cap audio_stream

tools/mtmd/mtmd-cli.cpp

@@ -567,7 +567,7 @@ int main(int argc, char ** argv) {
 
     if (g_is_interrupted) return 130;
 
-    if (mtmd_support_audio_stream(ctx.ctx_vision.get())) {
+    if (ctx.mtmd_cap.audio_stream) {
         if (params.image.size() != 1) {
             LOG_ERR("This model transcribes audio in realtime, pass exactly one --audio file\n");
             return 1;

tools/mtmd/mtmd-helper.cpp

@@ -1033,6 +1033,7 @@ mtmd_helper_cap mtmd_helper_get_capabilities(const llama_model * model, const mt
 
     cap.vision = mtmd_support_vision(mctx);
     cap.audio  = mtmd_support_audio(mctx);
+    cap.audio_stream = mtmd_support_audio_stream(mctx);
 #ifdef MTMD_VIDEO
     cap.video  = cap.vision;
 #endif

tools/mtmd/mtmd-helper.h

@@ -191,6 +191,9 @@ struct mtmd_helper_cap {
     bool audio;
     bool video; // also depending on MTMD_VIDEO build option
 
+    // realtime capabilities
+    bool audio_stream; // realtime transcription, see mtmd_audio_stream_init()
+
     // output capabilities
     bool audio_gen;
 };

tools/server/server-context.cpp

@@ -5130,7 +5130,7 @@ server_context_meta server_context::get_meta() const {
         /* has_inp_video           */ impl->chat_params.allow_video,
         /* has_cap_chat            */ impl->has_cap_chat(),
         /* has_cap_tts             */ impl->has_cap_tts(),
-        /* has_audio_stream        */ impl->mctx != nullptr && mtmd_support_audio_stream(impl->mctx),
+        /* has_audio_stream        */ impl->mtmd_cap.audio_stream,
         /* json_ui_settings        */ impl->json_ui_settings,
         /* slot_n_ctx              */ impl->n_ctx_slot(),
         /* pooling_type            */ llama_pooling_type(impl->ctx_tgt),