)
const (
+ // telegramRetryDelay is the wait before the first retry of a failed
+ // call.
+ telegramRetryDelay = time.Second
+
+ // telegramRetryLimit is the longest wait between retries of a failed
+ // call.
+ telegramRetryLimit = time.Minute
+
// telegramAPIURL is the Telegram Bot API service.
telegramAPIURL = "https://api.telegram.org"
telegramTextOnly = "Only text messages are supported."
)
+// TelegramConfig holds the settings of telegram.json: the bot token and the
+// user ID of the owner. The token is a secret, so the file must have mode 0600.
+type TelegramConfig struct {
+ Token string `json:"token"`
+ Owner int64 `json:"owner"`
+}
+
+// TelegramBot connects the private chat of the owner with the agent. It polls
+// the Bot API for new messages and sends owner text, images, and voice notes to
+// the agent. It ignores every other update. Its sender sends the replies, the
+// errors, the notices of the bot, and the user messages of the other clients to
+// the chat, and shows the typing action while a turn runs. Only the replies
+// have a format: their Markdown goes as rich messages, which Telegram renders
+// itself, and as message entities only when Telegram refuses a rich message.
+// The rest is plain text. Tool calls do not go to the chat.
+//
+// The offset file holds the ID of the next update. It moves past an update only
+// after the message of the update is durable in the log, so a restart does not
+// lose a message. A crash between the flush and the offset write makes Telegram
+// send the update again, and the message repeats once.
+//
+// A failed Bot API call waits and tries again, from telegramRetryDelay up to
+// telegramRetryLimit, or as long as the Bot API asks. It never stops the agent.
+// A TelegramOutage logs the first failure and the end of each run of failures,
+// and the retries between them log nothing.
+type TelegramBot struct {
+ api *TelegramAPI
+ voice *VoiceTranscriber
+ owner int64
+ offsetPath string
+ logger *slog.Logger
+
+ mutex sync.Mutex
+ queue []Event
+ ready chan struct{}
+}
+
+// TelegramOutage is one run of failures of a kind of Bot API call. The first
+// failure logs loss, its cause, and the wait of the next try. The retries log
+// nothing. The next answer of Telegram logs back and the length of the outage.
+// loss and back say what the user loses and gets again.
+type TelegramOutage struct {
+ logger *slog.Logger
+ loss string
+ back string
+ start time.Time
+}
+
+// TelegramPart is the text of one message to the chat, at most
+// telegramMessageUnits long, and its entities.
+type TelegramPart struct {
+ text string
+ entities []TelegramEntity
+}
+
+// TelegramOffsetFile is the content of the offset file.
+type TelegramOffsetFile struct {
+ Offset int64 `json:"offset"`
+}
+
// loadTelegramConfig reads telegram.json in dir. It returns false when the file
// does not exist. An error never holds the token.
func loadTelegramConfig(dir string) (TelegramConfig, bool, error) {
func newTelegramBot(api *TelegramAPI, owner int64, offsetPath string, logger *slog.Logger) *TelegramBot {
return &TelegramBot{
api: api,
+ voice: &VoiceTranscriber{},
owner: owner,
offsetPath: offsetPath,
logger: logger,
- minWait: time.Second,
- maxWait: time.Minute,
ready: make(chan struct{}, 1),
}
}
-// TelegramConfig holds the settings of telegram.json: the bot token and the
-// user ID of the owner. The token is a secret, so the file must have mode 0600.
-type TelegramConfig struct {
- Token string `json:"token"`
- Owner int64 `json:"owner"`
-}
-
// validate makes sure that the token has the form of a bot token and that the
// owner is a user ID. The token goes into a URL path, so it can only hold the
// characters of a bot token.
return nil
}
-// TelegramBot connects the private chat of the owner with the agent. It polls
-// the Bot API for new messages and sends each text message of the owner to the
-// agent. It ignores every other update. Its sender sends the replies, the
-// errors, the notices of the bot, and the user messages of the other clients to
-// the chat, and shows the typing action while a turn runs. Only the replies
-// have a format: their Markdown goes as rich messages, which Telegram renders
-// itself, and as message entities only when Telegram refuses a rich message.
