From a0043494a0ab9f8b7ee4c85fc4e72a383c86e370 Mon Sep 17 00:00:00 2001
From: Cursor Agent <cursoragent@cursor.com>
Date: Fri, 22 May 2026 21:33:07 +0000
Subject: [PATCH] fix(gemini realtime): deep-merge nested config in follow-up
 session update

Previously, the follow-up setup performed a shallow merge between the
original setup and new overrides. If a session.update touched any field
inside generationConfig (e.g. modalities), the entire generationConfig
would be replaced, silently dropping unrelated sub-keys like temperature
or maxOutputTokens. Apply the same deep-merge to realtimeInputConfig so
partial automatic-activity-detection updates don't drop other realtime
input config fields either.

Co-authored-by: Yassin Kortam <yassin@berri.ai>
---
 .../llms/gemini/realtime/transformation.py    | 21 +++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/litellm/llms/gemini/realtime/transformation.py b/litellm/llms/gemini/realtime/transformation.py
index a1df67e9a98..29c7df8f289 100644
--- a/litellm/llms/gemini/realtime/transformation.py
+++ b/litellm/llms/gemini/realtime/transformation.py
@@ -349,6 +349,27 @@ class GeminiRealtimeConfig(BaseRealtimeConfig):
             **new_overrides,
             "model": f"models/{model}",
         }
+        # Deep-merge nested config dicts so that a partial session.update
+        # (e.g. only ``modalities``) does not silently drop unrelated
+        # sub-keys (e.g. ``temperature``) from the original setup.
+        original_generation_config = original_setup.get("generationConfig")
+        new_generation_config = new_overrides.get("generationConfig")
+        if isinstance(original_generation_config, dict) and isinstance(
+            new_generation_config, dict
+        ):
+            follow_up_setup["generationConfig"] = {
+                **original_generation_config,
+                **new_generation_config,
+            }
+        original_realtime_input_config = original_setup.get("realtimeInputConfig")
+        new_realtime_input_config = new_overrides.get("realtimeInputConfig")
+        if isinstance(original_realtime_input_config, dict) and isinstance(
+            new_realtime_input_config, dict
+        ):
+            follow_up_setup["realtimeInputConfig"] = {
+                **original_realtime_input_config,
+                **new_realtime_input_config,
+            }
         verbose_logger.debug(
             "Gemini Realtime: Forwarding session.update as follow-up setup"
         )