fix(vertex_ai.py): support passing in result of tool call to vertex

Fixes https://github.com/BerriAI/litellm/issues/3709

litellm/llms/prompt_templates/factory.py

@@ -12,6 +12,7 @@ from typing import (
     Sequence,
 )
 import litellm
+import litellm.types
 from litellm.types.completion import (
     ChatCompletionUserMessageParam,
     ChatCompletionSystemMessageParam,
@@ -20,9 +21,12 @@ from litellm.types.completion import (
     ChatCompletionMessageToolCallParam,
     ChatCompletionToolMessageParam,
 )
+import litellm.types.llms
 from litellm.types.llms.anthropic import *
 import uuid
 
+import litellm.types.llms.vertex_ai
+
 
 def default_pt(messages):
     return " ".join(message["content"] for message in messages)
@@ -841,6 +845,175 @@ def anthropic_messages_pt_xml(messages: list):
 # ------------------------------------------------------------------------------
 
 
+def infer_protocol_value(
+    value: Any,
+) -> Literal[
+    "string_value",
+    "number_value",
+    "bool_value",
+    "struct_value",
+    "list_value",
+    "null_value",
+    "unknown",
+]:
+    if value is None:
+        return "null_value"
+    if isinstance(value, int) or isinstance(value, float):
+        return "number_value"
+    if isinstance(value, str):
+        return "string_value"
+    if isinstance(value, bool):
+        return "bool_value"
+    if isinstance(value, dict):
+        return "struct_value"
+    if isinstance(value, list):
+        return "list_value"
+
+    return "unknown"
+
+
+def convert_to_gemini_tool_call_invoke(
+    tool_calls: list,
+) -> List[litellm.types.llms.vertex_ai.PartType]:
+    """
+    OpenAI tool invokes:
+    {
+      "role": "assistant",
+      "content": null,
+      "tool_calls": [
+        {
+          "id": "call_abc123",
+          "type": "function",
+          "function": {
+            "name": "get_current_weather",
+            "arguments": "{\n\"location\": \"Boston, MA\"\n}"
+          }
+        }
+      ]
+    },
+    """
+    """
+    Gemini tool call invokes: - https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/function-calling#submit-api-output
+    content {
+        role: "model"
+        parts [
+        {
+            function_call {
+            name: "get_current_weather"
+            args {
+                fields {
+                    key: "unit"
+                    value {
+                    string_value: "fahrenheit"
+                    }
+                }
+                fields {
+                    key: "predicted_temperature"
+                    value {
+                    number_value: 45
+                    }
+                }
+                fields {
+                    key: "location"
+                    value {
+                    string_value: "Boston, MA"
+                    }
+                }
+            }
+        },
+        {
+            function_call {
+            name: "get_current_weather"
+            args {
+                fields {
+                key: "location"
+                value {
+                    string_value: "San Francisco"
+                }
+                }
+            }
+            }
+        }
+        ]
+    }
+    """
+
+    """
+    - json.load the arguments 
+    - iterate through arguments -> create a FunctionCallArgs for each field
+    """
+    try:
+        _parts_list: List[litellm.types.llms.vertex_ai.PartType] = []
+        for tool in tool_calls:
+            if "function" in tool:
+                name = tool["function"].get("name", "")
+                arguments = tool["function"].get("arguments", "")
+                arguments_dict = json.loads(arguments)
+                for k, v in arguments_dict.items():
+                    inferred_protocol_value = infer_protocol_value(value=v)
+                    _field = litellm.types.llms.vertex_ai.Field(
+                        key=k, value={inferred_protocol_value: v}
+                    )
+                    _fields = litellm.types.llms.vertex_ai.FunctionCallArgs(
+                        fields=_field
+                    )
+                    function_call = litellm.types.llms.vertex_ai.FunctionCall(
+                        name=name,
+                        args=_fields,
+                    )
+                _parts_list.append(
+                    litellm.types.llms.vertex_ai.PartType(function_call=function_call)
+                )
+        return _parts_list
+    except Exception as e:
+        raise Exception(
+            "Unable to convert openai tool calls={} to gemini tool calls. Received error={}".format(
+                tool_calls, str(e)
+            )
+        )
+
+
+def convert_to_gemini_tool_call_result(
+    message: dict,
+) -> litellm.types.llms.vertex_ai.PartType:
+    """
+    OpenAI message with a tool result looks like:
+    {
+        "tool_call_id": "tool_1",
+        "role": "tool",
+        "name": "get_current_weather",
+        "content": "function result goes here",
+    },
+
+    OpenAI message with a function call result looks like:
+    {
+        "role": "function",
+        "name": "get_current_weather",
+        "content": "function result goes here",
+    }
+    """
+    content = message.get("content", "")
+    name = message.get("name", "")
+
+    # We can't determine from openai message format whether it's a successful or
+    # error call result so default to the successful result template
+    inferred_content_value = infer_protocol_value(value=content)
+
+    _field = litellm.types.llms.vertex_ai.Field(
+        key="content", value={inferred_content_value: content}
+    )
+
+    _function_call_args = litellm.types.llms.vertex_ai.FunctionCallArgs(fields=_field)
+
+    _function_response = litellm.types.llms.vertex_ai.FunctionResponse(
+        name=name, response=_function_call_args
+    )
+
+    _part = litellm.types.llms.vertex_ai.PartType(function_response=_function_response)
+
+    return _part
+
+
 def convert_to_anthropic_tool_result(message: dict) -> dict:
     """
     OpenAI message with a tool result looks like:

litellm/llms/vertex_ai.py

@@ -3,10 +3,15 @@ import json
 from enum import Enum
 import requests  # type: ignore
 import time
-from typing import Callable, Optional, Union, List
+from typing import Callable, Optional, Union, List, Literal
 from litellm.utils import ModelResponse, Usage, CustomStreamWrapper, map_finish_reason
 import litellm, uuid
 import httpx, inspect  # type: ignore
+from litellm.types.llms.vertex_ai import *
+from litellm.llms.prompt_templates.factory import (
+    convert_to_gemini_tool_call_result,
+    convert_to_gemini_tool_call_invoke,
+)
 
 
 class VertexAIError(Exception):
@@ -283,6 +288,129 @@ def _load_image_from_url(image_url: str):
     return Image.from_bytes(data=image_bytes)
 
 
+def _convert_gemini_role(role: str) -> Literal["user", "model"]:
+    if role == "user":
+        return "user"
+    else:
+        return "model"
+
+
+def _process_gemini_image(image_url: str):
+    try:
+        import vertexai
+    except:
+        raise VertexAIError(
+            status_code=400,
+            message="vertexai import failed please run `pip install google-cloud-aiplatform`",
+        )
+    from vertexai.preview.generative_models import Part
+
+    if "gs://" in image_url:
+        # Case 1: Images with Cloud Storage URIs
+        # The supported MIME types for images include image/png and image/jpeg.
+        part_mime = "image/png" if "png" in image_url else "image/jpeg"
+        google_clooud_part = Part.from_uri(image_url, mime_type=part_mime)
+        return google_clooud_part
+    elif "https:/" in image_url:
+        # Case 2: Images with direct links
+        image = _load_image_from_url(image_url)
+        return image
+    elif ".mp4" in image_url and "gs://" in image_url:
+        # Case 3: Videos with Cloud Storage URIs
+        part_mime = "video/mp4"
+        google_clooud_part = Part.from_uri(image_url, mime_type=part_mime)
+        return google_clooud_part
+    elif "base64" in image_url:
+        # Case 4: Images with base64 encoding
+        import base64, re
+
+        # base 64 is passed as data:image/jpeg;base64,<base-64-encoded-image>
+        image_metadata, img_without_base_64 = image_url.split(",")
+
+        # read mime_type from img_without_base_64=data:image/jpeg;base64
+        # Extract MIME type using regular expression
+        mime_type_match = re.match(r"data:(.*?);base64", image_metadata)
+
+        if mime_type_match:
+            mime_type = mime_type_match.group(1)
+        else:
+            mime_type = "image/jpeg"
+        decoded_img = base64.b64decode(img_without_base_64)
+        processed_image = Part.from_data(data=decoded_img, mime_type=mime_type)
+        return processed_image
+
+
+def _gemini_convert_messages_text(messages: list) -> List[ContentType]:
+    """
+    Converts given messages from OpenAI format to Gemini format
+
+    - Parts must be iterable
+    - Roles must alternate b/w 'user' and 'model' (same as anthropic -> merge consecutive roles)
+    - Please ensure that function response turn comes immediately after a function call turn
+    """
+    user_message_types = {"user", "system"}
+    contents: List[ContentType] = []
+
+    msg_i = 0
+    while msg_i < len(messages):
+        user_content: List[PartType] = []
+        init_msg_i = msg_i
+        ## MERGE CONSECUTIVE USER CONTENT ##
+        while msg_i < len(messages) and messages[msg_i]["role"] in user_message_types:
+            if isinstance(messages[msg_i]["content"], list):
+                _parts: List[PartType] = []
+                for element in messages[msg_i]["content"]:
+                    if isinstance(element, dict):
+                        if element["type"] == "text":
+                            _part = PartType(text=element["text"])
+                            _parts.append(_part)
+                        elif element["type"] == "image_url":
+                            image_url = element["image_url"]["url"]
+                            _part = _process_gemini_image(image_url=image_url)
+                            _parts.append(_part)  # type: ignore