-// The rest is plain text. Tool calls do not go to the chat.
-//
-// The offset file holds the ID of the next update. It moves past an update
-// only after the message of the update is durable in the log, so a restart
-// does not lose a message. A crash between the flush and the offset write
-// makes Telegram send the update again, and the message repeats once.
-//
-// A failed Bot API call waits and tries again, from minWait up to maxWait, or
-// as long as the Bot API asks. It never stops the agent. A TelegramOutage logs
-// the first failure and the end of each run of failures, and the retries
-// between them log nothing.
-type TelegramBot struct {
- api *TelegramAPI
- owner int64
- offsetPath string
- logger *slog.Logger
- minWait time.Duration
- maxWait time.Duration
-
- mutex sync.Mutex
- queue []Event
- ready chan struct{}
-}
-
-// Notify queues event for the chat. It never blocks, so a slow network does not
-// stop the agent. The WhatsApp batches and the runs of the schedules do not go
-// to the chat: the reply of the agent tells the user what needs the user.
-func (b *TelegramBot) Notify(event Event) {
- switch event.Kind {
- case "busy", "idle", "reply", "error", "notice", "subagent":
- case "work":
- if event.Source == whatsappSource || event.Source == scheduleSource {
- return
- }
- case "user":
- if event.Source == telegramSource {
- return
- }
- default:
- return
- }
-
- b.mutex.Lock()
-
- b.queue = append(b.queue, event)
- b.mutex.Unlock()
-
- select {
- case b.ready <- struct{}{}:
- default:
- }
-}
-
// Run polls for the messages of the owner and sends the queued events to the
-// chat until ctx ends. It returns an error only when a message or the offset
-// cannot be written, because then the bot would lose or repeat messages.
+// chat until ctx ends. It returns an error only when a message or the offset or
+// image cannot be written, because then the bot would lose or repeat messages.
func (b *TelegramBot) Run(ctx context.Context, l *TurnLoop, whatsapp *WhatsApp) error {
ctx, cancel := context.WithCancel(ctx)
defer cancel()
}
}
-// poll gets the updates from the stored offset on and handles each one, until
-// ctx ends. Each update is handled to its end also when ctx ends during it, so
-// a stop does not repeat its message.
+// poll gets updates from the stored offset and handles them until ctx ends. A
+// failed image download retries from the same offset. A submitted message is
+// flushed even if ctx ends, and only then can the offset advance.
func (b *TelegramBot) poll(ctx context.Context, l *TurnLoop, whatsapp *WhatsApp) error {
offset, err := telegramLoadOffset(b.offsetPath)
if err != nil {
- return err
+ return fmt.Errorf("load Telegram offset: %w", err)
}
outage := &TelegramOutage{
back: "receive back",
}
- wait := b.minWait
+ imageOutage := &TelegramOutage{
+ logger: b.logger,
+ loss: "image receive failed",
+ back: "image receive back",
+ }
+
+ imageWait := telegramRetryDelay
+ wait := telegramRetryDelay
+
+poll:
for ctx.Err() == nil {
updates, err := b.api.GetUpdates(ctx, offset)
if ctx.Err() != nil {
outage.End()
- wait = b.minWait
+ wait = telegramRetryDelay
+
for _, update := range updates {
if ctx.Err() != nil {
return nil
}
- err = b.handle(context.WithoutCancel(ctx), l, whatsapp, update)
+ err = b.handle(ctx, l, whatsapp, update)
if err != nil {
- return err
+ retry, asked := telegramRetryable(err)
+ if retry {
+ imageWait = b.pause(ctx, imageOutage, err, imageWait, asked)
+ continue poll
+ }
+
+ if err != nil {
+ return fmt.Errorf("handle Telegram update: %w", err)
+ }
+
+ return nil
}
+ imageOutage.End()
+
+ imageWait = telegramRetryDelay
+
offset = update.UpdateID + 1
err = telegramSaveOffset(b.offsetPath, offset)
if err != nil {
- return err
+ return fmt.Errorf("save Telegram offset: %w", err)
}
}
}
// typing shows the typing action once. A failure is not logged, because the
// next message or poll logs the state of the network.
func (b *TelegramBot) typing(ctx context.Context) {
- _ = b.api.SendTyping(ctx, b.owner)
+ // The next message or poll reports a network failure.
