Files
heicode/cc-haha/adapters/telegram/index.ts
T
gongzhiyong 1efcf73992 chore: consolidate heicode workspace into single repository
Import cc-haha and new-api into a single repository layout for unified delivery.

Made-with: Cursor
2026-04-29 21:08:15 +08:00

755 lines
25 KiB
TypeScript

/**
* Telegram Adapter for Claude Code Desktop
*
* 基于 grammY 的轻量 Telegram Bot,直连服务端 /ws/:sessionId。
* 启动:TELEGRAM_BOT_TOKEN=xxx bun run telegram/index.ts
*/
import { Bot, InlineKeyboard, type Context } from 'grammy'
import * as path from 'node:path'
import { WsBridge, type ServerMessage } from '../common/ws-bridge.js'
import { MessageBuffer } from '../common/message-buffer.js'
import { MessageDedup } from '../common/message-dedup.js'
import { enqueue } from '../common/chat-queue.js'
import { loadConfig } from '../common/config.js'
import {
formatImHelp,
formatImStatus,
formatPermissionRequest,
splitMessage,
} from '../common/format.js'
import { SessionStore } from '../common/session-store.js'
import { AdapterHttpClient } from '../common/http-client.js'
import { isAllowedUser, tryPair } from '../common/pairing.js'
import { TelegramMediaService } from './media.js'
import { AttachmentStore } from '../common/attachment/attachment-store.js'
import { checkAttachmentLimit } from '../common/attachment/attachment-limits.js'
import type { AttachmentRef } from '../common/ws-bridge.js'
import { ImageBlockWatcher } from '../common/attachment/image-block-watcher.js'
import type { PendingUpload } from '../common/attachment/attachment-types.js'
import * as fs from 'node:fs/promises'
const TELEGRAM_TEXT_LIMIT = 4000 // leave margin below 4096
// ---------- init ----------
const config = loadConfig()
if (!config.telegram.botToken) {
console.error('[Telegram] Missing TELEGRAM_BOT_TOKEN. Set env or ~/.claude/adapters.json')
process.exit(1)
}
const bot = new Bot(config.telegram.botToken)
const bridge = new WsBridge(config.serverUrl, 'tg')
const dedup = new MessageDedup()
const sessionStore = new SessionStore()
const httpClient = new AdapterHttpClient(config.serverUrl)
const attachmentStore = new AttachmentStore()
const media = new TelegramMediaService(bot, attachmentStore)
attachmentStore.gc().catch((err) => {
console.warn('[Telegram] AttachmentStore.gc failed:', err instanceof Error ? err.message : err)
})
// Track placeholder messages for streaming updates
const placeholders = new Map<string, { chatId: string; messageId: number }>()
// Track accumulated text per chat for streaming
const accumulatedText = new Map<string, string>()
// Message buffers per chat
const buffers = new Map<string, MessageBuffer>()
// Track chats waiting for project selection
const pendingProjectSelection = new Map<string, boolean>()
const runtimeStates = new Map<string, ChatRuntimeState>()
/** Per-chat outbound image watcher for Agent-produced markdown images. */
const tgImageWatchers = new Map<string, ImageBlockWatcher>()
function getTgWatcher(chatId: string): ImageBlockWatcher {
let w = tgImageWatchers.get(chatId)
if (!w) {
w = new ImageBlockWatcher()
tgImageWatchers.set(chatId, w)
}
return w
}
type ChatRuntimeState = {
state: 'idle' | 'thinking' | 'streaming' | 'tool_executing' | 'permission_pending'
verb?: string
model?: string
pendingPermissionCount: number
}
// ---------- helpers ----------
function getBuffer(chatId: string): MessageBuffer {
let buf = buffers.get(chatId)
if (!buf) {
buf = new MessageBuffer(async (text, isComplete) => {
await flushToTelegram(chatId, text, isComplete)
})
buffers.set(chatId, buf)
}
return buf
}
function getRuntimeState(chatId: string): ChatRuntimeState {
let state = runtimeStates.get(chatId)
if (!state) {
state = { state: 'idle', pendingPermissionCount: 0 }
runtimeStates.set(chatId, state)
}
return state
}
function clearTransientChatState(chatId: string): void {
