fix(gemini.py): support streaming

litellm/llms/gemini.py

@@ -120,9 +120,7 @@ def completion(
 
     ## Load Config
     inference_params = copy.deepcopy(optional_params)
-    inference_params.pop(
+    stream = inference_params.pop("stream", None)
-        "stream", None
-    )  # palm does not support streaming, so we handle this by fake streaming in main.py
     config = litellm.GeminiConfig.get_config()
     for k, v in config.items():
         if (
@@ -139,10 +137,18 @@ def completion(
     ## COMPLETION CALL
     try:
         _model = genai.GenerativeModel(f"models/{model}")
-        response = _model.generate_content(
+        if stream != True:
-            contents=prompt,
+            response = _model.generate_content(
-            generation_config=genai.types.GenerationConfig(**inference_params),
+                contents=prompt,
-        )
+                generation_config=genai.types.GenerationConfig(**inference_params),
+            )
+        else:
+            response = _model.generate_content(
+                contents=prompt,
+                generation_config=genai.types.GenerationConfig(**inference_params),
+                stream=True,
+            )
+            return response
     except Exception as e:
         raise GeminiError(
             message=str(e),
@@ -177,16 +183,20 @@ def completion(
 
     try:
         completion_response = model_response["choices"][0]["message"].get("content")
-        if completion_response is None: 
+        if completion_response is None:
             raise Exception
     except:
         original_response = f"response: {response}"
-        if hasattr(response, "candidates"): 
+        if hasattr(response, "candidates"):
             original_response = f"response: {response.candidates}"
-            if "SAFETY" in original_response: 
+            if "SAFETY" in original_response:
-                original_response += "\nThe candidate content was flagged for safety reasons."
+                original_response += (
+                    "\nThe candidate content was flagged for safety reasons."
+                )
             elif "RECITATION" in original_response:
-                original_response += "\nThe candidate content was flagged for recitation reasons."
+                original_response += (
+                    "\nThe candidate content was flagged for recitation reasons."
+                )
         raise GeminiError(
             status_code=400,
             message=f"No response received. Original response - {original_response}",

litellm/main.py

@@ -1382,6 +1382,18 @@ def completion(
                 acompletion=acompletion,
                 custom_prompt_dict=custom_prompt_dict,
             )
+            if (
+                "stream" in optional_params
+                and optional_params["stream"] == True
+                and acompletion == False
+            ):
+                response = CustomStreamWrapper(
+                    iter(model_response),
+                    model,
+                    custom_llm_provider="gemini",
+                    logging_obj=logging,
+                )
+                return response
             response = model_response
         elif custom_llm_provider == "vertex_ai":
             vertex_ai_project = litellm.vertex_project or get_secret("VERTEXAI_PROJECT")

litellm/tests/test_streaming.py

@@ -398,6 +398,36 @@ def test_completion_palm_stream():
 # test_completion_palm_stream()
 
 
+def test_completion_gemini_stream():
+    try:
+        litellm.set_verbose = False
+        print("Streaming gemini response")
+        messages = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {
+                "role": "user",
+                "content": "how does a court case get to the Supreme Court?",
+            },
+        ]
+        print("testing gemini streaming")
+        response = completion(model="gemini/gemini-pro", messages=messages, stream=True)
+        print(f"type of response at the top: {response}")
+        complete_response = ""
+        # Add any assertions here to check the response
+        for idx, chunk in enumerate(response):
+            print(chunk)
+            # print(chunk.choices[0].delta)
+            chunk, finished = streaming_format_tests(idx, chunk)
+            if finished:
+                break
+            complete_response += chunk
+        if complete_response.strip() == "":
+            raise Exception("Empty response received")
+        print(f"completion_response: {complete_response}")
+    except Exception as e:
+        pytest.fail(f"Error occurred: {e}")
+
+
 def test_completion_mistral_api_stream():
     try:
         litellm.set_verbose = True

litellm/utils.py

@@ -7622,7 +7622,9 @@ class CustomStreamWrapper:
                             raise Exception("An unknown error occurred with the stream")
                         model_response.choices[0].finish_reason = "stop"
                         self.sent_last_chunk = True
-            elif self.custom_llm_provider and self.custom_llm_provider == "vertex_ai":
+            elif self.custom_llm_provider == "gemini":
+                completion_obj["content"] = chunk.text
+            elif self.custom_llm_provider and (self.custom_llm_provider == "vertex_ai"):
                 try:
                     # print(chunk)
                     if hasattr(chunk, "text"):