fix(tinybard): add background pre-warming of Kokoro TTS on server startup

.agents/tasks/TASK.2026-06-10_prewarm.md

@@ -0,0 +1,14 @@
+# Task: TinyBard Startup Pre-warming of TTS Models
+
+- Branch: fix/tinybard-startup-prewarm
+- Date: 2026-06-10
+
+## Chain-of-Draft
+
+1. Initialize task file.
+2. Add background thread prewarm.
+3. Merge and push.
+4. Verify on HF Space.
+
+####
+Proceeding with implementation.

projects/tinybard/app.py

@@ -1514,6 +1514,11 @@ mount_gradio_app(fastapi_app, gradio_blocks, path="/gradio")
 # Entrypoint
 # ---------------------------------------------------------------------------
 if __name__ == "__main__":
+    # Pre-warm Kokoro TTS in a background thread to cache models on startup
+    import threading
+    log.info("Pre-warming Kokoro TTS client in the background...")
+    threading.Thread(target=get_kokoro_client, daemon=True).start()
+
     # On HF Spaces, the platform handles the server — just launch Gradio
     if os.environ.get("SPACE_ID"):
         log.info("Running on HF Spaces — launching Gradio directly")