X-Git-Url: https://git.giacomolli.uk/onechat/blobdiff_plain/1e2df788bbb17653a6df44acd41da2edb109d403..195c3fbb075383ac47bba310eb701b02774e8d10:/telegram.go

diff --git a/telegram.go b/telegram.go
index 9d7600f..f3b3bb2 100644
--- a/telegram.go
+++ b/telegram.go
@@ -7,7 +7,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"log"
+	"log/slog"
 	"net/http"
 	"os"
 	"path/filepath"
@@ -18,6 +18,14 @@ import (
 )
 
 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"
 
@@ -73,8 +81,10 @@ func loadTelegramConfig(dir string) (TelegramConfig, bool, error) {
 	}
 
 	var config TelegramConfig
+
 	decoder := json.NewDecoder(bytes.NewReader(data))
 	decoder.DisallowUnknownFields()
+
 	err = decoder.Decode(&config)
 	if err == nil && !errors.Is(decoder.Decode(&struct{}{}), io.EOF) {
 		err = errors.New("data after the JSON object")
@@ -95,14 +105,12 @@ func loadTelegramConfig(dir string) (TelegramConfig, bool, error) {
 // newTelegramBot returns the bot of owner on api. The offset file is at
 // offsetPath, and logger gets the failures. No log line holds the text of a
 // message.
-func newTelegramBot(api *TelegramAPI, owner int64, offsetPath string, logger *log.Logger) *TelegramBot {
+func newTelegramBot(api *TelegramAPI, owner int64, offsetPath string, logger *slog.Logger) *TelegramBot {
 	return &TelegramBot{
 		api:        api,
 		owner:      owner,
 		offsetPath: offsetPath,
 		logger:     logger,
-		minWait:    time.Second,
-		maxWait:    time.Minute,
 		ready:      make(chan struct{}, 1),
 	}
 }
@@ -131,30 +139,28 @@ func (c TelegramConfig) validate() error {
 }
 
 // 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 Bot API for new messages and sends owner text and images 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.
+// 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
 	owner      int64
 	offsetPath string
-	logger     *log.Logger
-	minWait    time.Duration
-	maxWait    time.Duration
+	logger     *slog.Logger
 
 	mutex sync.Mutex
 	queue []Event
@@ -180,6 +186,7 @@ func (b *TelegramBot) Notify(event Event) {
 	}
 
 	b.mutex.Lock()
+
 	b.queue = append(b.queue, event)
 	b.mutex.Unlock()
 
@@ -190,14 +197,16 @@ func (b *TelegramBot) Notify(event Event) {
 }
 
 // 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()
 
 	var sender sync.WaitGroup
+
 	sender.Add(1)
+
 	go func() {
 		defer sender.Done()
 
@@ -205,6 +214,7 @@ func (b *TelegramBot) Run(ctx context.Context, l *TurnLoop, whatsapp *WhatsApp)
 	}()
 
 	err := b.poll(ctx, l, whatsapp)
+
 	cancel()
 	sender.Wait()
 
@@ -237,6 +247,7 @@ func (b *TelegramBot) send(ctx context.Context) {
 		switch event.Kind {
 		case "busy":
 			busy = true
+
 			b.typing(ctx)
 			ticker.Reset(telegramTypingInterval)
 		case "idle":
@@ -251,9 +262,9 @@ func (b *TelegramBot) send(ctx context.Context) {
 	}
 }
 
-// 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 {
@@ -262,11 +273,20 @@ func (b *TelegramBot) poll(ctx context.Context, l *TurnLoop, whatsapp *WhatsApp)
 
 	outage := &TelegramOutage{
 		logger: b.logger,
-		loss:   "cannot get messages from Telegram",
-		back:   "messages from Telegram come again",
+		loss:   "receive failed",
+		back:   "receive back",
+	}
+
+	imageOutage := &TelegramOutage{
+		logger: b.logger,
+		loss:   "image receive failed",
+		back:   "image receive back",
 	}
 
-	wait := b.minWait
+	imageWait := telegramRetryDelay
+	wait := telegramRetryDelay
+
+poll:
 	for ctx.Err() == nil {
 		updates, err := b.api.GetUpdates(ctx, offset)
 		if ctx.Err() != nil {
@@ -281,17 +301,29 @@ func (b *TelegramBot) poll(ctx context.Context, l *TurnLoop, whatsapp *WhatsApp)
 		}
 
 		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 {
+				retry, asked := telegramRetryable(err)
+				if retry {
+					imageWait = b.pause(ctx, imageOutage, err, imageWait, asked)
+					continue poll
+				}
+
 				return err
 			}
 
+			imageOutage.End()
+
+			imageWait = telegramRetryDelay
+
 			offset = update.UpdateID + 1
 			err = telegramSaveOffset(b.offsetPath, offset)
 			if err != nil {
@@ -345,7 +377,8 @@ func (b *TelegramBot) post(ctx context.Context, event Event) {
 }
 
 // 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)
@@ -358,26 +391,62 @@ func (b *TelegramBot) pause(ctx context.Context, outage *TelegramOutage, err err
 	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. 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 {
-		b.logger.Print("a Telegram message is ignored, because it is not from the owner in the private chat")
+		b.logger.Info("message ignored: not from the owner")
+
 		return nil
 	}
 
 	name, stop := stopSubagentName(message.Text)
 	args, isWhatsApp := whatsappCommandArgs(message.Text)
 	switch {
+	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:
@@ -401,7 +470,7 @@ func (b *TelegramBot) handle(ctx context.Context, l *TurnLoop, whatsapp *WhatsAp
 			return err
 		}
 