placeholders.delete(chatId)
accumulatedText.delete(chatId)
buffers.get(chatId)?.reset()
const runtime = getRuntimeState(chatId)
runtime.state = 'idle'
runtime.verb = undefined
runtime.pendingPermissionCount = 0
tgImageWatchers.delete(chatId)
}
async function ensureExistingSession(chatId: string): Promise<{ sessionId: string; workDir: string } | null> {
const stored = sessionStore.get(chatId)
if (!stored) return null
if (!bridge.hasSession(chatId)) {
bridge.connectSession(chatId, stored.sessionId)
bridge.onServerMessage(chatId, (msg) => handleServerMessage(chatId, msg))
const opened = await bridge.waitForOpen(chatId)
if (!opened) return null
}
return stored
}
async function buildStatusText(chatId: string): Promise<string> {
const stored = await ensureExistingSession(chatId)
if (!stored) return formatImStatus(null)
const runtime = getRuntimeState(chatId)
let projectName = path.basename(stored.workDir) || stored.workDir
let branch: string | null = null
try {
const gitInfo = await httpClient.getGitInfo(stored.sessionId)
projectName = gitInfo.repoName || path.basename(gitInfo.workDir) || projectName
branch = gitInfo.branch
} catch {
// Ignore git lookup failures and fall back to stored workDir
}
let taskCounts:
| {
total: number
pending: number
inProgress: number
completed: number
}
| undefined
try {
const tasks = await httpClient.getTasksForSession(stored.sessionId)
if (tasks.length > 0) {
taskCounts = {
total: tasks.length,
pending: tasks.filter((task) => task.status === 'pending').length,
inProgress: tasks.filter((task) => task.status === 'in_progress').length,
completed: tasks.filter((task) => task.status === 'completed').length,
}
}
} catch {
// Ignore task lookup failures in IM status summary
}
return formatImStatus({
sessionId: stored.sessionId,
projectName,
branch,
model: runtime.model,
state: runtime.state,
verb: runtime.verb,
pendingPermissionCount: runtime.pendingPermissionCount,
taskCounts,
})
}
async function flushToTelegram(chatId: string, newText: string, isComplete: boolean): Promise<void> {
const numericChatId = Number(chatId)
const prev = accumulatedText.get(chatId) ?? ''
const fullText = prev + newText
accumulatedText.set(chatId, fullText)
const placeholder = placeholders.get(chatId)
if (placeholder) {
if (isComplete) {
const chunks = splitMessage(fullText, TELEGRAM_TEXT_LIMIT)
try {
await bot.api.editMessageText(numericChatId, placeholder.messageId, chunks[0]!)
} catch { /* ignore */ }
for (let i = 1; i < chunks.length; i++) {
await bot.api.sendMessage(numericChatId, chunks[i]!)
}
} else {
const displayText = fullText.slice(0, TELEGRAM_TEXT_LIMIT - 2) + ' ▍'
try {
await bot.api.editMessageText(numericChatId, placeholder.messageId, displayText)
} catch { /* ignore */ }
}
} else if (isComplete && fullText.trim()) {
const chunks = splitMessage(fullText, TELEGRAM_TEXT_LIMIT)
for (const chunk of chunks) {
await bot.api.sendMessage(numericChatId, chunk)
}
}
if (isComplete) {
placeholders.delete(chatId)
accumulatedText.delete(chatId)
buffers.get(chatId)?.reset()
}
}
// ---------- session management ----------
async function ensureSession(chatId: string): Promise<boolean> {
if (bridge.hasSession(chatId)) return true
const stored = sessionStore.get(chatId)
if (stored) {
bridge.connectSession(chatId, stored.sessionId)
bridge.onServerMessage(chatId, (msg) => handleServerMessage(chatId, msg))
return await bridge.waitForOpen(chatId)
}
const workDir = config.defaultProjectDir
if (workDir) {
return await createSessionForChat(chatId, workDir)
}
await showProjectPicker(chatId)
return false
}
async function createSessionForChat(chatId: string, workDir: string): Promise<boolean> {
const numericChatId = Number(chatId)
try {
// Always tear down any stale WS connection before creating a new session.
// Without this, bridge.connectSession() below would short-circuit when an
// old OPEN connection still exists, leaving messages routed to the old session.
bridge.resetSession(chatId)
const sessionId = await httpClient.createSession(workDir)
sessionStore.set(chatId, sessionId, workDir)
bridge.connectSession(chatId, sessionId)
bridge.onServerMessage(chatId, (msg) => handleServerMessage(chatId, msg))
const opened = await bridge.waitForOpen(chatId)
if (!opened) {
await bot.api.sendMessage(numericChatId, '⚠️ 连接服务器超时,请重试。')
return false
}
return true
} catch (err) {
await bot.api.sendMessage(numericChatId,
`❌ 无法创建会话: ${err instanceof Error ? err.message : String(err)}`)
return false
}
}
async function showProjectPicker(chatId: string): Promise<void> {
const numericChatId = Number(chatId)
try {
const projects = await httpClient.listRecentProjects()
if (projects.length === 0) {
await bot.api.sendMessage(numericChatId,
'没有找到最近的项目。请先在 Desktop App 中打开一个项目,或在 Settings → IM 接入中配置默认项目。')
return
}
const lines = projects.slice(0, 10).map((p, i) =>
`${i + 1}. ${p.projectName}${p.branch ? ` (${p.branch})` : ''}\n ${p.realPath}`
)
pendingProjectSelection.set(chatId, true)
await bot.api.sendMessage(numericChatId,
`选择项目(回复编号):\n\n${lines.join('\n\n')}\n\n💡 下次可直接 /new <编号或名称> 快速新建会话`)
} catch (err) {
await bot.api.sendMessage(numericChatId,
`❌ 无法获取项目列表: ${err instanceof Error ? err.message : String(err)}`)
}
}
// ---------- outbound media dispatch ----------
/** Upload a PendingUpload found in streaming output and send it via
* bot.api.sendPhoto as an independent message. Runs fire-and-forget
* from the stream handler so streaming text isn't blocked. */
async function dispatchOutboundMedia(chatId: string, pending: PendingUpload): Promise<void> {
const numericChatId = Number(chatId)
try {
let buffer: Buffer
let mime = 'image/png'
switch (pending.source.kind) {
case 'base64': {
buffer = Buffer.from(pending.source.data, 'base64')
mime = pending.source.mime
break
}
case 'path': {
buffer = await fs.readFile(pending.source.path)
mime = pending.source.mime ?? 'image/png'
break
}
case 'url': {
const resp = await fetch(pending.source.url)
if (!resp.ok) {
throw new Error(`fetch ${pending.source.url} -> ${resp.status}`)
}
buffer = Buffer.from(await resp.arrayBuffer())
mime = pending.source.mime ?? resp.headers.get('content-type') ?? 'image/png'
break
}
}
const check = checkAttachmentLimit('image', buffer.length, mime)
if (!check.ok) {
console.warn('[Telegram] Outbound image rejected:', check.hint)
return
}
await media.sendPhoto(numericChatId, buffer, pending.alt)
} catch (err) {
console.error(
'[Telegram] dispatchOutboundMedia failed:',
err instanceof Error ? err.message : err,
)
}
}
// ---------- server message handler ----------
async function handleServerMessage(chatId: string, msg: ServerMessage): Promise<void> {
const numericChatId = Number(chatId)
const buf = getBuffer(chatId)
const runtime = getRuntimeState(chatId)
switch (msg.type) {
case 'connected':
break
case 'status':
runtime.state = msg.state
runtime.verb = typeof msg.verb === 'string' ? msg.verb : undefined
if (msg.state === 'thinking' && !placeholders.has(chatId)) {
const sent = await bot.api.sendMessage(numericChatId, '💭 思考中...')
placeholders.set(chatId, { chatId, messageId: sent.message_id })
accumulatedText.set(chatId, '')
}
break
case 'content_start':
if (msg.blockType === 'text') {
if (!placeholders.has(chatId)) {
const sent = await bot.api.sendMessage(numericChatId, '▍')
placeholders.set(chatId, { chatId, messageId: sent.message_id })
accumulatedText.set(chatId, '')
}
} else if (msg.blockType === 'tool_use') {
// Finalize current text placeholder before tool calls,
// so text after tools gets a fresh message
await buf.complete()
// If placeholder still exists (buffer was already empty), clean up directly
if (placeholders.has(chatId)) {
const text = accumulatedText.get(chatId)
if (text?.trim()) {
try {
await bot.api.editMessageText(numericChatId, placeholders.get(chatId)!.messageId, text)
} catch { /* ignore */ }
}
placeholders.delete(chatId)
accumulatedText.delete(chatId)
buffers.get(chatId)?.reset()
}
}
break
case 'content_delta':
if (msg.text) {
buf.append(msg.text)
const newUploads = getTgWatcher(chatId).feed(msg.text)
for (const pending of newUploads) {
void dispatchOutboundMedia(chatId, pending)
}
}
break
case 'thinking':
if (placeholders.has(chatId)) {
try {
await bot.api.editMessageText(
numericChatId,
placeholders.get(chatId)!.messageId,
`💭 ${msg.text.slice(0, 200)}...`,
)
} catch { /* ignore */ }
}
break
case 'tool_use_complete':
// Tool details are noise for IM users; visible in Desktop if needed.
break
case 'tool_result':
// Tool errors are handled internally by the AI (retries etc.)
// No need to notify the user for every failed attempt.
break
case 'permission_request': {
runtime.pendingPermissionCount += 1
runtime.state = 'permission_pending'
const text = formatPermissionRequest(msg.toolName, msg.input, msg.requestId)
const keyboard = new InlineKeyboard()
.text('✅ 允许', `permit:${msg.requestId}:yes`)
.text('❌ 拒绝', `permit:${msg.requestId}:no`)
await bot.api.sendMessage(numericChatId, text, { reply_markup: keyboard })
break
}
case 'message_complete':
runtime.state = 'idle'
runtime.verb = undefined
await buf.complete()
// Ensure placeholder is always cleaned up even if buffer was already empty
if (placeholders.has(chatId)) {
const text = accumulatedText.get(chatId)
if (text?.trim()) {
try {
const chunks = splitMessage(text, TELEGRAM_TEXT_LIMIT)
await bot.api.editMessageText(numericChatId, placeholders.get(chatId)!.messageId, chunks[0]!)
for (let i = 1; i < chunks.length; i++) {
await bot.api.sendMessage(numericChatId, chunks[i]!)
}
} catch { /* ignore */ }
}
placeholders.delete(chatId)
accumulatedText.delete(chatId)
buffers.get(chatId)?.reset()
}
break
case 'error':
runtime.state = 'idle'
runtime.verb = undefined
// Auto-recover from stale thinking block signatures by creating a fresh session.
// This happens when the API key or provider changed since the session was created.
if (msg.message && /Invalid.*signature.*thinking/i.test(msg.message)) {
const stored = sessionStore.get(chatId)
const workDir = stored?.workDir || config.defaultProjectDir
if (workDir) {
await bot.api.sendMessage(numericChatId, '⚠️ 会话上下文已失效,正在自动重建...')
clearTransientChatState(chatId)
bridge.resetSession(chatId)
sessionStore.delete(chatId)
const ok = await createSessionForChat(chatId, workDir)
if (ok) {
await bot.api.sendMessage(numericChatId, '✅ 已重建会话,请重新发送消息。')
} else {
await bot.api.sendMessage(numericChatId, '❌ 重建会话失败,请发送 /new 手动新建。')
}
} else {
await bot.api.sendMessage(numericChatId, '⚠️ 会话上下文已失效,请发送 /new 新建会话。')
}
} else {
await bot.api.sendMessage(numericChatId, `❌ ${msg.message}`)
}
break
case 'system_notification':
if (msg.subtype === 'init' && msg.data && typeof msg.data === 'object') {
const model = (msg.data as Record<string, unknown>).model
if (typeof model === 'string' && model.trim()) {
runtime.model = model
}
}
break
}
}
// ---------- bot handlers ----------
async function sendHelp(ctx: Context): Promise<void> {
await ctx.reply(`👋 Claude Code Bot 已就绪。\n\n${formatImHelp()}`)
}
bot.command('start', (ctx) => void sendHelp(ctx))
bot.command('help', (ctx) => void sendHelp(ctx))
/** Reset session state and start a new session for chatId.
