fix(ask): prototype 拦截器替代 waterfall hook + TG 原生按钮作答

根因:dsh 0.1.5-rc.1 把 'user-questions/request' waterfall 分发到
per-fiber events 注册表,其他 fiber 的 ctx.on 监听(含本插件)永远收不到
分发(实测 root/global/agent 作用域监听全部落空)。

修复:包装 UserQuestionService.prototype.ask(类跨实例共享),无论哪个
fiber/实例发起的 ask 都能拦截到 agent+questions;TG 答案与原 ask 竞速,
保留 dual-path 语义;stop 时恢复原型,热重载不叠加。

功能(应用户要求):
- 提问前显式发送预告文本消息
- 选项改用 TG 原生 inline keyboard 按钮(单选即点即答,多选带提交键)
- 文本作答保留并与按钮作答合并,逐题累计、齐题自动提交
- 附带修复:Telegram API 调用加 15s 客户端超时(代理挂死会卡死通知队列)

Co-authored-by: Shirobana Agent Bot <tg-bridge>
This commit is contained in:
root
2026-09-17 11:15:16 +08:00
co-authored by Shirobana Agent Bot <tg-bridge>
parent 4039dec03c
commit da06485d64
8 changed files with 547 additions and 152 deletions
+240 -70
View File
@@ -6,7 +6,7 @@ import { dirname, join } from 'node:path'
import type { Context } from '@deepseek-ai/cordis'
import type { Agent } from '@deepseek-ai/dsh-agent'
import { createUserMessage } from '@deepseek-ai/dsh-llm/message'
import type { ContentBlock, TextBlock } from '@deepseek-ai/dsh-llm/types'
import type { ContentBlock } from '@deepseek-ai/dsh-llm/types'
import { SessionId, type SessionEvent } from '@deepseek-ai/dsh-session/types'
import { isAuthorized } from './auth.js'
import { resolveApiProxy } from './apiproxy.js'
@@ -28,7 +28,7 @@ import {
type TelegramUpdate,
} from './client.js'
import { MSG, LAST_CB, parseCommand } from './commands.js'
import { markdownToHtml, splitMessage } from './format.js'
import { escapeHtml, markdownToHtml, splitMessage } from './format.js'
import { splitRichMarkdown } from './rich-format.js'
import { formatLastTurn, loadLastTurn } from './history.js'
import { describeAgent, displayLabel, workspaceName } from './label.js'
@@ -97,6 +97,10 @@ interface TgPendingAsk {
questions: TgQuestionItem[]
ask: TgAnswerHandle
at: number
/** question index → option indices chosen via inline-keyboard taps. */
kbSel: Map<number, Set<number>>
/** question index → answer parsed from a text reply (letters or custom). */
textAnswers: Map<number, { id?: string; selected: string[]; custom?: string }>
}
/** chatId → pending approval request mirrored to Telegram. */
@@ -105,15 +109,6 @@ interface TgPendingApproval {
ask: TgAnswerHandle
}
/** Structural view of the user-questions seam (types live in the host's dsh-user-questions package). */
interface UserQuestionsServiceLike {
provider?: {
ask: (request: unknown) => Promise<unknown>
__tgHooked?: boolean
__tgRealAsk?: (request: unknown) => Promise<unknown>
}
}
/** Loose view of the approval request passed through the `approval/request` waterfall hook. */
interface ApprovalRequestLike {
agent?: Agent
@@ -194,7 +189,11 @@ const MODEL_CB = 'mdl:'
/** Use eff: (not me:) — short prefix, no collision with other callbacks. */
const EFFORT_CB = 'eff:'
const BACK_MODEL_CB = 'mb'
/** ask:{questionIndex}:{optionIndex|ok} — inline-keyboard taps for ask_user_question. */
const ASK_CB = 'ask:'
const MAX_BUTTONS = 40
/** Marker stored on UserQuestionService.prototype for the ask interceptor. */
const kTgAskWrapped = Symbol.for('dsh-telegram-channel.askWrapped')
function lastContextKeyboard(): InlineKeyboardMarkup {
return {
@@ -204,11 +203,28 @@ function lastContextKeyboard(): InlineKeyboardMarkup {
const defaultSleep = (ms: number) => new Promise<void>((resolve) => setTimeout(resolve, ms))
/** Cap a single reasoning block so a long chain of thought can't flood the phone. */
const REASONING_QUOTE_MAX_CHARS = 800
/**
* Assistant message → markdown for delivery. Reasoning (thinking) blocks are
* kept, one `> ` quote line each — Telegram renders them as the accent-bar
* quote style the phone user expects for bash/thinking-style chatter.
