fix: stream in litellm + adk and add corresponding integration tests

Fixes https://github.com/google/adk-python/issues/1368

PiperOrigin-RevId: 772218385

src/google/adk/models/lite_llm.py

@@ -739,11 +739,12 @@ class LiteLlm(BaseLlm):
                 _message_to_generate_content_response(
                     ChatCompletionAssistantMessage(
                         role="assistant",
-                        content="",
+                        content=text,
                         tool_calls=tool_calls,
                     )
                 )
             )
+            text = ""
             function_calls.clear()
           elif finish_reason == "stop" and text:
             aggregated_llm_response = _message_to_generate_content_response(

tests/integration/models/test_litellm_no_function.py

@@ -20,12 +20,26 @@ from google.genai.types import Content
 from google.genai.types import Part
 import pytest
 
-_TEST_MODEL_NAME = "vertex_ai/meta/llama-4-maverick-17b-128e-instruct-maas"
 
+_TEST_MODEL_NAME = "vertex_ai/meta/llama-3.1-405b-instruct-maas"
 
 _SYSTEM_PROMPT = """You are a helpful assistant."""
 
 
+def get_weather(city: str) -> str:
+  """Simulates a web search. Use it get information on weather.
+
+  Args:
+      city: A string containing the location to get weather information for.
+
+  Returns:
+      A string with the simulated weather information for the queried city.
+  """
+  if "sf" in city.lower() or "san francisco" in city.lower():
+    return "It's 70 degrees and foggy."
+  return "It's 80 degrees and sunny."
+
+
 @pytest.fixture
 def oss_llm():
   return LiteLlm(model=_TEST_MODEL_NAME)
@@ -44,6 +58,48 @@ def llm_request():
   )
 
 
+@pytest.fixture
+def llm_request_with_tools():
+  return LlmRequest(
+      model=_TEST_MODEL_NAME,
+      contents=[
+          Content(
+              role="user",
+              parts=[
+                  Part.from_text(text="What is the weather in San Francisco?")
+              ],
+          )
+      ],
+      config=types.GenerateContentConfig(
+          temperature=0.1,
+          response_modalities=[types.Modality.TEXT],
+          system_instruction=_SYSTEM_PROMPT,
+          tools=[
+              types.Tool(
+                  function_declarations=[
+                      types.FunctionDeclaration(
+                          name="get_weather",
+                          description="Get the weather in a given location",
+                          parameters=types.Schema(
+                              type=types.Type.OBJECT,
+                              properties={
+                                  "city": types.Schema(
+                                      type=types.Type.STRING,
+                                      description=(
+                                          "The city to get the weather for."
+                                      ),
+                                  ),
+                              },
+                              required=["city"],
+                          ),
+                      )
+                  ]
+              )
+          ],
+      ),
+  )
+
+
 @pytest.mark.asyncio
 async def test_generate_content_async(oss_llm, llm_request):
   async for response in oss_llm.generate_content_async(llm_request):
@@ -51,10 +107,8 @@ async def test_generate_content_async(oss_llm, llm_request):
     assert response.content.parts[0].text
 
 
-# Note that, this test disabled streaming because streaming is not supported
-# properly in the current test model for now.
 @pytest.mark.asyncio
-async def test_generate_content_async_stream(oss_llm, llm_request):
+async def test_generate_content_async(oss_llm, llm_request):
   responses = [
       resp
       async for resp in oss_llm.generate_content_async(
@@ -63,3 +117,50 @@ async def test_generate_content_async_stream(oss_llm, llm_request):
   ]
   part = responses[0].content.parts[0]
   assert len(part.text) > 0
+
+
+@pytest.mark.asyncio
+async def test_generate_content_async_with_tools(
+    oss_llm, llm_request_with_tools
+):
+  responses = [
+      resp
+      async for resp in oss_llm.generate_content_async(
+          llm_request_with_tools, stream=False
+      )
+  ]
+  function_call = responses[0].content.parts[0].function_call
+  assert function_call.name == "get_weather"
+  assert function_call.args["city"] == "San Francisco"
+
+
+@pytest.mark.asyncio
+async def test_generate_content_async_stream(oss_llm, llm_request):
+  responses = [
+      resp
+      async for resp in oss_llm.generate_content_async(llm_request, stream=True)
+  ]
+  text = ""
+  for i in range(len(responses) - 1):
+    assert responses[i].partial is True
+    assert responses[i].content.parts[0].text
+    text += responses[i].content.parts[0].text
+
+  # Last message should be accumulated text
+  assert responses[-1].content.parts[0].text == text
+  assert not responses[-1].partial
+
+
+@pytest.mark.asyncio
+async def test_generate_content_async_stream_with_tools(
+    oss_llm, llm_request_with_tools
+):
+  responses = [
+      resp
+      async for resp in oss_llm.generate_content_async(
+          llm_request_with_tools, stream=True
+      )
+  ]
+  function_call = responses[-1].content.parts[0].function_call
+  assert function_call.name == "get_weather"
+  assert function_call.args["city"] == "San Francisco"

tests/integration/models/test_litellm_with_function.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 from google.adk.models import LlmRequest
-from google.adk.models import LlmResponse
 from google.adk.models.lite_llm import LiteLlm
 from google.genai import types
 from google.genai.types import Content
@@ -23,12 +22,11 @@ import pytest
 
 litellm.add_function_to_prompt = True
 
-_TEST_MODEL_NAME = "vertex_ai/meta/llama-4-maverick-17b-128e-instruct-maas"
-
+_TEST_MODEL_NAME = "vertex_ai/meta/llama-3.1-405b-instruct-maas"
 
 _SYSTEM_PROMPT = """
 You are a helpful assistant, and call tools optionally.
-If call tools, the tool format should be in json, and the tool arguments should be parsed from users inputs.
+If call tools, the tool format should be in json body, and the tool argument values should be parsed from users inputs.
 """
 
 
@@ -40,7 +38,7 @@ _FUNCTIONS = [{
         "properties": {
             "city": {
                 "type": "string",
-                "description": "The city, e.g. San Francisco",
+                "description": "The city to get the weather for.",
             },
         },
         "required": ["city"],
@@ -87,8 +85,6 @@ def llm_request():
   )
 
 
-# Note that, this test disabled streaming because streaming is not supported
-# properly in the current test model for now.
 @pytest.mark.asyncio
 async def test_generate_content_asyn_with_function(
     oss_llm_with_function, llm_request
@@ -102,3 +98,18 @@ async def test_generate_content_asyn_with_function(
   function_call = responses[0].content.parts[0].function_call
   assert function_call.name == "get_weather"
   assert function_call.args["city"] == "San Francisco"
+
+
+@pytest.mark.asyncio
+async def test_generate_content_asyn_stream_with_function(
+    oss_llm_with_function, llm_request
+):
+  responses = [
+      resp
+      async for resp in oss_llm_with_function.generate_content_async(
+          llm_request, stream=True
+      )
+  ]
+  function_call = responses[-1].content.parts[0].function_call
+  assert function_call.name == "get_weather"
+  assert function_call.args["city"] == "San Francisco"