* If `query` is provided, match a project by index or name;
* otherwise use defaultProjectDir or show the picker. */
async function startNewSession(chatId: string, query?: string): Promise<void> {
const numericChatId = Number(chatId)
bridge.resetSession(chatId)
sessionStore.delete(chatId)
placeholders.delete(chatId)
accumulatedText.delete(chatId)
buffers.get(chatId)?.reset()
buffers.delete(chatId)
pendingProjectSelection.delete(chatId)
runtimeStates.delete(chatId)
tgImageWatchers.delete(chatId)
if (query) {
try {
const { project, ambiguous } = await httpClient.matchProject(query)
if (project) {
const ok = await createSessionForChat(chatId, project.realPath)
if (ok) {
await bot.api.sendMessage(numericChatId,
`✅ 已新建会话:${project.projectName}${project.branch ? ` (${project.branch})` : ''}`)
}
return
}
if (ambiguous) {
const list = ambiguous.map((p, i) => `${i + 1}. ${p.projectName} — ${p.realPath}`).join('\n')
await bot.api.sendMessage(numericChatId, `匹配到多个项目,请更精确:\n\n${list}`)
return
}
await bot.api.sendMessage(numericChatId, `未找到匹配 "${query}" 的项目。发送 /projects 查看完整列表。`)
} catch (err) {
await bot.api.sendMessage(numericChatId,
`❌ ${err instanceof Error ? err.message : String(err)}`)
}
} else {
const workDir = config.defaultProjectDir
if (workDir) {
const ok = await createSessionForChat(chatId, workDir)
if (ok) {
await bot.api.sendMessage(numericChatId, '✅ 已新建会话,可以开始对话了。')
}
} else {
await showProjectPicker(chatId)
}
}
}
bot.command('new', async (ctx) => {
const chatId = String(ctx.chat.id)
await startNewSession(chatId, ctx.match?.trim() || undefined)
})
bot.command('projects', async (ctx) => {
const chatId = String(ctx.chat.id)
await showProjectPicker(chatId)
})
bot.command('stop', (ctx) => {
const chatId = String(ctx.chat.id)
void (async () => {
const stored = await ensureExistingSession(chatId)
if (!stored) {
await ctx.reply(formatImStatus(null))
return
}
bridge.sendStopGeneration(chatId)
await ctx.reply('⏹ 已发送停止信号。')
})()
})
bot.command('status', async (ctx) => {
const chatId = String(ctx.chat.id)
await ctx.reply(await buildStatusText(chatId))
})
bot.command('clear', (ctx) => {
const chatId = String(ctx.chat.id)
void (async () => {
const stored = await ensureExistingSession(chatId)
if (!stored) {
await ctx.reply(formatImStatus(null))
return
}
clearTransientChatState(chatId)
const sent = bridge.sendUserMessage(chatId, '/clear')
if (!sent) {
await ctx.reply('⚠️ 无法发送 /clear,请先发送 /new 重新连接会话。')
return
}
await ctx.reply('🧹 已清空当前会话上下文。')
})()
})
/** Shared per-user-message pipeline: dedup, pairing check, project-pick
* routing, enqueue, ensureSession, sendUserMessage with attachments.
* Caller has already extracted text and attachments from the context. */
async function routeUserMessage(
ctx: Context,
text: string,
attachments: AttachmentRef[],
): Promise<void> {
if (!ctx.from || ctx.chat?.type !== 'private') return
if (!dedup.tryRecord(String(ctx.message?.message_id))) return
const chatId = String(ctx.chat.id)
const userId = ctx.from.id
if (!isAllowedUser('telegram', userId)) {
const displayName = [ctx.from.first_name, ctx.from.last_name].filter(Boolean).join(' ')
const success = tryPair(text.trim(), { userId, displayName }, 'telegram')
if (success) {
await ctx.reply('✅ 配对成功!现在可以开始聊天了。\n\n发送消息即可与 Claude 对话。')
} else {
await ctx.reply('🔒 未授权。请在 Claude Code 桌面端生成配对码后发送给我。')
}
return
}
enqueue(chatId, async () => {
if (pendingProjectSelection.has(chatId)) {
if (text.trim()) await startNewSession(chatId, text.trim())
return
}
const ready = await ensureSession(chatId)
if (!ready) return
const effective =
text || (attachments.length > 0 ? '(用户发送了附件)' : '')
if (!effective && attachments.length === 0) return
const sent = bridge.sendUserMessage(chatId, effective, attachments.length ? attachments : undefined)