*/
function contentToText(content: readonly ContentBlock[]): string {
return content
.filter((block): block is TextBlock => block.type === 'text')
.map((block) => block.text)
.join('')
const parts: string[] = []
for (const block of content) {
if (block.type === 'text' && typeof block.text === 'string') {
parts.push(block.text)
} else if (block.type === 'reasoning' && typeof block.text === 'string') {
let thinking = block.text.trim()
if (!thinking) continue
if (thinking.length > REASONING_QUOTE_MAX_CHARS)
thinking = `${thinking.slice(0, REASONING_QUOTE_MAX_CHARS)}…`
parts.push(thinking.split('\n').map((line) => `> ${line}`).join('\n'))
}
}
return parts.join('\n\n')
}
function isRichUnsupportedError(err: unknown): boolean {
@@ -252,8 +268,9 @@ export class TelegramBridge {
private hookTimer: ReturnType<typeof setTimeout> | undefined
private readonly pendingApprovalsTG = new Map<string, TgPendingApproval>()
private disposeApprovalHook: (() => void) | undefined
/** dsh 0.1.5 起 ask 走 'user-questions/request' waterfall(不再有 userQuestions.provider)。 */
private disposeUserQuestionHook: (() => void) | undefined
/** Prototype-ask interceptor state (see installAskInterceptor). */
private askInstalled = false
private askRestore: (() => void) | undefined
/** callId → tool name (tool/result failure notices) */
private readonly callNames = new Map<string, string>()
/** sessionId → latest todo snapshot (/mission) */
@@ -295,10 +312,12 @@ export class TelegramBridge {
// Restore persisted bindings (survive hot reload / restart) and install
// the TG answering hooks (ask_user dual-path + approval dual-path).
this.loadBindings()
this.disposeUserQuestionHook?.()
this.disposeUserQuestionHook = (this.ctx as unknown as {
on: (event: string, handler: (req: unknown, next: () => Promise<unknown>) => Promise<unknown>) => () => void
}).on('user-questions/request', (req, next) => this.onUserQuestionRequest(req, next))
// ask_user interception: see installAskInterceptor — a prototype wrap of
// UserQuestionService.prototype.ask. Plain ctx.on('user-questions/request')
// registries never see the dispatch (the waterfall fires on a different
// per-fiber events registry in dsh 0.1.5-rc.1), so the prototype wrap —
// which sees EVERY ask regardless of fiber/instance — is the reliable seam.
