From aa6fdff11ebb4f674e3f4d2137518a7de1a2a41d Mon Sep 17 00:00:00 2001
From: Hua Audio <jiahua.zhang@kuleuven.be>
Date: Sat, 21 Mar 2026 09:42:09 +0100
Subject: [PATCH] Apply suggestions from code review

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
---
 pkg/bus/types.go              |  2 +-
 pkg/channels/discord/voice.go | 39 +++++++++++++++++++++++++++----
 pkg/tts/tts.go                | 44 +++++++++++++++++++++++++++++++----
 3 files changed, 75 insertions(+), 10 deletions(-)

diff --git a/pkg/bus/types.go b/pkg/bus/types.go
index 794db5b0f..c648fb0ce 100644
--- a/pkg/bus/types.go
+++ b/pkg/bus/types.go
@@ -70,5 +70,5 @@ type AudioChunk struct {
 type VoiceControl struct {
 	SessionID string `json:"session_id"`
 	Type      string `json:"type"`   // "state", "command"
-	Action    string `json:"action"` // "idle", "listening", "start", "stop"
+	Action    string `json:"action"` // "idle", "listening", "start", "stop", "leave"
 }
diff --git a/pkg/channels/discord/voice.go b/pkg/channels/discord/voice.go
index d5d5d303f..37c882b25 100644
--- a/pkg/channels/discord/voice.go
+++ b/pkg/channels/discord/voice.go
@@ -73,17 +73,46 @@ func streamOggOpusToDiscord(ctx context.Context, vc *discordgo.VoiceConnection,
 func (c *DiscordChannel) receiveVoice(vc *discordgo.VoiceConnection, guildID string, chatID string) {
 	logger.InfoCF("discord", "Started listening for voice", map[string]any{"guild": guildID})
 
-	go func() {
+	go func(ctx context.Context, vc *discordgo.VoiceConnection) {
+		// Recover from potential panics if OpusSend is closed mid-send.
+		defer func() {
+			if rec := recover(); rec != nil {
+				logger.WarnCF("discord", "Recovered from panic while sending wake-up frames", map[string]any{
+					"error": rec,
+					"guild": guildID,
+				})
+			}
+		}()
+
+		// If the voice connection or OpusSend are not available, nothing to do.
+		if vc == nil || vc.OpusSend == nil {
+			return
+		}
+
 		time.Sleep(250 * time.Millisecond) // Wait a bit for connection to settle
+
+		// Abort if the context has already been cancelled.
+		select {
+		case <-ctx.Done():
+			return
+		default:
+		}
+
 		vc.Speaking(true)
+		defer vc.Speaking(false)
+
+		silenceFrame := []byte{0xF8, 0xFF, 0xFE}
 		for i := 0; i < 5; i++ {
-			vc.OpusSend <- []byte{0xF8, 0xFF, 0xFE}
+			select {
+			case <-ctx.Done():
+				return
+			case vc.OpusSend <- silenceFrame:
+			}
 			time.Sleep(20 * time.Millisecond)
 		}
-		vc.Speaking(false)
-		logger.DebugCF("discord", "Sent wake-up silence frames", nil)
-	}()
 
+		logger.DebugCF("discord", "Sent wake-up silence frames", map[string]any{"guild": guildID})
+	}(c.ctx, vc)
 	sessionID := fmt.Sprintf("discord_vc_%s", guildID)
 
 	c.bus.PublishVoiceControl(c.ctx, bus.VoiceControl{
diff --git a/pkg/tts/tts.go b/pkg/tts/tts.go
index 63b4ecd24..25b533326 100644
--- a/pkg/tts/tts.go
+++ b/pkg/tts/tts.go
@@ -29,11 +29,47 @@ type OpenAITTSProvider struct {
 }
 
 func NewOpenAITTSProvider(apiKey string, apiBase string, proxyURL string) *OpenAITTSProvider {
-	if apiBase == "" || apiBase == "https://api.openai.com/v1" {
+	// Normalize apiBase to avoid malformed endpoints like
+	// "https://api.openai.com/audio/speech" when "/v1" is required.
+	if apiBase == "" {
 		apiBase = "https://api.openai.com/v1/audio/speech"
-	} else if !strings.HasSuffix(apiBase, "/audio/speech") {
-		// Just in case they provide openrouter base or standard base
-		apiBase = strings.TrimSuffix(apiBase, "/") + "/audio/speech"
+	} else {
+		if u, err := url.Parse(apiBase); err == nil && u.Scheme != "" && u.Host != "" {
+			path := u.Path
+			if u.Host == "api.openai.com" {
+				// For the official OpenAI host, ensure exactly one /v1 prefix and
+				// that the path ends with /audio/speech.
+				if path == "" || path == "/" || path == "/v1" {
+					path = "/v1/audio/speech"
+				} else {
+					if !strings.HasPrefix(path, "/") {
+						path = "/" + path
+					}
+					if !strings.HasPrefix(path, "/v1/") {
+						path = "/v1" + strings.TrimSuffix(path, "/")
+					}
+					if !strings.HasSuffix(path, "/audio/speech") {
+						path = strings.TrimSuffix(path, "/") + "/audio/speech"
+					}
+				}
+			} else {
+				// For non-OpenAI hosts (e.g., proxies), preserve the existing base
+				// path and only ensure it ends with /audio/speech.
+				if !strings.HasSuffix(path, "/audio/speech") {
+					path = strings.TrimSuffix(path, "/") + "/audio/speech"
+				}
+			}
+			u.Path = path
+			apiBase = u.String()
+		} else {
+			// Fallback to the previous string-based behavior if parsing fails.
+			if apiBase == "https://api.openai.com/v1" {
+				apiBase = "https://api.openai.com/v1/audio/speech"
+			} else if !strings.HasSuffix(apiBase, "/audio/speech") {
+				// Just in case they provide openrouter base or standard base
+				apiBase = strings.TrimSuffix(apiBase, "/") + "/audio/speech"
+			}
+		}
 	}
 
 	client := &http.Client{