if (!sent) {
await bot.api.sendMessage(Number(chatId), '⚠️ 消息发送失败,连接可能已断开。请发送 /new 重新开始。')
}
})
}
/** Scan ctx.message for photo/document/video/audio/voice, download
* each via TelegramMediaService, apply size/mime limits, and produce
* a ready-to-send AttachmentRef[] plus any rejection hints. */
async function collectAttachmentsFromCtx(
ctx: Context,
): Promise<{ attachments: AttachmentRef[]; rejections: string[] }> {
const msg = ctx.message
if (!msg || !ctx.chat) return { attachments: [], rejections: [] }
const sessionId = sessionStore.get(String(ctx.chat.id))?.sessionId ?? String(ctx.chat.id)
const attachments: AttachmentRef[] = []
const rejections: string[] = []
const runOne = async (
fileId: string,
fileName?: string,
mimeType?: string,
): Promise<void> => {
try {
const local = await media.downloadFile(fileId, sessionId, { fileName, mimeType })
const check = checkAttachmentLimit(local.kind, local.size, local.mimeType)
if (!check.ok) {
rejections.push(check.hint)
return
}
if (local.kind === 'image') {
attachments.push({
type: 'image',
name: local.name,
data: local.buffer.toString('base64'),
mimeType: local.mimeType,
})
} else {
attachments.push({
type: 'file',
name: local.name,
path: local.path,
mimeType: local.mimeType,
})
}
} catch (err) {
console.error('[Telegram] downloadFile failed:', err)
rejections.push('📎 附件下载失败,请稍后重试')
}
}
// Photos: grammY exposes an array of sizes, largest last.
if (msg.photo && msg.photo.length > 0) {
const largest = msg.photo[msg.photo.length - 1]!
await runOne(largest.file_id, `photo-${largest.file_unique_id}.jpg`, 'image/jpeg')
}
if (msg.document) {
await runOne(msg.document.file_id, msg.document.file_name, msg.document.mime_type)
}
if (msg.video) {
await runOne(msg.video.file_id, msg.video.file_name, msg.video.mime_type)
}
if (msg.audio) {
await runOne(msg.audio.file_id, msg.audio.file_name, msg.audio.mime_type)
}
if (msg.voice) {
await runOne(
msg.voice.file_id,
`voice-${msg.voice.file_unique_id}.ogg`,
msg.voice.mime_type ?? 'audio/ogg',
)
}
return { attachments, rejections }
}
bot.on('message:text', async (ctx) => {
await routeUserMessage(ctx, ctx.message.text, [])
})
bot.on(
['message:photo', 'message:document', 'message:video', 'message:audio', 'message:voice'],
async (ctx) => {
const caption = ctx.message.caption ?? ''
const { attachments, rejections } = await collectAttachmentsFromCtx(ctx)
for (const r of rejections) {
await ctx.reply(r).catch(() => {})
}
if (attachments.length === 0 && !caption.trim()) return
await routeUserMessage(ctx, caption, attachments)
},
)
bot.on('callback_query:data', async (ctx) => {
const data = ctx.callbackQuery.data
if (!data.startsWith('permit:')) return
const parts = data.split(':')
if (parts.length !== 3) return
const requestId = parts[1]!
const allowed = parts[2] === 'yes'
const chatId = String(ctx.callbackQuery.message?.chat.id)
bridge.sendPermissionResponse(chatId, requestId, allowed)
const runtime = getRuntimeState(chatId)
runtime.pendingPermissionCount = Math.max(0, runtime.pendingPermissionCount - 1)
const statusText = allowed ? '✅ 已允许' : '❌ 已拒绝'
try {
await ctx.editMessageText(
ctx.callbackQuery.message?.text + `\n\n${statusText}`,
)
} catch { /* ignore */ }
await ctx.answerCallbackQuery(statusText)
})
// ---------- start ----------
console.log('[Telegram] Starting bot...')
console.log(`[Telegram] Server: ${config.serverUrl}`)
console.log(`[Telegram] Allowed users: ${config.telegram.allowedUsers.length === 0 ? 'all' : config.telegram.allowedUsers.join(', ')}`)
bot.start({
onStart: () => console.log('[Telegram] Bot is running!'),
})
// Graceful shutdown
process.on('SIGINT', () => {
console.log('[Telegram] Shutting down...')
bot.stop()
bridge.destroy()
dedup.destroy()
process.exit(0)
})