fix(agno): handle streaming mode in InvokeAsync to prevent pickle error

memori/llm/_invoke.py

@@ -71,9 +71,26 @@ async def invoke(self, **kwargs):
         )
 
         raw_response = await self._method(**kwargs)
+
+        # Check if streaming is enabled - if so, return a wrapped async generator
+        # that handles post-processing after the stream is consumed.
+        # This fixes the "cannot pickle '_thread.RLock' object" error when using
+        # OpenAI with Agno in streaming mode (Issue #214)
+        if kwargs.get("stream", False):
+            return self._wrap_stream(raw_response, kwargs, start)
+
         self.handle_post_response(kwargs, start, raw_response)
         return raw_response
 
+    async def _wrap_stream(self, stream, kwargs, start):
+        """Wrap an async stream to handle post-processing after consumption."""
+        raw_response = {}
+        async for chunk in stream:
+            raw_response = merge_chunk(raw_response, chunk.__dict__)
+            yield chunk
+
+        self.handle_post_response(kwargs, start, raw_response)
+
 
 class InvokeAsyncIterator(BaseInvoke):
     async def invoke(self, **kwargs):