feat: add prompt startup actions and shared vision config

backend/services/pipecat/pipeline.py

@@ -270,7 +270,6 @@ async def run_pipeline(
             build_workflow_voice_switcher(cfg, "TTS", tts)
         )
 
-    greeting = await brain.greeting(cfg)
     system_content = brain.system_prompt(cfg)
 
     worker_holder: dict = {}
@@ -683,6 +682,14 @@ async def run_pipeline(
             flow_global_functions=flow_global_functions,
         ),
     )
+    try:
+        await brain.run_preflight()
+    except Exception:
+        if recorder:
+            await recorder.finish(status="failed")
+        raise
+    # Preflight tools may assign variables used by the opening speech.
+    greeting = await brain.greeting(cfg)
 
     async def submit_user_input(value: UserInput) -> None:
         if not value.has_camera_frame: