【增强】重建 session_index.jsonl:确保所有会话都在索引中(切账号后必要)
function rebuildSessionIndex(records, log = () => {}) {
if (!records || !records.length) {
log("索引重建:无会话记录,跳过");
return { rebuilt: 0 };
}
// 备份旧索引
if (fs.existsSync(SESSION_INDEX)) {
const backup = SESSION_INDEX + ".bak-" + ts();
try {
fs.copyFileSync(SESSION_INDEX, backup);
log(`已备份旧索引 → ${path.basename(backup)}`);
} catch (e) {
log(`索引备份失败:${e.message || e}`);
}
}
// 读取现有索引,保留用户重命名的标题
const existingTitles = {};
try {
const lines = fs.readFileSync(SESSION_INDEX, "utf-8").split("\n");
for (const line of lines) {
if (!line.trim()) continue;
try {
const d = JSON.parse(line);
if (d.id && d.thread_name) existingTitles[d.id] = d.thread_name;
} catch {}
}
} catch {}
// 重建索引:每个会话一行 JSON
const lines = [];
for (const r of records) {
if (!r.id) continue;
const indexEntry = {
id: r.id,
thread_name: existingTitles[r.id] || r.title || "(无标题对话)",
project_path: r.cwd || null,
created_at: new Date(r.created_at_ms).toISOString(),
source: r.source || "vscode",
archived: r.archived ? true : false,
};
lines.push(JSON.stringify(indexEntry));
}
// 写入新索引
try {
fs.writeFileSync(SESSION_INDEX, lines.join("\n") + "\n", "utf-8");
log(`索引重建完成:${lines.length} 条会话`);
return { rebuilt: lines.length };
} catch (e) {
log(`索引重建失败:${e.message || e}`);
return { rebuilt: 0, error: e.message || String(e) };
}
}
async function syncProviderToCurrent(log = () => {}, onProgress = () => {}) {
if (codexRunning()) throw new Error("检测到 Codex 正在运行,请先完全退出 Codex 再同步。");
const target = targetProvider();
@@ -394,6 +452,11 @@ async function syncProviderToCurrent(log = () => {}, onProgress = () => {}) {
}
}
log(`数据库更新:provider ${provRows} 行、has_user_event ${ueRows} 行、新增线程 ${insertedRows} 行、项目索引 +${projectState.addedProjects}`);
// 【增强】4) 重建 session_index.jsonl,确保所有会话都在索引中
onProgress({ phase: "重建索引", done: 0, total: 0 });
const indexResult = rebuildSessionIndex(records, log);
onProgress({ phase: "完成", done: 0, total: 0 });
return {
@@ -407,6 +470,7 @@ async function syncProviderToCurrent(log = () => {}, onProgress = () => {}) {
insertedRows,
projectRows: projectState.addedProjects,
projectlessRemoved: projectState.removedProjectless,
indexRebuilt: indexResult.rebuilt || 0,
backup,
};
}
@@ -457,6 +521,11 @@ async function syncProviderToCurrent(log = () => {}, onProgress = () => {}) {
}
}
log(`数据库更新:provider ${provRows} 行、has_user_event ${ueRows} 行、cwd ${cwdRows} 行`);
// 【增强】4) 重建 session_index.jsonl,确保所有会话都在索引中
onProgress({ phase: "重建索引", done: 0, total: 0 });
const indexResult = rebuildSessionIndex(records, log);
onProgress({ phase: "完成", done: 0, total: 0 });
return {
@@ -470,6 +539,7 @@ async function syncProviderToCurrent(log = () => {}, onProgress = () => {}) {
insertedRows: 0,
projectRows: projectState.addedProjects,
projectlessRemoved: projectState.removedProjectless,
indexRebuilt: indexResult.rebuilt || 0,
backup,
};
}
+72 -10
View File
@@ -304,6 +304,16 @@ function llmGroup() {
return config.groups.find(usable) || null;
}
// 供「语音输入」转写用:优先当前激活且勾选 STT 的分组,否则取任一勾选 STT 的分组。
// 需配齐 baseUrl+key 且勾了 stt;返回的 group 带 sttModel(如 speech-01 / whisper-1)。
function sttGroup() {
load();
const ok = (g) => g && !g.isDefault && g.baseUrl && g.key && g.stt;
const act = config.groups.find((g) => g.id === config.activeId);
if (ok(act)) return act;
return config.groups.find(ok) || null;
}
// 拉取某个中转 API 的模型列表(试 /v1/models 再 /models)
async function fetchModels(baseUrl, key) {
if (!baseUrl) throw new Error("请先填 API 地址");
@@ -343,6 +353,64 @@ function writeCapture(name, data) {
} catch {}
}
function requestTextForIntent(req) {
const parts = [];
const walk = (v) => {
if (v == null) return;
if (typeof v === "string") {
parts.push(v);
return;
}
if (Array.isArray(v)) {
for (const it of v) walk(it);
return;
}
if (typeof v === "object") {
for (const key of ["text", "content", "input", "output", "arguments"]) walk(v[key]);
}
};
walk(req && req.input);
return parts.join("\n");
}
function hasExplicitGenImageIntent(req) {
return /\$gen[\s_-]*image\b/i.test(requestTextForIntent(req));
}
function stripGenImageSkill(instructions) {
if (typeof instructions !== "string" || !instructions.includes("gen-image")) return instructions;
return instructions
.split(/\r?\n/)
.filter((line) => !/^\s*-\s+gen-image\s*:/.test(line))
.join("\n");
}
function rewriteResponsesBody(body, group) {
if (!body.length) return body;
let j;
try {
j = JSON.parse(body.toString("utf-8"));
} catch {
return body;
}
if (!j || typeof j !== "object") return body;
let changed = false;
if (group.model && "model" in j && j.model !== group.model) {
j.model = group.model;
changed = true;
}
// 普通聊天分组没有 image-generation 权限时,上游会因 gen-image skill 声明误判为生图请求。
// 只有用户显式输入 "$gen image" 才保留该 skill,让真正生图走脚本里的图片接口分组。
if (!hasExplicitGenImageIntent(j)) {
const nextInstructions = stripGenImageSkill(j.instructions);
