internal/service/telegram/bot.go (1)

191-206: Duplicate stopTypingLoop call after CancelSession.

Line 193 stops the typing loop immediately after clearing pending messages. Line 205 calls stopTypingLoop again after CancelSession. Since stopTypingLoop is idempotent (sets typingCancel/typingDone to nil under lock), the second call is a harmless no-op but redundant—nothing between lines 193 and 205 restarts the loop.

🔧 Optional: Remove the redundant call

 	case "cancel":
 		b.clearPendingMessages(cs)
 		b.stopTypingLoop(cs)
 		sid, ownerUID := cs.ActiveSession()

 		if sid == "" {
 			b.sendText(chatID, "No active session.")
 			return
 		}

 		if err := b.agentAPI.CancelSession(ctx, sid, ownerUID); err != nil {
 			b.sendText(chatID, "Failed to cancel session: "+err.Error())
 			return
 		}
-		b.stopTypingLoop(cs)
 		b.sendText(chatID, "Session cancelled.")

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@internal/service/telegram/bot.go` around lines 191 - 206, The duplicate
stopTypingLoop call is redundant: remove the second stopTypingLoop invocation
after b.agentAPI.CancelSession to avoid unnecessary no-op; keep the initial
b.stopTypingLoop(cs) right after b.clearPendingMessages(cs) and leave the
CancelSession call and subsequent sendText/ActiveSession logic unchanged (look
for stopTypingLoop, clearPendingMessages, ActiveSession, CancelSession, and
sendText in this case "cancel" branch to locate the code).

internal/service/telegram/bot_test.go (1)

356-385: Test is valid but consider adding agentAPI for robustness.

The test omits agentAPI from the Bot struct. This works because the test relies on the early return at line 196-198 when no active session exists, never reaching b.agentAPI.CancelSession(...). However, if the /cancel logic is later modified to access agentAPI before checking ActiveSession(), this test would panic.

Consider adding the fake agent service for defensive test design:

🔧 Optional: Add agentAPI to test setup

 func TestBot_HandleCommand_CancelStopsPreFlushTypingWithoutSession(t *testing.T) {
 	t.Parallel()

 	api := &fakeTelegramAPI{}
 	logger := slog.New(slog.NewTextHandler(io.Discard, nil))
+	service := newFakeTelegramAgentService("ignored")
 	bot := &Bot{
+		agentAPI:     service,
 		botAPI:       api,
 		logger:       logger,
 		allowedChats: map[int64]struct{}{123: {}},
 	}

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@internal/service/telegram/bot_test.go` around lines 356 - 385, The test
TestBot_HandleCommand_CancelStopsPreFlushTypingWithoutSession constructs a Bot
without populating bot.agentAPI which could cause a panic if /cancel logic
changes; update the test setup to assign a fake agent service to the Bot (e.g.,
create and set a fakeAgentAPI instance on bot.agentAPI) so bot.handleCommand has
a non-nil agentAPI during execution and the existing assertions around
getOrCreateChat, startTypingLoop, and assertTypingStops remain unchanged.