+                user_content.extend(_parts)
+            else:
+                _part = PartType(text=messages[msg_i]["content"])
+                user_content.append(_part)
+
+            msg_i += 1
+
+        if user_content:
+            contents.append(ContentType(role="user", parts=user_content))
+        assistant_content = []
+        ## MERGE CONSECUTIVE ASSISTANT CONTENT ##
+        while msg_i < len(messages) and messages[msg_i]["role"] == "assistant":
+            assistant_text = (
+                messages[msg_i].get("content") or ""
+            )  # either string or none
+            if assistant_text:
+                assistant_content.append(PartType(text=assistant_text))
+            if messages[msg_i].get(
+                "tool_calls", []
+            ):  # support assistant tool invoke convertion
+                assistant_content.extend(
+                    convert_to_gemini_tool_call_invoke(messages[msg_i]["tool_calls"])
+                )
+            msg_i += 1
+
+        if assistant_content:
+            contents.append(ContentType(role="model", parts=assistant_content))
+
+        ## APPEND TOOL CALL MESSAGES ##
+        if messages[msg_i]["role"] == "tool":
+            _part = convert_to_gemini_tool_call_result(messages[msg_i])
+            contents.append(ContentType(parts=[_part]))  # type: ignore
+            msg_i += 1
+        if msg_i == init_msg_i:  # prevent infinite loops
+            raise Exception(
+                "Invalid Message passed in - {}. File an issue https://github.com/BerriAI/litellm/issues".format(
+                    messages[msg_i]
+                )
+            )
+
+    return contents
+
+
 def _gemini_vision_convert_messages(messages: list):
     """
     Converts given messages for GPT-4 Vision to Gemini format.
@@ -574,11 +702,9 @@ def completion(
             print_verbose("\nMaking VertexAI Gemini Pro / Pro Vision Call")
             print_verbose(f"\nProcessing input messages = {messages}")
             tools = optional_params.pop("tools", None)
-            prompt, images = _gemini_vision_convert_messages(messages=messages)
+            content = _gemini_convert_messages_text(messages=messages)
-            content = [prompt] + images
             stream = optional_params.pop("stream", False)
             if stream == True:
-
                 request_str += f"response = llm_model.generate_content({content}, generation_config=GenerationConfig(**{optional_params}), safety_settings={safety_settings}, stream={stream})\n"
                 logging_obj.pre_call(
                     input=prompt,
@@ -590,7 +716,7 @@ def completion(
                 )
 
                 model_response = llm_model.generate_content(
-                    contents=content,
+                    contents={"content": content},
                     generation_config=optional_params,
                     safety_settings=safety_settings,
                     stream=True,

litellm/tests/log.txt

@@ -1,15 +1,21 @@
 ============================= test session starts ==============================
-platform darwin -- Python 3.11.9, pytest-7.3.1, pluggy-1.3.0
+platform darwin -- Python 3.11.4, pytest-8.2.0, pluggy-1.5.0
-rootdir: /Users/krrishdholakia/Documents/litellm/litellm/tests
+rootdir: /Users/krrishdholakia/Documents/litellm
-plugins: timeout-2.2.0, asyncio-0.23.2, anyio-3.7.1, xdist-3.3.1
+configfile: pyproject.toml
+plugins: asyncio-0.23.6, mock-3.14.0, anyio-4.2.0
 asyncio: mode=Mode.STRICT
 collected 1 item
 
-test_router_timeout.py .                                                 [100%]
+test_amazing_vertex_completion.py Success: model=gemini-1.5-pro-preview-0514 in model_cost_map
+prompt_tokens=88; completion_tokens=31
+Returned custom cost for model=gemini-1.5-pro-preview-0514 - prompt_tokens_cost_usd_dollar: 5.5e-05, completion_tokens_cost_usd_dollar: 5.8125e-05
+final cost: 0.000113125; prompt_tokens_cost_usd_dollar: 5.5e-05; completion_tokens_cost_usd_dollar: 5.8125e-05
+success callbacks: []
+.                                      [100%]
 
