wip

Author: truemagic-coder

README.md

@@ -299,18 +299,18 @@ async def generate():
         rt_encode_output=True,
         rt_voice="marin",
         output_format="audio",
-        audio_output_format="m4a",
+        audio_output_format="mp3",
         audio_input_format="mp4",
     ):
         yield chunk
 
 return StreamingResponse(
     content=generate(),
-    media_type="audio/mp4",
+    media_type="audio/mp3",
     headers={
         "Cache-Control": "no-store",
         "Pragma": "no-cache",
-        "Content-Disposition": "inline; filename=stream.m4a",
+        "Content-Disposition": "inline; filename=stream.mp3",
         "X-Accel-Buffering": "no",
     },
 )

docs/index.rst

@@ -224,18 +224,18 @@ This example will work using expo-audio on Android and iOS.
          rt_encode_output=True,
          rt_voice="marin",
          output_format="audio",
-         audio_output_format="m4a",
-         audio_input_format="mp4",
+         audio_output_format="mp3",
+         audio_input_format="m4a",
       ):
          yield chunk
 
    return StreamingResponse(
       content=generate(),
-      media_type="audio/mp4",
+      media_type="audio/mp3",
       headers={
          "Cache-Control": "no-store",
          "Pragma": "no-cache",
-         "Content-Disposition": "inline; filename=stream.m4a",
+         "Content-Disposition": "inline; filename=stream.mp3",
          "X-Accel-Buffering": "no",
       },
    )

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "solana-agent"
-version = "31.2.4-dev1"
+version = "31.2.4-dev2"
 description = "AI Agents for Solana"
 authors = ["Bevan Hunt <bevan@bevanhunt.com>"]
 license = "MIT"

solana_agent/adapters/openai_realtime_ws.py

@@ -1037,6 +1037,15 @@ def _strip_tool_strict(tools_val):
         if "tools" in patch:
             patch["tools"] = _strip_tool_strict(patch["tools"])  # idempotent
 
+        # Per server requirements, always include session.type and output_modalities
+        try:
+            patch["type"] = "realtime"
+            # Preserve caller-provided output_modalities if present, otherwise default to audio
+            if "output_modalities" not in patch:
+                patch["output_modalities"] = ["audio"]
+        except Exception:
+            pass
+
         payload = {"type": "session.update", "session": patch}
         # Mark awaiting updated and store last patch
         self._last_session_patch = patch or {}