if (nextInstructions !== j.instructions) {
j.instructions = nextInstructions;
changed = true;
}
}
return changed ? Buffer.from(JSON.stringify(j), "utf-8") : body;
}
function handle(req, res) {
const g = activeGroup();
if (!g) {
@@ -360,16 +428,7 @@ function handle(req, res) {
req.on("end", () => {
let body = Buffer.concat(chunks);
const doCapture = /responses/.test(req.url || "");
// 分组指定了模型 → 覆盖请求体里的 model
if (g.model && body.length) {
try {
const j = JSON.parse(body.toString("utf-8"));
if (j && typeof j === "object" && "model" in j) {
j.model = g.model;
body = Buffer.from(JSON.stringify(j), "utf-8");
}
} catch {}
}
if (doCapture) body = rewriteResponsesBody(body, g);
// —— chat 模式:Responses ⇄ Chat Completions 协议转换(DeepSeek/GLM/Kimi 等)——
if (g.mode === "chat" && /responses/.test(req.url || "")) {
let responsesReq = {};
@@ -577,6 +636,8 @@ function status() {
hasKey: g.isDefault ? true : !!g.key,
isDefault: !!g.isDefault,
codexImage: !!g.codexImage,
stt: !!g.stt,
sttModel: g.sttModel || "",
})),
};
}
@@ -880,5 +941,6 @@ module.exports = {
groupById,
usableGroups,
llmGroup,
sttGroup,
fetchModels,
};
+187
View File
@@ -0,0 +1,187 @@
// session_sync_enhanced.js —— 增强版会话同步:确保本地所有会话一个不剩地同步到新 Codex 账号。
// 在 provider_sync.js 的基础上增强,确保:
// 1. 扫描所有 sessions/ 和 archived_sessions/ 下的会话
// 2. 强制插入到当前账号的数据库(UPSERT 逻辑)
// 3. 重建 session_index.jsonl
// 4. 更新全局项目索引
// 5. 提供详细的同步报告
const fs = require("fs");
const path = require("path");
const os = require("os");
const providerSync = require("./provider_sync");
const CODEX_HOME = path.join(os.homedir(), ".codex");
const SESSION_INDEX = path.join(CODEX_HOME, "session_index.jsonl");
// 完整同步方案:确保所有本地会话都进入新账号
async function syncAllSessionsToNewAccount(opts = {}) {
const log = opts.log || (() => {});
const onProgress = opts.onProgress || (() => {});
log("=== 开始全量会话同步 ===");
log("目标:将本地所有会话同步到当前 Codex 账号");
// 1. 检查 Codex 是否在运行
if (providerSync.codexRunning()) {
throw new Error("检测到 Codex 正在运行,请先完全退出 Codex 再同步。");
}
// 2. 获取当前 provider(从 config.toml)
const targetProvider = providerSync.targetProvider();
log(`当前账号 provider:${targetProvider}`);
// 3. 执行核心同步(provider_sync 已经做了 UPSERT)
log("开始同步本地会话到数据库...");
const result = await providerSync.syncProviderToCurrent(log, onProgress);
// 4. 重建 session_index.jsonl(确保索引完整)
log("重建会话索引...");
const indexResult = rebuildSessionIndex(log);
// 5. 生成同步报告
const report = {
success: true,
provider: result.target,
totalSessions: result.totalConversations,
syncedToDatabase: result.insertedRows,
updatedProvider: result.provRows,
updatedUserEvent: result.ueRows,
indexRebuilt: indexResult.total,
projects: result.projectRows,
backup: result.backup,
details: {
changedFiles: result.changedFiles,
userEventSessions: result.userEvents,
archivedSessions: indexResult.archived,
activeSessions: indexResult.active,
}
};
log("\n=== 同步完成 ===");
log(`✅ 总会话数:${report.totalSessions}`);
log(`✅ 新增到数据库:${report.syncedToDatabase} 条`);
log(`✅ 更新 provider:${report.updatedProvider} 条`);
log(`✅ 重建索引:${report.indexRebuilt} 条`);
log(`✅ 项目分组:${report.projects} 个`);
log(`✅ 备份位置:${report.backup}`);
log("\n现在可以启动 Codex,所有会话应该都能看到了。");
return report;
}
// 重建 session_index.jsonl:扫描所有会话文件,生成完整索引
function rebuildSessionIndex(log = () => {}) {
const sessions = require("./sessions");
// 列出所有会话(包括 archived)
const allConversations = sessions.listConversations({
includeSubagent: true,
includeArchived: true,
});
log(`扫描到 ${allConversations.length} 个会话文件`);
// 备份旧索引
if (fs.existsSync(SESSION_INDEX)) {
const backup = SESSION_INDEX + ".bak-" + Date.now();
fs.copyFileSync(SESSION_INDEX, backup);
log(`已备份旧索引:${path.basename(backup)}`);
}
// 重建索引:每个会话一行 JSON
const lines = [];
let active = 0;
let archived = 0;
for (const conv of allConversations) {
if (!conv.id) continue;
const indexEntry = {
id: conv.id,
thread_name: conv.title || "(无标题对话)",
project_path: conv.cwd || null,
created_at: conv.ts || new Date().toISOString(),
source: conv.source || "vscode",
archived: conv.archived || false,
};
lines.push(JSON.stringify(indexEntry));
if (conv.archived) archived++;
else active++;
}
// 写入新索引
fs.writeFileSync(SESSION_INDEX, lines.join("\n") + "\n", "utf-8");
log(`已重建索引:活跃 ${active} 条,归档 ${archived} 条`);
return { total: lines.length, active, archived };
}
// 诊断:检查会话同步状态
function diagnoseSessionSync(log = () => {}) {
const sessions = require("./sessions");
log("=== 会话同步诊断 ===");
// 1. 统计本地会话文件
const allConversations = sessions.listConversations({
includeSubagent: true,
includeArchived: true,
});
const activeSessions = allConversations.filter(c => !c.archived);
const archivedSessions = allConversations.filter(c => c.archived);
log(`本地会话文件:${allConversations.length} 个`);
log(` - 活跃:${activeSessions.length} 个`);
log(` - 归档:${archivedSessions.length} 个`);
// 2. 检查数据库
const dbFiles = fs.readdirSync(CODEX_HOME).filter(f => /^state_.*\.sqlite$/.test(f));