this.installAskInterceptor()
this.disposeApprovalHook?.()
this.disposeApprovalHook = (this.ctx as unknown as {
on: (event: string, handler: (req: ApprovalRequestLike, next: () => Promise<string>) => Promise<string>) => () => void
@@ -357,8 +376,9 @@ export class TelegramBridge {
clearTimeout(this.hookTimer)
this.hookTimer = undefined
}
this.disposeUserQuestionHook?.()
this.disposeUserQuestionHook = undefined
this.askRestore?.()
this.askRestore = undefined
this.askInstalled = false
this.disposeApprovalHook?.()
this.disposeApprovalHook = undefined
if (this.pollPromise) {
@@ -492,6 +512,10 @@ export class TelegramBridge {
const data = cq.data ?? ''
const picker = this.pickers.get(String(chatId))
if (data.startsWith(ASK_CB)) {
await this.handleAskCallback(chatId, cq)
return
}
if (data === LAST_CB) {
await this.client.answerCallbackQuery(cq.id)
await this.sendLastTurn(chatId)
@@ -1259,6 +1283,7 @@ export class TelegramBridge {
}
private async onSessionEvent(session: SessionLike, event: SessionEvent): Promise<void> {
if (!this.askInstalled) this.installAskInterceptor()
const id = String(session.id)
const targets = [...this.bindings.values()].filter((b) => b.sessionId === id)
if (targets.length === 0) return
@@ -1458,17 +1483,28 @@ export class TelegramBridge {
}
/**
* Send a notice; a leading `> ` keeps the quote look via a plain-text `> `
* prefix. NB: NOT `<blockquote>` HTML — older Telegram clients cannot render
* the blockquote entity at all and show the whole message as “not supported”.
* Falls back to plain text when the send fails — a notice is never lost.
* Send a notice; a leading `> ` marks the body for quote rendering as a real
* `<blockquote>` (parse_mode HTML) — every current Telegram client draws the
* accent-colored left bar. Old-client compatibility is no longer required,
* but a failed HTML send still falls back to plain text — a notice is never lost.
*/
private async deliverNotice(chatId: number, text: string): Promise<void> {
const quoted = text.startsWith('> ')
const body = quoted ? text.slice(2) : text
if (quoted) {
try {
await this.withRetry(() =>
this.client.sendMessage(chatId, `<blockquote>${escapeHtml(body)}</blockquote>`, 'HTML'),
)
return
} catch (err) {
this.ctx.logger.warn(
`dsh-telegram-channel: blockquote notice failed (${this.redact(err)}), falling back to plain text`,
)
}
}
try {
await this.withRetry(() => this.client.sendMessage(chatId, quoted ? `> ${body}` : body))
return
await this.withRetry(() => this.client.sendMessage(chatId, body))
} catch (err) {
this.ctx.logger.error(this.redact(err))
}
@@ -1505,28 +1541,64 @@ export class TelegramBridge {
this.heartbeats.clear()
}
// ── ask_user_question: TG answering via dual-path race with the UI ──
// ── ask_user_question: TG answering via a prototype interceptor ──
//
// dsh 0.1.5-rc.1 routes the 'user-questions/request' waterfall through a
// per-fiber events registry: plain ctx.on listeners in other fibers (this
// plugin included) never see the dispatch, so the waterfall hook cannot
// mirror questions to TG. The class is shared across instances though, so
// wrapping UserQuestionService.prototype.ask intercepts EVERY ask with its
// agent + questions, regardless of fiber. The wrap races the TG answer
// against the original ask (GUI path), preserving the old dual-path
// semantics. Restore-on-stop keeps hot reloads from stacking wrappers.
