Use vLLM audio URL payloads

internal/transcription/client.go

@@ -120,14 +120,13 @@ type audioLLMChatMessage struct {
 }
 
 type audioLLMContentPart struct {
-	Type       string         `json:"type"`
-	Text       string         `json:"text,omitempty"`
-	InputAudio *audioLLMAudio `json:"input_audio,omitempty"`
+	Type     string          `json:"type"`
+	Text     string          `json:"text,omitempty"`
+	AudioURL *audioLLMURLRef `json:"audio_url,omitempty"`
 }
 
-type audioLLMAudio struct {
-	Data   string `json:"data"`
-	Format string `json:"format,omitempty"`
+type audioLLMURLRef struct {
+	URL string `json:"url"`
 }
 
 type audioLLMChatResponse struct {
@@ -567,11 +566,8 @@ func (c *Client) transcribeAudioLLM(ctx context.Context, provider ProviderConfig
 				Content: []audioLLMContentPart{
 					{Type: "text", Text: prompt},
 					{
-						Type: "input_audio",
-						InputAudio: &audioLLMAudio{
-							Data:   base64.StdEncoding.EncodeToString(audio),
-							Format: audioFormat(filename),
-						},
+						Type:     "audio_url",
+						AudioURL: &audioLLMURLRef{URL: audioDataURL(audio, filename)},
 					},
 				},
 			},
@@ -634,6 +630,10 @@ func audioFormat(filename string) string {
 	}
 }
 
+func audioDataURL(audio []byte, filename string) string {
+	return "data:audio/" + audioFormat(filename) + ";base64," + base64.StdEncoding.EncodeToString(audio)
+}
+
 func firstNonEmpty(values ...string) string {
 	for _, value := range values {
 		if strings.TrimSpace(value) != "" {

internal/transcription/client_test.go

@@ -58,6 +58,14 @@ func TestNewWithOptionsBuildsComparisonProviders(t *testing.T) {
 	}
 }
 
+func TestAudioDataURLUsesVLLMAudioURLFormat(t *testing.T) {
+	got := audioDataURL([]byte("abc"), "call.wav")
+	want := "data:audio/wav;base64,YWJj"
+	if got != want {
+		t.Fatalf("audio data url = %q, want %q", got, want)
+	}
+}
+
 func near(got, want float64) bool {
 	return math.Abs(got-want) < 0.000001
 }