log(`数据库文件:${dbFiles.length} 个`);
for (const db of dbFiles) {
log(` - ${db}`);
}
// 3. 检查索引文件
const indexExists = fs.existsSync(SESSION_INDEX);
log(`session_index.jsonl:${indexExists ? "存在" : "不存在"}`);
if (indexExists) {
const indexContent = fs.readFileSync(SESSION_INDEX, "utf-8");
const indexLines = indexContent.split("\n").filter(l => l.trim());
log(` - 索引条目:${indexLines.length} 条`);
}
// 4. 检查 provider
const currentProvider = providerSync.targetProvider();
log(`当前 provider:${currentProvider}`);
// 5. 检查会话文件中的 provider 分布
const providerStats = {};
for (const conv of allConversations) {
const p = conv.provider || "(未设置)";
providerStats[p] = (providerStats[p] || 0) + 1;
}
log("会话文件 provider 分布:");
for (const [p, count] of Object.entries(providerStats)) {
log(` - ${p}: ${count} 个`);
}
return {
totalSessions: allConversations.length,
activeSessions: activeSessions.length,
archivedSessions: archivedSessions.length,
databases: dbFiles.length,
indexExists,
currentProvider,
providerStats,
};
}
module.exports = {
syncAllSessionsToNewAccount,
rebuildSessionIndex,
diagnoseSessionSync,
};
+617
View File
@@ -0,0 +1,617 @@
// voice_cdp.js —— Codex 语音输入注入:主进程侧。
// 用 CDP(Chrome DevTools Protocol)连接带 --remote-debugging-port 的 Codex 桌面版,
// 往其页面注入麦克风按钮脚本,录音经 binding 回传 → 调中转站 Whisper 转写 → 文字回填 composer。
// 不改 Codex 源码、不写补丁文件。无第三方依赖:CDP over WebSocket 用 Node 内置 net 手写最小 WS 客户端
// (Electron 34 主进程的 Node 无全局 WebSocket);HTTP 用全局 fetch;multipart 用全局 FormData/Blob。
//
// 机制参考 CodexPlusPlus 的 bridge.rs:Runtime.evaluate / Page.addScriptToEvaluateOnNewDocument /
// Runtime.addBinding + Runtime.bindingCalled。
const { spawn, execFile } = require("child_process");
const os = require("os");
const path = require("path");
const net = require("net");
const crypto = require("crypto");
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 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 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)
// —— 状态机 ——
// unknown / checking / not_running / running_no_port / restarting / connecting /
// injecting / connected / disconnected / error
let state = {
phase: "unknown",
port: DEFAULT_PORT,
codexPath: "",
lastError: "",
target: null, // { id, url, title }
};
let session = null; // 当前 CdpSession
let onStatusCb = null;
function snapshot() {
const { phase, port, codexPath, lastError, model, target } = state;
return { phase, port, codexPath, lastError, model, target };
}
function setPhase(phase, extra = {}) {
state = { ...state, phase, ...extra };
if (onStatusCb) {
try {
onStatusCb(snapshot());
} catch {}
}
}
// —— 探测 Codex 可执行路径 ——
function codexExecutable() {
const home = os.homedir();
if (process.platform === "darwin") {
const cands = [
path.join(home, "Applications/ChatGPT.app/Contents/MacOS/ChatGPT"),
"/Applications/ChatGPT.app/Contents/MacOS/ChatGPT",
"/Applications/Codex.app/Contents/MacOS/Codex",
];
for (const p of cands) {
try {
if (require("fs").existsSync(p)) return p;
} catch {}
}
return "";
}
if (process.platform === "win32") {
const local = process.env.LOCALAPPDATA || "";
const pf = process.env["ProgramFiles"] || "C:\\Program Files";
const cands = [
local && path.join(local, "Programs\\ChatGPT\\ChatGPT.exe"),
path.join(pf, "ChatGPT\\ChatGPT.exe"),
].filter(Boolean);
for (const p of cands) {
try {
if (require("fs").existsSync(p)) return p;
} catch {}
}
return "";
}
return "";
}
// —— HTTP:GET /json 取 targets ——
async function listTargets(port) {
const ctrl = new AbortController();
const t = setTimeout(() => ctrl.abort(), CDP_HTTP_TIMEOUT);
try {
const res = await fetch(`http://127.0.0.1:${port}/json`, { signal: ctrl.signal });
if (!res.ok) return [];
const arr = await res.json();
return Array.isArray(arr) ? arr : [];
} catch {
return [];
} finally {
clearTimeout(t);
}
}
// —— 挑可注入的 Codex 主页面 target ——
// Codex 可能有多个 page:主对话页 app://-/index.html,以及 avatar-overlay / global-dictation
// 等带 ?initialRoute= 的子路由覆盖层。优先选「无 query 的主入口」,次选普通 codex 页,排除覆盖层。
function pickTarget(targets) {
let main = null; // url 恰为 app://-/index.html(无 query)
let fallback = null; // 其它非覆盖层 codex 页
for (const t of targets) {
if (t.type !== "page") continue;
const ws = t.webSocketDebuggerUrl;
if (!ws) continue;
const url = (t.url || "").toLowerCase();
const title = (t.title || "").toLowerCase();
// 覆盖层:带 initialRoute 的子路由(avatar-overlay / global-dictation 等)
const isOverlayRoute = url.includes("initialroute=");
const isCodex =
url.startsWith("app://-/index.html") || url.includes("index.html") || title.includes("codex");
if (!isCodex) continue;
if (isOverlayRoute) continue; // 覆盖层不注入
if (url === "app://-/index.html" || url === "app://-/index.html/") {
main = main || t;
} else {
fallback = fallback || t;
}
}
return main || fallback;
}
// —— 最小 WebSocket 客户端(Node 内置 net,无需依赖;仅支持 CDP 需要的文本帧)——
// 解析 ws://host:port/path,TCP 连接 → HTTP 升级握手(带 Sec-WebSocket-Key)→ 文本帧收发。
// 客户端发送必须掩码;CDP 服务端发送不掩码。帧可能分片到达,用缓冲区拼。
function wsConnect(wsUrl) {
const u = new URL(wsUrl);
const host = u.hostname;
const port = u.port || (u.protocol === "wss:" ? "443" : "80");
const pathStr = u.pathname + u.search;
const key = crypto.randomBytes(16).toString("base64");
const socket = net.connect({ host, port: Number(port) });
const events = { open: [], message: [], close: [], error: [] };
const emit = (name, arg) => (events[name] || []).forEach((cb) => cb(arg));
// —— 发送一帧文本(客户端必须掩码)——
function sendFrame(str) {
const payload = Buffer.from(str, "utf8");
const len = payload.length;
// opcode 1 = 文本;FIN=1
let header;
const mask = crypto.randomBytes(4);
if (len < 126) {
header = Buffer.alloc(6);
header[0] = 0x81;
header[1] = 0x80 | len;
mask.copy(header, 2);
} else if (len < 65536) {
header = Buffer.alloc(8);
header[0] = 0x81;
header[1] = 0x80 | 126;
header.writeUInt16BE(len, 2);
mask.copy(header, 4);
} else {
header = Buffer.alloc(14);
header[0] = 0x81;
header[1] = 0x80 | 127;
header.writeUInt32BE(0, 2);
header.writeUInt32BE(len, 6);
mask.copy(header, 10);
}
const masked = Buffer.alloc(len);
for (let i = 0; i < len; i++) masked[i] = payload[i] ^ mask[i & 3];
socket.write(Buffer.concat([header, masked]));
}
// —— 接收缓冲 + 帧解析 ——
let buf = Buffer.alloc(0);
let handshakeDone = false;
function parseFrames() {
while (true) {
if (buf.length < 2) return;
const b0 = buf[0];
const b1 = buf[1];
const opcode = b0 & 0x0f;
const masked = (b1 & 0x80) !== 0;
let payloadLen = b1 & 0x7f;
let offset = 2;
if (payloadLen === 126) {
if (buf.length < 4) return;
payloadLen = buf.readUInt16BE(2);
offset = 4;
} else if (payloadLen === 127) {
if (buf.length < 10) return;
// 高 32 位忽略(CDP 帧不会那么大)
payloadLen = Number(buf.readUInt32BE(6));
offset = 10;
}
let maskKey = null;
if (masked) {
if (buf.length < offset + 4) return;
maskKey = buf.slice(offset, offset + 4);
offset += 4;
}
if (buf.length < offset + payloadLen) return;
let payload = buf.slice(offset, offset + payloadLen);
if (masked) {
const un = Buffer.alloc(payload.length);
for (let i = 0; i < payload.length; i++) un[i] = payload[i] ^ maskKey[i & 3];
payload = un;
}
buf = buf.slice(offset + payloadLen);
// 只处理文本(1)/ping(9)/pong(10)/关闭(8)
if (opcode === 0x8) {
// 关闭帧
emit("close");
try {
socket.end();
} catch {}
return;
} else if (opcode === 0x9) {
// ping → 回 pong(无掩码服务端不要求,但回时按服务端→客户端本可不掩码;此处回文本帧不规范,CDP 不发 ping)
continue;
} else if (opcode === 0xa) {
continue; // pong,忽略
} else {
// 文本帧(含分片首帧 opcode 0 也按文本拼;CDP 单帧文本足够,简单按单帧处理)
emit("message", payload.toString("utf8"));
}
}
}
socket.on("connect", () => {
const req =
`GET ${pathStr} HTTP/1.1\r\n` +
`Host: ${host}:${port}\r\n` +
`Upgrade: websocket\r\n` +
`Connection: Upgrade\r\n` +
`Sec-WebSocket-Key: ${key}\r\n` +
`Sec-WebSocket-Version: 13\r\n\r\n`;
socket.write(req);
});
socket.on("data", (chunk) => {
buf = Buffer.concat([buf, chunk]);
if (!handshakeDone) {
const idx = buf.indexOf("\r\n\r\n");
if (idx < 0) return;
const head = buf.slice(0, idx).toString("utf8");
buf = buf.slice(idx + 4);
if (!/101/.test(head.split("\r\n")[0])) {
emit("error", new Error("WS 握手失败: " + head.split("\r\n")[0]));
try {
socket.destroy();
} catch {}
return;
}
handshakeDone = true;
emit("open");
parseFrames();
} else {
parseFrames();
}
});
socket.on("close", () => emit("close"));
socket.on("error", (e) => emit("error", e));
return {
send: sendFrame,
close: () => {
try {
socket.destroy();
} catch {}
},
on: (name, cb) => {
(events[name] = events[name] || []).push(cb);
},
_socket: socket,
};
}
// —— CDP WebSocket 会话 ——
function openSession(wsUrl, onBindingCalled) {
const pending = new Map(); // id → {resolve, reject}
const handlers = new Map(); // method → Set<cb>
let nextId = 1;
let closed = false;
const ws = wsConnect(wsUrl);
function send(method, params = {}) {
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 });
ws.send(JSON.stringify({ id, method, params }));
});
}
function on(method, cb) {
if (!handlers.has(method)) handlers.set(method, new Set());
handlers.get(method).add(cb);
}
function close() {
closed = true;
try {
ws.close();
} catch {}
}
ws.on("message", (data) => {
let msg;
try {
msg = JSON.parse(data);
} catch {
return;
}
if (msg.id != null) {
const p = pending.get(msg.id);
if (!p) return;
pending.delete(msg.id);
if (msg.error) p.reject(new Error(msg.error.message || "CDP error"));
else p.resolve(msg.result);
return;
}
if (msg.method) {
if (msg.method === "Runtime.bindingCalled" && onBindingCalled) {
try {
onBindingCalled(msg.params);
} catch {}
}
const set = handlers.get(msg.method);
if (set) for (const cb of set) cb(msg.params);
}
});
ws.on("close", () => {
closed = true;
for (const { reject } of pending.values()) {
try {
reject(new Error("CDP 会话关闭"));
} catch {}
}
pending.clear();
});