private userQuestions(): UserQuestionsServiceLike | undefined {
return (this.ctx as unknown as { userQuestions?: UserQuestionsServiceLike }).userQuestions
private installAskInterceptor(): void {
if (this.askInstalled) return
let svc: { constructor: { prototype: Record<string | symbol, unknown> } } | undefined
try {
svc = (this.ctx as unknown as { get: (name: string) => unknown }).get('userQuestions') as
| { constructor: { prototype: Record<string | symbol, unknown> } }
| undefined
} catch {
return // service not ready yet — retried lazily from onSessionEvent
}
if (!svc) return
const proto = svc.constructor.prototype
const prev = proto[kTgAskWrapped] as { restore: () => void; bridge: unknown } | undefined
if (prev?.bridge === this) {
this.askInstalled = true
this.askRestore = prev.restore
return
}
prev?.restore() // stale wrapper from a disposed bridge instance
const orig = proto.ask as (this: unknown, request: unknown) => Promise<unknown>
const bridge = this
const wrapped = function (this: unknown, request: unknown): Promise<unknown> {
return bridge.aroundUserQuestionAsk(this, request, orig)
}
proto.ask = wrapped
const restore = (): void => {
if (proto.ask === wrapped) proto.ask = orig
}
proto[kTgAskWrapped] = { restore, bridge: this }
this.askRestore = restore
this.askInstalled = true
}
/**
* dsh 0.1.5 移除了 `userQuestions.provider`:ask 现在走 'user-questions/request'
* waterfall(与 approval/request 同形),启动时注册一次即可,无需轮询 provider。
* 让 TG 答案与 next()(UI 转发)赛跑;未绑定聊天时 TG promise 永不 settle,
* 于是 race 完全跟随 next()。
*/
private onUserQuestionRequest(req: unknown, next: () => Promise<unknown>): Promise<unknown> {
const tg = this.registerTgAsk(req)
private aroundUserQuestionAsk(
service: unknown,
request: unknown,
orig: (this: unknown, request: unknown) => Promise<unknown>,
): Promise<unknown> {
const tg = this.registerTgAsk(request)
let gui: Promise<unknown>
try {
gui = Promise.resolve().then(next)
gui = Promise.resolve().then(() => orig.call(service, request))
} catch (err) {
tg.reject(err as Error)
throw err
}
void gui.then(() => this.settleGuiSide(req), () => this.settleGuiSide(req))
void gui.then(() => this.settleGuiSide(request), () => this.settleGuiSide(request))
return Promise.race([tg.promise, gui])
}
@@ -1556,16 +1628,88 @@ export class TelegramBridge {
let asked = false
for (const [chatId, binding] of this.bindings) {
if (binding.sessionId !== sessionId) continue
this.pendingAsks.set(String(chatId), { req: request, sessionId, questions, ask, at: Date.now() })
this.enqueueNotice(Number(chatId), this.formatAskPending(questions))
const pending: TgPendingAsk = {
req: request,
sessionId,
questions,
ask,
at: Date.now(),
kbSel: new Map(),
textAnswers: new Map(),
}
this.pendingAsks.set(String(chatId), pending)
void this.deliverAskPrompt(Number(chatId), pending)
asked = true
}
if (asked) return { promise, reject: rejectFn }
}
// No bound TG chat: never settle — the UI remains the only answer path.
return { promise, reject: rejectFn }
}
/** Announce the incoming question, then one message per question (options as native inline-keyboard buttons). */
private async deliverAskPrompt(chatId: number, pending: TgPendingAsk): Promise<void> {
const { questions } = pending
try {
await this.withRetry(() =>
this.client.sendMessage(
chatId,
`❓ 即将向你提问(共 ${questions.length} 题)——点按钮或回复文本作答;/cancel 关闭 TG 作答`,
),
)
for (let i = 0; i < questions.length; i++) {
const q = questions[i]!
const opts = Array.isArray(q.options) ? q.options : []
const header = questions.length > 1 ? `Q${i + 1}. ` : ''
if (opts.length === 0) {
await this.withRetry(() =>
this.client.sendMessage(chatId, `${header}${q.question ?? ''}\n(开放问题,请直接回复文本)`),
)
continue
}
const keyboard: InlineKeyboardMarkup = {
inline_keyboard: [
...opts.map((o, j) => [{
text: `${j + 1}. ${o.label}`.slice(0, 48),
callback_data: `${ASK_CB}${i}:${j}`,
}]),
...(q.multiSelect ? [[{ text: '✅ 提交所选', callback_data: `${ASK_CB}${i}:ok` }]] : []),
],
}
await this.withRetry(() => this.client.sendMessage(chatId, `${header}${q.question ?? ''}`, undefined, keyboard))
}
} catch (err) {
this.ctx.logger.error(this.redact(err))
}
}
/** Resolve the pending ask once EVERY question has an answer (keyboard and/or text). */
private finalizeAskIfComplete(chatId: number, pending: TgPendingAsk): boolean {
const { questions, ask } = pending
const answers: Array<{ id?: string; selected: string[]; custom?: string }> = []
for (let i = 0; i < questions.length; i++) {
const q = questions[i]!
