Merge fda103e8b5 into ce73cf8b9d

litellm/interactions/main.py

@@ -103,6 +103,7 @@ async def acreate(
     custom_llm_provider: str | None = None,
     *,
     agent_config: Mapping[str, object] | None = None,
+    continuation_token: str | None = None,
     **kwargs,
 ) -> InteractionsAPIResponse | AsyncIterator[InteractionsAPIStreamingResponse]:
     """
@@ -129,6 +130,8 @@ async def acreate(
         response_format: JSON schema for response format
         response_mime_type: MIME type of the response
         previous_interaction_id: ID of previous interaction for continuation
+        continuation_token: Token from an interaction that stopped at the per-request output
+            cap, sent back to resume that decode
         extra_headers: Additional headers
         extra_body: Additional body parameters
         timeout: Request timeout
@@ -164,6 +167,7 @@ async def acreate(
             response_format=response_format,
             response_mime_type=response_mime_type,
             previous_interaction_id=previous_interaction_id,
+            continuation_token=continuation_token,
             extra_headers=extra_headers,
             extra_body=extra_body,
             timeout=timeout,
@@ -232,6 +236,7 @@ def create(
     custom_llm_provider: str | None = None,
     *,
     agent_config: Mapping[str, object] | None = None,
+    continuation_token: str | None = None,
     **kwargs,
 ) -> (
     InteractionsAPIResponse
@@ -262,6 +267,8 @@ def create(
         response_format: JSON schema for response format
         response_mime_type: MIME type of the response
         previous_interaction_id: ID of previous interaction for continuation
+        continuation_token: Token from an interaction that stopped at the per-request output
+            cap, sent back to resume that decode
         extra_headers: Additional headers
         extra_body: Additional body parameters
         timeout: Request timeout

litellm/interactions/utils.py

@@ -20,6 +20,7 @@ INTERACTIONS_API_OPTIONAL_PARAMS: Final = {
     "response_format",
     "response_mime_type",
     "previous_interaction_id",
+    "continuation_token",
     "agent_config",
 }
 

litellm/llms/gemini/interactions/transformation.py

@@ -126,6 +126,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig):
             "response_format",
             "response_mime_type",
             "previous_interaction_id",
+            "continuation_token",
         ]
 
     def validate_environment(
@@ -206,6 +207,7 @@ class GoogleAIStudioInteractionsConfig(BaseInteractionsAPIConfig):
             "environment",
             "response_modalities",
             "previous_interaction_id",
+            "continuation_token",
         ]
         dropped_keys: Final = (
             frozenset({"agent_config"})

tests/integration/providers/test_gemini_interactions_wire.py

@@ -20,6 +20,7 @@ _MODEL_BODY: Final = {
     "response_mime_type": "application/json",
     "response_format": {"type": "object", "properties": {"plan": {"type": "string"}}},
     "previous_interaction_id": "interaction-prev-123",
+    "continuation_token": "Y29udGludWUtZnJvbS1oZXJl",
     "store": True,
     "background": True,
 }
@@ -120,6 +121,7 @@ def test_interaction_forwards_native_body_to_gemini(gateway: Gateway, prefix: st
             "store": True,
             "background": True,
             "previous_interaction_id": "interaction-prev-123",
+            "continuation_token": "Y29udGludWUtZnJvbS1oZXJl",
             "generation_config": {"temperature": 0.3},
             "response_format": [
                 {

tests/unit/interactions/test_main.py

@@ -559,29 +559,76 @@ class TestInteractionsAcreateOffline:
 
     @pytest.mark.usefixtures("fake_provider_credentials")
     @pytest.mark.asyncio
-    async def test_acreate_gemini_keeps_continuation_token_out_of_the_request(self, respx_mock: MockRouter) -> None:
+    async def test_acreate_gemini_resumes_with_the_token_it_handed_the_caller(self, respx_mock: MockRouter) -> None:
+        """The token of a cut interaction comes back on the response and reaches Google on the resend.
+
+        continuation_token is an input of CreateModelInteractionParams and an output of Interaction in
+        https://ai.google.dev/static/api/interactions.openapi.json, read 2026-10-30
+        """
+        cut: Final = {
+            "id": "interaction-cut",
+            "object": "interaction",
+            "model": "gemini-3.8-flash",
+            "status": "requires_continuation",
+            "steps": [{"type": "model_output", "content": [{"type": "text", "text": "Rivers"}]}],
+            "usage": {"input_tokens": 6, "output_tokens": 3},
+            "continuation_token": "dG9rLTE=",
+        }
+        rest: Final = {**cut, "id": "interaction-rest", "status": "completed", "continuation_token": None}
+        route: Final = respx_mock.post("https://generativelanguage.googleapis.com/v1beta/interactions").mock(
+            side_effect=[httpx.Response(200, json=cut), httpx.Response(200, json=rest)]
+        )
+        first: Final = await interactions.acreate(
+            model="gemini/gemini-3.8-flash",
+            input="Prove it",
+            api_key="gemini-offline",
+        )
+        assert first.status == "requires_continuation"
+        assert first.continuation_token == "dG9rLTE="
+        assert "continuation_token" not in json.loads(route.calls[0].request.content)
+
+        resumed: Final = await interactions.acreate(
+            model="gemini/gemini-3.8-flash",
+            input="Prove it",
+            continuation_token=first.continuation_token,
+            api_key="gemini-offline",
+        )
+        assert json.loads(route.calls.last.request.content) == {
+            "model": "gemini-3.8-flash",
+            "input": "Prove it",
+            "continuation_token": "dG9rLTE=",
+        }
+        assert resumed.status == "completed"
+
+    @pytest.mark.usefixtures("fake_provider_credentials")
+    @pytest.mark.asyncio
+    async def test_acreate_gemini_binds_the_fifteenth_positional_argument_to_extra_headers(
+        self, respx_mock: MockRouter
+    ) -> None:
+        """A caller passing model, agent, input and the eleven params after them positionally still
+        lands its headers in ``extra_headers``, so adding a create parameter cannot shift that slot."""
         route: Final = respx_mock.post("https://generativelanguage.googleapis.com/v1beta/interactions").mock(
             return_value=httpx.Response(
                 200,
                 json={
-                    "id": "interaction-offline",
+                    "id": "interaction-positional",
                     "object": "interaction",
                     "model": "gemini-3.8-flash",
                     "status": "completed",
                     "steps": [{"type": "model_output", "content": [{"type": "text", "text": "Hi"}]}],
-                    "usage": {"input_tokens": 6, "output_tokens": 3},
                 },
             )
         )
-        response: Final = await interactions.acreate(
-            model="gemini/gemini-3.8-flash",
-            input="Prove it",
-            continuation_token="dG9rLTE=",
+        await interactions.acreate(
+            "gemini/gemini-3.8-flash",
+            None,
+            "Prove it",
+            *([None] * 11),
+            {"X-Trace-Id": "trace-9"},
             api_key="gemini-offline",
         )
-        # The Interactions API answered 400 "Unknown parameter 'continuation_token'." on 2026-10-13
+        assert route.calls.last.request.headers["x-trace-id"] == "trace-9"
         assert json.loads(route.calls.last.request.content) == {"model": "gemini-3.8-flash", "input": "Prove it"}
-        assert response.status == "completed"
 
     @pytest.mark.usefixtures("fake_provider_credentials")
     @pytest.mark.asyncio