return new Promise((resolve, reject) => {
ws.on("open", () => resolve({ send, on, close, ws }));
ws.on("error", (e) => reject(new Error("CDP 连接失败:" + (e && e.message ? e.message : "ws error"))));
});
}
// —— 转写:中转站 /v1/audio/transcriptions(用 sttGroup:勾选了 STT 的分组)——
async function runStt(audioB64, mime) {
const g = router.sttGroup();
if (!g || !g.baseUrl || !g.key) throw new Error("未配置支持语音转文字的中转分组(在路由中转里勾选「支持语音转文字」)");
const base = (g.baseUrl || "").replace(/\/+$/, "");
const model = g.sttModel || "whisper-1";
const buf = Buffer.from(audioB64, "base64");
const ext = mime && mime.includes("webm") ? "webm" : mime && mime.includes("mp4") ? "mp4" : "wav";
const form = new FormData();
form.append("file", new Blob([buf], { type: mime || "audio/webm" }), `voice.${ext}`);
form.append("model", model);
form.append("response_format", "json");
const res = await fetch(base + "/v1/audio/transcriptions", {
method: "POST",
headers: { Authorization: "Bearer " + g.key },
body: form,
});
dlog("转写请求 →", base + "/v1/audio/transcriptions", "model=" + model, "size=" + buf.length, "HTTP " + res.status);
if (!res.ok) {
const detail = await res.text().catch(() => "");
throw new Error(`转写接口 HTTP ${res.status}:${detail.slice(0, 300)}`);
}
const j = await res.json();
return (j && (j.text || "")).trim();
}
// —— 主进程收到页面 binding 回传:处理转写并把文字送回页面 ——
async function handleBinding(params) {
const name = params && params.name;
if (name !== BINDING_NAME) return;
let payload;
try {
payload = JSON.parse(params.payload);
} catch {
return;
}
if (!payload || payload.kind !== "transcribe") return;
const { id, audioB64, mime } = payload;
dlog("binding 收到转写请求 id=" + id, "mime=" + mime, "audioB64长度=" + (audioB64?audioB64.length:0));
setPhase("transcribing");
let text = "";
let err = "";
try {
text = await runStt(audioB64, mime);
dlog("转写结果:", JSON.stringify(text).slice(0, 120));
} catch (e) {
err = String((e && e.message) || e);
dlog("转写失败:", err);
}
setPhase(connectedPhase());
// 把结果回调给页面
const expr = `window.__voiceInjectHandle && window.__voiceInjectHandle(${JSON.stringify(
{ id, text, err }
)})`;
try {
if (session) await session.send("Runtime.evaluate", { expression: expr, awaitPromise: false });
} catch {}
}
function connectedPhase() {
return session ? "connected" : "disconnected";
}
// —— 注入:addBinding + addScriptToEvaluateOnNewDocument + 立即 evaluate ——
async function inject(sessionObj) {
await sessionObj.send("Runtime.enable");
try {
await sessionObj.send("Runtime.removeBinding", { name: BINDING_NAME });
} catch {}
await sessionObj.send("Runtime.addBinding", { name: BINDING_NAME });
await sessionObj.send("Page.enable");
await sessionObj.send("Page.addScriptToEvaluateOnNewDocument", { source: INJECT_SCRIPT });
await sessionObj.send("Runtime.evaluate", { expression: INJECT_SCRIPT, awaitPromise: false });
}
// —— 连接到已带端口的 Codex ——
async function connectToTarget(port) {
const targets = await listTargets(port);
const tgt = pickTarget(targets);
if (!tgt) return false;
state.target = { id: tgt.id, url: tgt.url, title: tgt.title };
setPhase("connecting");
session = await openSession(tgt.webSocketDebuggerUrl, handleBinding);
session.on("Runtime.bindingCalled", handleBinding);
session.ws.on("close", () => {
session = null;
setPhase("disconnected");
});
setPhase("injecting");
await inject(session);
setPhase("connected", { target: state.target });
return true;
}
// —— 轮询直到出现 page target(重启后用)——
async function waitForReady(port) {
const t0 = Date.now();
while (Date.now() - t0 < READY_POLL_MAX_MS) {
const targets = await listTargets(port);
const tgt = pickTarget(targets);
if (tgt) return tgt;
await new Promise((r) => setTimeout(r, READY_POLL_INTERVAL));
}
return null;
}
// —— 启动接管:杀当前 Codex → 带端口重启 ——
function killCodex() {
return new Promise((resolve) => {
if (process.platform === "darwin") {
execFile("osascript", ["-e", 'tell application "ChatGPT" to quit'], () => {
// 兜底:osascript 没杀掉就 pkill
execFile("pgrep", ["-f", "ChatGPT.app/Contents/MacOS/ChatGPT"], (_e, out) => {
if (out && out.trim()) {
execFile("pkill", ["-f", "ChatGPT.app/Contents/MacOS/ChatGPT"], () =>
setTimeout(resolve, 800)
);
} else resolve();
});
});
} else if (process.platform === "win32") {
execFile("taskkill", ["/IM", "ChatGPT.exe", "/F"], () =>
setTimeout(resolve, 800)
);
} else resolve();
});
}
function spawnCodex(exePath, port) {
const child = spawn(exePath, [`--remote-debugging-port=${port}`], {
detached: true,
stdio: "ignore",
});
try {
child.unref();
} catch {}
}
// —— 对外:status / connect / disconnect ——
function status() {
return snapshot();
}
function onStatus(cb) {
onStatusCb = cb;
}
// connect(onStatusCb):流程见计划。forceRestart=true 时直接重启(前端已确认)。
async function connect(cb, opts = {}) {
if (cb) onStatusCb = cb;
const port = opts.port || DEFAULT_PORT;
state.port = port;
setPhase("checking");
// 1. 端口已开?