const opts = Array.isArray(q.options) ? q.options : []
const sel = pending.kbSel.get(i)
if (sel && sel.size > 0) {
answers.push({ id: q.id, selected: [...sel].sort((a, b) => a - b).map((j) => opts[j]!.label) })
continue
}
const text = pending.textAnswers.get(i)
if (text !== undefined) {
answers.push(text)
continue
}
return false
}
ask.answered = true
for (const [cid, p] of this.pendingAsks) {
if (p.ask === ask) this.pendingAsks.delete(cid)
}
ask.resolve({ answers })
this.enqueueNotice(chatId, '已提交你的回答')
return true
}
private settleGuiSide(request: unknown): void {
for (const [chatId, p] of [...this.pendingAsks]) {
if (p.req !== request) continue
@@ -1576,24 +1720,8 @@ export class TelegramBridge {
}
}
private formatAskPending(questions: TgQuestionItem[]): string {
const lines = ['待回答(回复字母选项,或直接输入自定义答案;/cancel 关闭 TG 作答)']
questions.forEach((q, i) => {
const prefix = questions.length > 1 ? `Q${i + 1}. ` : ''
lines.push(`${prefix}${q.question ?? ''}`)
const opts = Array.isArray(q.options) ? q.options : []
if (opts.length === 0) {
lines.push(' (开放问题,直接输入答案)')
} else {
opts.forEach((o, j) => lines.push(` [${String.fromCharCode(65 + j)}] ${o.label}${o.description ? ` — ${o.description}` : ''}`))
if (q.multiSelect) lines.push(' (可多选,如:A,B)')
}
})
return lines.join('\n')
}
private async handleTgAnswer(chatId: number, pending: TgPendingAsk, text: string): Promise<void> {
const { questions, ask } = pending
const { questions } = pending
let parts: Array<[number | null, string]>
if (questions.length === 1) {
parts = [[0, text.trim()]]
@@ -1610,7 +1738,6 @@ export class TelegramBridge {
return
}
}
const answers: Array<{ id?: string; selected: string[]; custom?: string }> = []
for (const [idx, raw] of parts) {
if (idx === null || idx < 0 || idx >= questions.length) {
await this.enqueueNotice(chatId, `题号 ${(idx ?? 0) + 1} 不存在(共 ${questions.length} 题)`)
@@ -1633,17 +1760,60 @@ export class TelegramBridge {
}
selected.push(opts[i]!.label)
}
answers.push({ id: q.id, selected })
pending.textAnswers.set(idx, { id: q.id, selected })
} else {
answers.push({ id: q.id, selected: [], custom: cleaned })
pending.textAnswers.set(idx, { id: q.id, selected: [], custom: cleaned })
}
}
ask.answered = true
for (const [cid, p] of this.pendingAsks) {
if (p.ask === ask) this.pendingAsks.delete(cid)
// Keyboard selections take precedence per question; resolve when complete.