-		err = l.Flush(ctx)
+		err = l.Flush(context.WithoutCancel(ctx))
 		if err != nil {
 			return fmt.Errorf("flush the message from Telegram: %w", err)
 		}
@@ -416,7 +485,7 @@ func (b *TelegramBot) handle(ctx context.Context, l *TurnLoop, whatsapp *WhatsAp
 // 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)
@@ -429,15 +498,18 @@ func (b *TelegramBot) deliver(ctx context.Context, part TelegramPart) {
 		}
 
 		var answer *TelegramError
+
 		if len(part.entities) > 0 && errors.As(err, &answer) && answer.Status == http.StatusBadRequest {
-			b.logger.Printf("a reply goes as plain text, because Telegram refused its formatting: %v", err)
+			b.logger.Warn("reply sent as plain text", "err", err)
+
 			part.entities = nil
 
 			continue
 		}
 
 		if !retry {
-			b.logger.Printf("a reply to Telegram is lost: %v", err)
+			b.logger.Error("reply lost", "err", err)
+
 			return
 		}
 
@@ -454,7 +526,7 @@ func (b *TelegramBot) deliver(ctx context.Context, part TelegramPart) {
 // 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)
@@ -467,7 +539,8 @@ func (b *TelegramBot) deliverRich(ctx context.Context, markdown string) {
 		}
 
 		if telegramRejected(err) {
-			b.logger.Printf("a reply goes with simpler formatting, because Telegram refused its full formatting: %v", err)
+			b.logger.Warn("reply sent with simple formatting", "err", err)
+
 			for _, part := range telegramReplyParts(telegramUnescapeTags(markdown)) {
 				b.deliver(ctx, part)
 			}
@@ -476,7 +549,8 @@ func (b *TelegramBot) deliverRich(ctx context.Context, markdown string) {
 		}
 
 		if !retry {
-			b.logger.Printf("a reply to Telegram is lost: %v", err)
+			b.logger.Error("reply lost", "err", err)
+
 			return
 		}
 
@@ -488,8 +562,8 @@ func (b *TelegramBot) deliverRich(ctx context.Context, markdown string) {
 func (b *TelegramBot) sendOutage() *TelegramOutage {
 	return &TelegramOutage{
 		logger: b.logger,
-		loss:   "cannot send replies to Telegram",
-		back:   "replies go to Telegram again",
+		loss:   "send failed",
+		back:   "send back",
 	}
 }
 
@@ -506,7 +580,7 @@ func (b *TelegramBot) notice(text string) {
 // 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 *log.Logger
+	logger *slog.Logger
 	loss   string
 	back   string
 	start  time.Time
@@ -519,7 +593,7 @@ func (o *TelegramOutage) Fail(err error, delay time.Duration) {
 	}
 
 	o.start = time.Now()
-	o.logger.Printf("%s: %v (next try in %s)", o.loss, err, durationText(delay))
+	o.logger.Warn(o.loss, "err", err, "retry", delay)
 }
 
 // End logs the end of the outage that runs, with its length.
@@ -528,7 +602,8 @@ func (o *TelegramOutage) End() {
 		return
 	}
 
-	o.logger.Printf("%s, after %s", o.back, durationText(time.Since(o.start)))
+	o.logger.Info(o.back, "after", time.Since(o.start))
+
 	o.start = time.Time{}
 }
 
@@ -557,6 +632,7 @@ func telegramLoadOffset(path string) (int64, error) {
 	}
 
 	var file TelegramOffsetFile
+
 	err = json.Unmarshal(data, &file)
 	if err != nil {
 		return 0, fmt.Errorf("%s: %w", path, err)
@@ -588,6 +664,7 @@ func telegramSaveOffset(path string, offset int64) error {
 // never cuts a character.
 func telegramSplitMessage(text string, limit int) []string {
 	var parts []string
+
 	for {
 		end := telegramPrefixEnd(text, limit)
 		if end == len(text) {
@@ -637,6 +714,7 @@ func telegramText(event Event) string {
 // not cut an entity of telegramEscapeTags. It never cuts a character.
 func telegramSplitRich(markdown string, limit int) []string {
 	var parts []string
+
 	for len(markdown) > limit {
 		cut, fence, opening := telegramRichCut(markdown, limit)
 		if fence == "" {
@@ -692,7 +770,9 @@ func telegramPrefixEnd(text string, limit int) int {
 // so each cut makes the rest shorter.
 func telegramRichCut(markdown string, limit int) (int, string, string) {
 	blank, line, code := 0, 0, 0
+
 	var fence, opening, inFence, inOpening string
+
 	for start := 0; start < limit; {
 		end := strings.IndexByte(markdown[start:], '\n')
 		if end < 0 || start+end+1 > limit {
@@ -701,10 +781,10 @@ func telegramRichCut(markdown string, limit int) (int, string, string) {
 
 		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
@@ -751,10 +831,13 @@ func telegramRichCut(markdown string, limit int) (int, string, string) {
 // An entity that crosses a cut becomes one entity in each part.
 func telegramSplitEntities(text string, entities []TelegramEntity, limit int) []TelegramPart {
 	var parts []TelegramPart
+
 	start := 0
 	for _, part := range telegramSplitMessage(text, limit) {
 		end := start + telegramUTF16Units(part)
+
 		var kept []TelegramEntity
+
 		for _, entity := range entities {
 			from := max(entity.Offset, start)
 			to := min(entity.Offset+entity.Length, end)