const ok = await connectToTarget(port);
if (ok) return snapshot();
// 2. 判断 Codex 是否在跑
const running = providerSync.codexRunning ? providerSync.codexRunning() : false;
if (!running && !opts.forceRestart) {
// 没在跑也没开端口:直接尝试拉起带端口的实例(无需杀进程)
const exe = codexExecutable();
if (!exe) {
setPhase("error", { lastError: "未找到 Codex 可执行文件" });
return snapshot();
}
state.codexPath = exe;
setPhase("restarting", { codexPath: exe });
spawnCodex(exe, port);
const tgt = await waitForReady(port);
if (!tgt) {
setPhase("error", { lastError: "Codex 启动后未开放调试端口,请重试" });
return snapshot();
}
const ok2 = await connectToTarget(port);
if (!ok2) setPhase("error", { lastError: "已启动但未找到可注入页面" });
return snapshot();
}
// 3. 在跑但没端口:需用户在前端确认后 forceRestart 重启
if (running && !opts.forceRestart) {
state.codexPath = codexExecutable();
setPhase("running_no_port", { codexPath: state.codexPath });
return snapshot(); // 前端据此弹"重启确认"
}
// 4. forceRestart:杀进程 → 重启
const exe = codexExecutable();
if (!exe) {
setPhase("error", { lastError: "未找到 Codex 可执行文件" });
return snapshot();
}
state.codexPath = exe;
setPhase("restarting");
await killCodex();
spawnCodex(exe, port);
const tgt = await waitForReady(port);
if (!tgt) {
setPhase("error", { lastError: "重启 Codex 后未开放调试端口" });
return snapshot();
}
const ok3 = await connectToTarget(port);
if (!ok3) setPhase("error", { lastError: "已启动但未找到可注入页面" });
return snapshot();
}
function disconnect() {
if (session) {
try {
session.close();
} catch {}
session = null;
}
setPhase("disconnected");
return snapshot();
}
// 当前语音转写配置(来自路由里勾选 STT 的分组),供前端只读展示。
function sttInfo() {
const g = router.sttGroup();
if (!g) return { ready: false, groupName: "", model: "" };
return {
ready: true,
groupName: g.name || "",
model: g.sttModel || "whisper-1",
baseUrl: (g.baseUrl || "").replace(/\/+$/, ""),
};
}
// 诊断:用每个非默认分组的真 key 探测 /v1/audio/transcriptions 是否存在。
// 404 = 路径不存在(不支持STT);400/401/4xx-body非404/200 = 路径存在(支持,鉴权/参数问题另算)。
async function probeStt() {
const groups = router.usableGroups ? router.usableGroups() : [];
const out = [];
const body = Buffer.from([0x1a, 0x45, 0xdf, 0xa3]); // 假 webm 头,只为触发路径判断
for (const g of groups) {
const base = (g.baseUrl || "").replace(/\/+$/, "");
if (!base || !g.key) { out.push({ name: g.name, baseUrl: base, status: 0, note: "缺地址或密钥" }); continue; }
try {
const form = new FormData();
form.append("file", new Blob([body], { type: "audio/webm" }), "v.webm");
form.append("model", g.sttModel || "whisper-1");
const res = await fetch(base + "/v1/audio/transcriptions", {
method: "POST",
headers: { Authorization: "Bearer " + g.key },
body: form,
});
const text = await res.text().catch(() => "");
const notFound = res.status === 404 || /404 page not found|not found/i.test(text.slice(0, 60));
out.push({
name: g.name,
baseUrl: base,
status: res.status,
supported: !notFound,
note: text.slice(0, 120).replace(/\s+/g, " "),
});
} catch (e) {
out.push({ name: g.name, baseUrl: base, status: 0, supported: false, note: String((e && e.message) || e).slice(0, 120) });
}
}
return out;
}
module.exports = {
status,
connect,
disconnect,
onStatus,
sttInfo,
probeStt,
// 供临时实测脚本复用
_listTargets: listTargets,
_pickTarget: pickTarget,
_openSession: openSession,
_inject: inject,
DEFAULT_PORT,
};
+350
View File
@@ -0,0 +1,350 @@
// voice_inject.js —— 注入到 Codex 页面(app://-/index.html)的脚本,以字符串形式导出。
// 由 voice_cdp.js 经 CDP Page.addScriptToEvaluateOnNewDocument(常驻) + Runtime.evaluate(立即) 注入。
// 职责:在 .composer-footer 插麦克风按钮;点击录音→MediaRecorder 出 Blob→base64 经
// window.__voiceBridge(CDP binding) 回传主进程转写→回传文字→execCommand('insertText') 写回 composer。
// 写回手法参考 CodexPlusPlus 粘贴修复(renderer-inject.js 约第 8952 行):document.execCommand('insertText', false, text)。
// 注意:脚本体用普通模板字符串包裹;内部避免出现 ${(已无)与裸反引号(内部的模板字符串反引号需转义为 \`)。
const INJECT_SCRIPT = `
(() => {
const VER = "voice-inject-v6";
if (window.__voiceInjectInstalled === VER) return;
// 版本变了或首次:清理旧按钮/旧 observer,重新安装
if (window.__voiceInjectObserver) { try { window.__voiceInjectObserver.disconnect(); } catch {} }
const oldBtn = document.getElementById("codex-voice-input-btn");
if (oldBtn) oldBtn.remove();
window.__voiceInjectInstalled = VER;
const TAG = "[VoiceInput]";
const BTN_ID = "codex-voice-input-btn";
const STYLE_ID = "codex-voice-input-style-v6";
// —— 样式 ——
if (!document.getElementById(STYLE_ID)) {
const st = document.createElement("style");
st.id = STYLE_ID;
st.textContent = \`
.codex-voice-input-btn{
display:inline-flex;align-items:center;justify-content:center;flex:none;
width:32px;height:32px;padding:0;border:0;border-radius:999px;
background:rgba(17,24,39,.04);color:var(--text-secondary,#6b7280);
cursor:pointer;line-height:1;transition:background .15s,color .15s,box-shadow .15s,transform .15s;
-webkit-app-region:no-drag;