+ b.api.SendTyping(ctx, b.owner) // nolint: errcheck
}
// post sends event to the chat. A reply goes as rich messages of its Markdown,
}
// pause gives err to outage and waits the longer of wait and asked, or until
-// ctx ends. It returns the wait of the next failure: twice wait, up to maxWait.
+// ctx ends. It returns the wait of the next failure: twice wait, up to
+// telegramRetryLimit.
func (b *TelegramBot) pause(ctx context.Context, outage *TelegramOutage, err error, wait, asked time.Duration) time.Duration {
delay := max(wait, asked)
outage.Fail(err, delay)
case <-ctx.Done():
}
- return min(2*wait, b.maxWait)
+ return min(2*wait, telegramRetryLimit)
}
// handle runs one update. A text message of the owner in the private chat goes
// to the agent, and handle returns when it is durable in the log. The line
// /stop interrupts the turn, the line /stop NAME stops the subagent NAME, and a
// line that starts with /whatsapp runs a WhatsApp command on whatsapp, whose
-// answers go to the chat only. Any other message of the owner gets the answer
-// that only text is supported. Every other update is ignored, and its content
-// is not logged.
+// answers go to the chat only. Photos and image documents enter the shared
+// image store with their caption. Voice notes keep their audio and local
+// transcript. Other media keep the text-only answer. Every other update is
+// ignored, and its content is not logged.
func (b *TelegramBot) handle(ctx context.Context, l *TurnLoop, whatsapp *WhatsApp, update TelegramUpdate) error {
message := update.Message
if message == nil || message.Chat.Type != "private" || message.Chat.ID != b.owner || message.From == nil || message.From.ID != b.owner {
name, stop := stopSubagentName(message.Text)
args, isWhatsApp := whatsappCommandArgs(message.Text)
switch {
+ case message.Voice != nil:
+ text := telegramVoiceText(ctx, b.api, l.agent.Voices(), b.voice, message)
+ _, err := l.Submit(telegramSource, text)
+ if err != nil {
+ return fmt.Errorf("submit Telegram voice: %w", err)
+ }
+
+ err = l.Flush(context.WithoutCancel(ctx))
+ if err != nil {
+ return fmt.Errorf("flush Telegram voice: %w", err)
+ }
+ case len(message.Photo) > 0 || message.Document != nil:
+ text, err := telegramImageText(ctx, b.api, l.agent.Images(), message)
+ if err != nil {
+ retry, _ := telegramRetryable(err)
+ if retry {
+ return fmt.Errorf("receive image: %w", err)
+ }
+
+ switch {
+ case errors.Is(err, errTelegramOtherMedia):
+ b.notice(telegramTextOnly)
+ case errors.Is(err, errImageSize), errors.Is(err, errImageFormat), errors.Is(err, errTelegramPhotoSize):
+ b.notice(err.Error())
+ default:
+ var answer *TelegramError
+
+ if !errors.As(err, &answer) {
+ return fmt.Errorf("receive Telegram image: %w", err)
+ }
+
+ b.notice("Cannot download image.")
+ }
+
+ return nil
+ }
+
+ _, err = l.Submit(telegramSource, text)
+ if err != nil {
+ return fmt.Errorf("submit Telegram image: %w", err)
+ }
+
+ err = l.Flush(context.WithoutCancel(ctx))
+ if err != nil {
+ return fmt.Errorf("flush Telegram image: %w", err)
+ }
case message.Text == "":
b.notice(telegramTextOnly)
case strings.TrimSpace(message.Text) == stopCommand:
default:
_, err := l.Submit(telegramSource, message.Text)
if err != nil {
- return err
+ return fmt.Errorf("submit Telegram text: %w", err)
}
- err = l.Flush(ctx)
+ err = l.Flush(context.WithoutCancel(ctx))
if err != nil {
return fmt.Errorf("flush the message from Telegram: %w", err)
}
// later try can fix, until ctx ends. It drops the part after any other failure.