 =============================== warnings summary ===============================
-../../../../../../opt/homebrew/lib/python3.11/site-packages/pydantic/_internal/_config.py:284: 25 warnings
+../proxy/myenv/lib/python3.11/site-packages/pydantic/_internal/_config.py:284: 25 warnings
-  /opt/homebrew/lib/python3.11/site-packages/pydantic/_internal/_config.py:284: PydanticDeprecatedSince20: Support for class-based `config` is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.7/migration/
+  /Users/krrishdholakia/Documents/litellm/litellm/proxy/myenv/lib/python3.11/site-packages/pydantic/_internal/_config.py:284: PydanticDeprecatedSince20: Support for class-based `config` is deprecated, use ConfigDict instead. Deprecated in Pydantic V2.0 to be removed in V3.0. See Pydantic V2 Migration Guide at https://errors.pydantic.dev/2.7/migration/
     warnings.warn(DEPRECATION_MESSAGE, DeprecationWarning)
 
 ../proxy/_types.py:255
@@ -68,9 +74,5 @@ test_router_timeout.py .                                                 [100%]
   /Users/krrishdholakia/Documents/litellm/litellm/utils.py:60: DeprecationWarning: open_text is deprecated. Use files() instead. Refer to https://importlib-resources.readthedocs.io/en/latest/using.html#migrating-from-legacy for migration advice.
     with resources.open_text("litellm.llms.tokenizers", "anthropic_tokenizer.json") as f:
 
-test_router_timeout.py::test_router_timeouts_bedrock
-  /opt/homebrew/lib/python3.11/site-packages/httpx/_content.py:204: DeprecationWarning: Use 'content=<...>' to upload raw bytes/text content.
-    warnings.warn(message, DeprecationWarning)
-
 -- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
-======================== 1 passed, 40 warnings in 0.99s ========================
+======================== 1 passed, 39 warnings in 2.92s ========================

litellm/tests/test_amazing_vertex_completion.py

@@ -16,6 +16,7 @@ from litellm.tests.test_streaming import streaming_format_tests
 import json
 import os
 import tempfile
+from litellm.llms.vertex_ai import _gemini_convert_messages_text
 
 litellm.num_retries = 3
 litellm.cache = None
@@ -98,7 +99,7 @@ def load_vertex_ai_credentials():
 
 
 @pytest.mark.asyncio
-async def get_response():
+async def test_get_response():
     load_vertex_ai_credentials()
     prompt = '\ndef count_nums(arr):\n    """\n    Write a function count_nums which takes an array of integers and returns\n    the number of elements which has a sum of digits > 0.\n    If a number is negative, then its first signed digit will be negative:\n    e.g. -123 has signed digits -1, 2, and 3.\n    >>> count_nums([]) == 0\n    >>> count_nums([-1, 11, -11]) == 1\n    >>> count_nums([1, 1, 2]) == 3\n    """\n'
     try:
@@ -589,35 +590,73 @@ def test_gemini_pro_vision_base64():
             pytest.fail(f"An exception occurred - {str(e)}")
 