}
.codex-voice-input-btn:hover{
background:rgba(17,24,39,.08);color:var(--text-primary,#111827);
box-shadow:0 1px 4px rgba(15,23,42,.08);
}
.codex-voice-input-btn:active{transform:scale(.96);}
.codex-voice-input-btn:focus-visible{outline:2px solid rgba(59,130,246,.55);outline-offset:2px;}
.codex-voice-input-btn .cv-ico{width:17px;height:17px;display:block;}
.codex-voice-input-btn .cv-ico svg{width:100%;height:100%;display:block;stroke:currentColor;}
.codex-voice-input-btn[data-state="recording"]{
background:rgba(239,68,68,.12);color:#ef4444;
animation:codexVoicePulse 1.1s ease-in-out infinite;
}
.codex-voice-input-btn[data-state="transcribing"]{background:rgba(59,130,246,.1);color:#2563eb;cursor:wait;}
.codex-voice-input-btn[data-state="transcribing"] .cv-ico{animation:codexVoiceSpin .8s linear infinite;}
.codex-voice-input-btn[data-state="error"]{background:rgba(239,68,68,.1);color:#dc2626;}
@media (prefers-color-scheme:dark){
.codex-voice-input-btn{background:rgba(255,255,255,.05);color:var(--text-secondary,#a1a1aa);}
.codex-voice-input-btn:hover{background:rgba(255,255,255,.1);color:var(--text-primary,#f4f4f5);box-shadow:none;}
}
@keyframes codexVoicePulse{0%,100%{box-shadow:0 0 0 0 rgba(239,68,68,.42)}50%{box-shadow:0 0 0 7px rgba(239,68,68,0)}}
@keyframes codexVoiceSpin{to{transform:rotate(360deg)}}
.codex-voice-toast{
position:fixed;left:50%;bottom:96px;transform:translateX(-50%);
background:rgba(24,24,27,.96);color:#f3f4f6;padding:9px 16px;border-radius:10px;
font:13px system-ui,sans-serif;z-index:999999;box-shadow:0 8px 30px rgba(0,0,0,.5);
max-width:80vw;
}
\`;
(document.head || document.documentElement).appendChild(st);
}
function toast(msg, ms = 2400) {
const t = document.createElement("div");
t.className = "codex-voice-toast";
t.textContent = msg;
document.body.appendChild(t);
setTimeout(() => t.remove(), ms);
}
// —— binding 通道(window.__voiceBridge 由主进程 CDP addBinding 注入)——
const pending = new Map();
let seq = 0;
window.__voiceInjectHandle = (msg) => {
let p;
try {
p = typeof msg === "string" ? JSON.parse(msg) : msg;
} catch {
return;
}
const entry = pending.get(p.id);
if (!entry) return;
pending.delete(p.id);
if (p.err) entry.reject(new Error(p.err));
else entry.resolve(p.text || "");
};
function sendToBackend(audioB64, mime) {
const id = String(++seq);
return new Promise((resolve, reject) => {
pending.set(id, { resolve, reject });
if (typeof window.__voiceBridge !== "function") {
pending.delete(id);
reject(new Error("语音通道未就绪,请重新连接工具箱"));
return;
}
try {
window.__voiceBridge(JSON.stringify({ id, kind: "transcribe", audioB64, mime }));
} catch (e) {
pending.delete(id);
reject(e);
}
});
}
// —— 找底部输入工具栏:从编辑器往上找含 footer 的工具行容器 ——
function findToolbar() {
const rich = document.querySelector("DIV.ProseMirror, [contenteditable], textarea");
if (!rich) return null;
let p = rich;
for (let i = 0; i < 6 && p; i++) {
const foot = p.querySelector("[class*='_footer_'], [class*='composer-footer'], footer.composer, [class*='footer'][class*='composer']");
if (foot && foot.querySelectorAll("button, [role='button']").length >= 2) return foot;
p = p.parentElement;
}
return null;
}
function findEditor() {
return (
document.querySelector("DIV.ProseMirror") ||
document.querySelector("[contenteditable='true'],[contenteditable='']") ||
document.querySelector("textarea")
);
}
// —— 录音 ——
let mediaRecorder = null;
let chunks = [];
let curBtn = null;
let stopTimer = null;
function setBtnState(btn, st) {
if (!btn) return;
btn.dataset.state = st || "";
const label =
st === "recording" ? "停止录音" :
st === "transcribing" ? "转写中…" :
st === "error" ? "语音输入出错" :
"语音输入";
const icon =
st === "recording"
? '<svg viewBox="0 0 24 24" fill="none" stroke-width="2.4" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><rect x="8" y="8" width="8" height="8" rx="1.5"></rect></svg>'
: st === "transcribing"
? '<svg viewBox="0 0 24 24" fill="none" stroke-width="2.3" stroke-linecap="round" aria-hidden="true"><path d="M12 3a9 9 0 1 1-8.2 5.3"></path></svg>'
: '<svg viewBox="0 0 24 24" fill="none" stroke-width="2.2" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true"><path d="M12 3a3 3 0 0 0-3 3v5a3 3 0 0 0 6 0V6a3 3 0 0 0-3-3Z"></path><path d="M19 10v1a7 7 0 0 1-14 0v-1"></path><path d="M12 18v3"></path><path d="M8 21h8"></path></svg>';
btn.innerHTML = '<span class="cv-ico">' + icon + '</span>';
btn.title = label;
btn.setAttribute("aria-label", label);
}
async function startRecording(btn) {
if (!navigator.mediaDevices || !navigator.mediaDevices.getUserMedia) {
toast("此环境不支持麦克风");
setBtnState(btn, "error");
return;
}
let stream;
try {
stream = await navigator.mediaDevices.getUserMedia({ audio: true });
} catch (e) {
const name = (e && e.name) || "";