func (b *TelegramBot) deliver(ctx context.Context, part TelegramPart) {
outage := b.sendOutage()
- wait := b.minWait
+ wait := telegramRetryDelay
for {
err := b.api.SendMessage(ctx, b.owner, part.text, part.entities)
retry, asked := telegramRetryable(err)
// fix, until ctx ends. It drops the message after any other failure.
func (b *TelegramBot) deliverRich(ctx context.Context, markdown string) {
outage := b.sendOutage()
- wait := b.minWait
+ wait := telegramRetryDelay
for {
err := b.api.SendRichMessage(ctx, b.owner, markdown)
retry, asked := telegramRetryable(err)
}
}
-// sendOutage returns a new outage of the messages to the chat.
-func (b *TelegramBot) sendOutage() *TelegramOutage {
- return &TelegramOutage{
- logger: b.logger,
- loss: "send failed",
- back: "send back",
- }
-}
-
// notice queues a short answer of the bot for the chat.
func (b *TelegramBot) notice(text string) {
b.Notify(Event{
})
}
-// TelegramOutage is one run of failures of a kind of Bot API call. The first
-// failure logs loss, its cause, and the wait of the next try. The retries log
-// nothing. The next answer of Telegram logs back and the length of the outage.
-// loss and back say what the user loses and gets again.
-type TelegramOutage struct {
- logger *slog.Logger
- loss string
- back string
- start time.Time
+// sendOutage returns a new outage of the messages to the chat.
+func (b *TelegramBot) sendOutage() *TelegramOutage {
+ return &TelegramOutage{
+ logger: b.logger,
+ loss: "send failed",
+ back: "send back",
+ }
}
-// Fail logs the start of an outage with err and delay, unless an outage runs.
-func (o *TelegramOutage) Fail(err error, delay time.Duration) {
- if !o.start.IsZero() {
+// Notify queues event for the chat. It never blocks, so a slow network does not
+// stop the agent. The WhatsApp batches and the runs of the schedules do not go
+// to the chat: the reply of the agent tells the user what needs the user.
+func (b *TelegramBot) Notify(event Event) {
+ switch event.Kind {
+ case "busy", "idle", "reply", "error", "notice", "subagent":
+ case "work":
+ if event.Source == whatsappSource || event.Source == scheduleSource {
+ return
+ }
+ case "user":
+ if event.Source == telegramSource {
+ return
+ }
+ default:
return
}
- o.start = time.Now()
- o.logger.Warn(o.loss, "err", err, "retry", delay)
+ b.mutex.Lock()
+
+ b.queue = append(b.queue, event)
+ b.mutex.Unlock()
+
+ select {
+ case b.ready <- struct{}{}:
+ default:
+ }
}
// End logs the end of the outage that runs, with its length.
o.start = time.Time{}
}
-// TelegramPart is the text of one message to the chat, at most
-// telegramMessageUnits long, and its entities.
-type TelegramPart struct {
- text string
- entities []TelegramEntity
-}
+// Fail logs the start of an outage with err and delay, unless an outage runs.
+func (o *TelegramOutage) Fail(err error, delay time.Duration) {
+ if !o.start.IsZero() {
+ return
+ }
-// TelegramOffsetFile is the content of the offset file.
-type TelegramOffsetFile struct {
- Offset int64 `json:"offset"`
+ o.start = time.Now()
+ o.logger.Warn(o.loss, "err", err, "retry", delay)
}
// telegramLoadOffset reads the update offset at path. A missing file gives 0,
end += start + 1
text := strings.TrimRight(markdown[start:end], "\r\n")
- _, opens := telegramCodeFence(text)
+ _, opens := markdownCodeFence(text)
switch {
case fence != "":
- if telegramClosesFence(text, fence) {
+ if markdownClosesFence(text, fence) {
fence = ""
} else if end > len(opening)+1 && end+len(fence)+1 <= limit {
code, inFence, inOpening = end, fence, opening
projects / onechat / blobdiff