if (!this.finalizeAskIfComplete(chatId, pending)) {
const done = questions.filter((_, i) => pending.kbSel.get(i)?.size || pending.textAnswers.has(i)).length
await this.enqueueNotice(chatId, `已记录 ${done}/${questions.length} 题,请继续作答(按钮或 1:/2: 文本)`)
}
}
/** Inline-keyboard tap on an ask message: toggle/record the choice, resolve when every question is answered. */
private async handleAskCallback(chatId: number, cq: { id: string; data?: string }): Promise<void> {
const pending = this.pendingAsks.get(String(chatId))
if (!pending) {
await this.client.answerCallbackQuery(cq.id, '该提问已失效(可能已在 Web 端作答)')
return
}
const payload = (cq.data ?? '').slice(ASK_CB.length)
const sep = payload.indexOf(':')
const qi = Number(payload.slice(0, sep))
const action = payload.slice(sep + 1)
const q = pending.questions[qi]
if (!q) {
await this.client.answerCallbackQuery(cq.id, '题目已失效')
return
}
const opts = Array.isArray(q.options) ? q.options : []
if (action === 'ok') {
if (!pending.kbSel.get(qi)?.size) {
await this.client.answerCallbackQuery(cq.id, '请先点选选项')
return
}
} else {
const j = Number(action)
if (!Number.isInteger(j) || j < 0 || j >= opts.length) {
await this.client.answerCallbackQuery(cq.id, '选项不存在')
return
}
if (q.multiSelect) {
const set = pending.kbSel.get(qi) ?? new Set<number>()
if (set.has(j)) set.delete(j)
else set.add(j)
pending.kbSel.set(qi, set)
await this.client.answerCallbackQuery(cq.id, set.size > 0 ? `已选 ${set.size} 项,选完点"提交所选"` : '已取消该选项')
} else {
pending.kbSel.set(qi, new Set([j]))
await this.client.answerCallbackQuery(cq.id, `已选:${opts[j]!.label}`)
}
}
if (!this.finalizeAskIfComplete(chatId, pending) && action === 'ok') {
await this.client.answerCallbackQuery(cq.id, '已提交本题选择,请继续作答其他问题')
}
ask.resolve({ answers })
this.enqueueNotice(chatId, '已提交你的回答')
}
// ── approval: TG answering via dual-path race on the request waterfall ──
+8 -2
View File
@@ -153,12 +153,17 @@ export class TelegramClient implements TelegramClientLike {
return message.split(this.token).join('***')
}
private async call<T>(method: string, body?: Record<string, unknown>): Promise<T> {
private async call<T>(method: string, body?: Record<string, unknown>, timeoutMs = 15_000): Promise<T> {
const url = `${this.baseUrl}/bot${this.token}/${method}`
try {
const init: RequestInit = {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
// Without a client-side timeout a stalled proxy connection hangs the
// request forever — and the serialized notice queue wedges behind it,
// silently swallowing every later tool/ask notice. TimeoutError is
// caught below and surfaces as a normal (retryable) failure.
signal: AbortSignal.timeout(timeoutMs),
}
if (body !== undefined) {
init.body = JSON.stringify(body)
@@ -187,7 +192,8 @@ export class TelegramClient implements TelegramClientLike {
if (offset !== undefined) {
body.offset = offset
}
return this.call<TelegramUpdate[]>('getUpdates', body)
// Long-poll: client-side abort must outlive the server-side hold.
return this.call<TelegramUpdate[]>('getUpdates', body, this.pollingTimeoutSec * 1000 + 10_000)
}
async sendMessage(
+23 -1
View File
@@ -14,6 +14,28 @@ function inlineToHtml(text: string): string {
return html
}
/** Convert `> …` line groups into `<blockquote>` (Telegram renders it with the accent-colored left bar). */
function blocksToHtml(text: string): string {
const lines = text.split('\n')
const out: string[] = []
let quote: string[] = []
const flush = (): void => {
if (quote.length === 0) return
out.push(`<blockquote>${inlineToHtml(quote.join('\n'))}</blockquote>`)
quote = []
}
for (const line of lines) {
if (line.startsWith('> ')) quote.push(line.slice(2))
else if (line === '>') quote.push('')
else {
flush()
out.push(inlineToHtml(line))
}
}
flush()
return out.join('\n')
}
export function markdownToHtml(text: string): string {
const parts = text.split(/```[\w-]*/)
if (parts.length % 2 === 0) return inlineToHtml(text)
@@ -23,7 +45,7 @@ export function markdownToHtml(text: string): string {
const code = parts[i]!.replace(/^\n/, '').replace(/\n$/, '')
html += `<pre>${escapeHtml(code)}</pre>`
} else {
html += inlineToHtml(parts[i]!)
html += blocksToHtml(parts[i]!)
}
}
return html