if (name === "NotAllowedError" || name === "SecurityError") {
toast("麦克风被拒绝:系统设置 → 隐私 → 麦克风 → 允许 ChatGPT", 4200);
} else {
toast("无法访问麦克风:" + (e && e.message ? e.message : name));
}
setBtnState(btn, "error");
return;
}
curBtn = btn;
chunks = [];
const mime = MediaRecorder.isTypeSupported("audio/webm")
? "audio/webm"
: MediaRecorder.isTypeSupported("audio/mp4")
? "audio/mp4"
: "";
try {
mediaRecorder = new MediaRecorder(stream, mime ? { mimeType: mime } : undefined);
} catch (e) {
toast("录音器初始化失败");
setBtnState(btn, "error");
stream.getTracks().forEach((t) => t.stop());
return;
}
mediaRecorder.ondataavailable = (e) => {
if (e.data && e.data.size) chunks.push(e.data);
};
mediaRecorder.onstop = () => {
stream.getTracks().forEach((t) => t.stop());
const blob = new Blob(chunks, { type: mime || "audio/webm" });
if (blob.size < 800) {
toast("未录到声音");
setBtnState(btn, "");
return;
}
setBtnState(btn, "transcribing");
const reader = new FileReader();
reader.onload = async () => {
const dataUrl = String(reader.result || "");
const comma = dataUrl.indexOf(",");
const b64 = comma >= 0 ? dataUrl.slice(comma + 1) : dataUrl;
try {
const text = await sendToBackend(b64, mime || "audio/webm");
if (text) writeBack(text);
else toast("转写结果为空");
} catch (err) {
toast("转写失败:" + (err && err.message ? err.message : err), 4200);
setBtnState(btn, "error");
return;
}
setBtnState(btn, "");
};
reader.onerror = () => {
toast("音频读取失败");
setBtnState(btn, "error");
};
reader.readAsDataURL(blob);
};
mediaRecorder.start();
setBtnState(btn, "recording");
stopTimer = setTimeout(() => stopRecording(), 60000);
}
function stopRecording() {
if (stopTimer) {
clearTimeout(stopTimer);
stopTimer = null;
}
if (mediaRecorder && mediaRecorder.state !== "inactive") {
try {
mediaRecorder.stop();
} catch {}
}
}
function writeBack(text) {
const ed = findEditor();
if (!ed) {
toast("未找到输入框,文字已无法填入", 3600);
return;
}
try {
ed.focus();
} catch {}
const inserted = document.execCommand("insertText", false, text);
if (!inserted) {
// 兜底:textarea / input 用 setRangeText
if (ed.tagName === "TEXTAREA" || ed.tagName === "INPUT") {
const s = ed.selectionStart || 0;
const e = ed.selectionEnd || 0;
const v = ed.value || "";
ed.value = v.slice(0, s) + text + v.slice(e);
const pos = s + text.length;
try {
ed.setSelectionRange(pos, pos);
} catch {}
ed.dispatchEvent(new Event("input", { bubbles: true }));
} else {
// contenteditable 兜底
try {
const sel = window.getSelection();
const range = sel && sel.rangeCount ? sel.getRangeAt(0) : null;
if (range) {
range.deleteContents();
range.insertNode(document.createTextNode(text));
}
} catch {}
}
}
}
function onBtnClick(btn) {
if (mediaRecorder && mediaRecorder.state === "recording") {
stopRecording();
return;
}
startRecording(btn);
}
function makeButton() {
const b = document.createElement("button");
b.type = "button";
b.id = BTN_ID;
b.className = "codex-voice-input-btn";
b.setAttribute("aria-label", "语音输入");
b.addEventListener("click", (e) => {
e.preventDefault();
e.stopPropagation();
onBtnClick(b);
});
setBtnState(b, "");
return b;
}
// 在工具栏里找「发送按钮」:默认取工具栏最后一个 button(Codex 里发送在最右)。
// 新版 Codex 会把模型选择和发送按钮包在右侧 action 容器里,语音按钮要挂到
// 发送按钮的直接父容器,才能落在「模型选择」与「发送」之间。
function findSendButton(toolbar) {
const btns = toolbar.querySelectorAll("button, [role='button']");
if (!btns || !btns.length) return null;
return btns[btns.length - 1];
}
function containsNode(parent, child) {
return !!(parent && child && (parent === child || parent.contains(child)));
}
let firstMount = true;
function tryMount() {
if (!firstMount && document.getElementById(BTN_ID)) return; // 已在,避免频繁重建丢焦点
firstMount = false;
const existing = document.getElementById(BTN_ID);
if (existing) existing.remove(); // 重新注入时刷新一次,确保新样式/位置
const toolbar = findToolbar();
if (!toolbar) { firstMount = true; return; } // 工具栏还没出来,下次再试
const sendBtn = findSendButton(toolbar);
const btn = makeButton();
const sendParent = sendBtn && sendBtn.parentElement;
if (sendParent && containsNode(toolbar, sendParent)) {
sendParent.insertBefore(btn, sendBtn); // 插在发送按钮前,保持在右侧 action 组内
} else {
toolbar.appendChild(btn);
}
}
// —— MutationObserver:composer/页面重建后重挂 ——
let mo = null;
function startObserver() {
if (mo) return;
mo = new MutationObserver(() => {
if (!document.getElementById(BTN_ID)) { firstMount = true; tryMount(); }
});
window.__voiceInjectObserver = mo; // 供重新安装时 disconnect
const root = document.body || document.documentElement;
if (root) mo.observe(root, { childList: true, subtree: true });
// 首次尝试
tryMount();
}
function ensureStarted() {
if (document.readyState === "loading") {
document.addEventListener("DOMContentLoaded", startObserver, { once: true });
} else {
startObserver();
}
}
ensureStarted();
})();
`;
module.exports = { INJECT_SCRIPT };