 
-@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.parametrize("sync_mode", [True])
 @pytest.mark.asyncio
 async def test_gemini_pro_function_calling(sync_mode):
     try:
         load_vertex_ai_credentials()
-        data = {
+        litellm.set_verbose = True
-            "model": "vertex_ai/gemini-pro",
+
-            "messages": [
+        messages = [
+            {
+                "role": "system",
+                "content": "Your name is Litellm Bot, you are a helpful assistant",
+            },
+            # User asks for their name and weather in San Francisco
             {
                 "role": "user",
-                    "content": "Call the submit_cities function with San Francisco and New York",
+                "content": "Hello, what is your name and can you tell me the weather?",
+            },
+            # Assistant replies with a tool call
+            {
+                "role": "assistant",
+                "content": "",
+                "tool_calls": [
+                    {
+                        "id": "call_123",
+                        "type": "function",
+                        "index": 0,
+                        "function": {
+                            "name": "get_weather",
+                            "arguments": '{"location":"San Francisco, CA"}',
+                        },
                     }
                 ],
-            "tools": [
+            },
+            # The result of the tool call is added to the history
+            {
+                "role": "tool",
+                "tool_call_id": "call_123",
+                "name": "get_weather",
+                "content": "27 degrees celsius and clear in San Francisco, CA",
+            },
+            # Now the assistant can reply with the result of the tool call.
+        ]
+
+        tools = [
             {
                 "type": "function",
                 "function": {
-                        "name": "submit_cities",
+                    "name": "get_weather",
-                        "description": "Submits a list of cities",
+                    "description": "Get the current weather in a given location",
                     "parameters": {
                         "type": "object",
                         "properties": {
-                                "cities": {"type": "array", "items": {"type": "string"}}
+                            "location": {
+                                "type": "string",
+                                "description": "The city and state, e.g. San Francisco, CA",
+                            }
                         },
-                            "required": ["cities"],
+                        "required": ["location"],
                     },
                 },
             }
-            ],
+        ]
+
+        data = {
+            "model": "vertex_ai/gemini-1.5-pro-preview-0514",
+            "messages": messages,
+            "tools": tools,
         }
         if sync_mode:
             response = litellm.completion(**data)
@@ -712,7 +751,7 @@ async def test_gemini_pro_async_function_calling():
                 "type": "function",
                 "function": {
                     "name": "get_current_weather",
-                    "description": "Get the current weather in a given location",
+                    "description": "Get the current weather in a given location. Response with a prediction on temperature as well.",
                     "parameters": {
                         "type": "object",
                         "properties": {
@@ -724,8 +763,9 @@ async def test_gemini_pro_async_function_calling():
                                 "type": "string",
                                 "enum": ["celsius", "fahrenheit"],
                             },
+                            "predicted_temperature": {"type": "integer"},
                         },
-                        "required": ["location"],
+                        "required": ["location", "predicted_temperature"],
                     },
                 },
             }
@@ -733,7 +773,7 @@ async def test_gemini_pro_async_function_calling():
         messages = [
             {
                 "role": "user",
-                "content": "What's the weather like in Boston today in fahrenheit?",
+                "content": "What's the weather like in Boston today in fahrenheit, give me a prediction?",
             }
         ]
         completion = await litellm.acompletion(
@@ -742,8 +782,10 @@ async def test_gemini_pro_async_function_calling():
         print(f"completion: {completion}")
         assert completion.choices[0].message.content is None
         assert len(completion.choices[0].message.tool_calls) == 1
-    except litellm.APIError as e:
+
-        pass
+        raise Exception
+    # except litellm.APIError as e:
+    #     pass
     except litellm.RateLimitError as e:
         pass
     except Exception as e:
@@ -893,3 +935,46 @@ async def test_vertexai_aembedding():
 #         traceback.print_exc()
 #         raise e
 # test_gemini_pro_vision_async()
+
+
+def test_prompt_factory():
+    messages = [
+        {
+            "role": "system",
+            "content": "Your name is Litellm Bot, you are a helpful assistant",
+        },
+        # User asks for their name and weather in San Francisco
+        {
+            "role": "user",
+            "content": "Hello, what is your name and can you tell me the weather?",
+        },
+        # Assistant replies with a tool call
+        {
+            "role": "assistant",
+            "content": "",
+            "tool_calls": [
+                {
+                    "id": "call_123",
+                    "type": "function",
+                    "index": 0,
+                    "function": {
+                        "name": "get_weather",
+                        "arguments": '{"location":"San Francisco, CA"}',
+                    },
+                }
+            ],
+        },
+        # The result of the tool call is added to the history
+        {
+            "role": "tool",
+            "tool_call_id": "call_123",
+            "name": "get_weather",
+            "content": "27 degrees celsius and clear in San Francisco, CA",
+        },
+        # Now the assistant can reply with the result of the tool call.
+    ]
+
+    translated_messages = _gemini_convert_messages_text(messages=messages)
+
+    print(f"\n\ntranslated_messages: {translated_messages}\ntranslated_messages")
+    raise Exception

litellm/types/llms/vertex_ai.py

@@ -0,0 +1,53 @@
+from typing import TypedDict, Any, Union, Optional, List, Literal
+import json
+from typing_extensions import (
+    Self,
+    Protocol,
+    TypeGuard,
+    override,
+    get_origin,
+    runtime_checkable,
+    Required,
+)
+
+
+class Field(TypedDict):
+    key: str
+    value: dict[str, Any]
+
+
+class FunctionCallArgs(TypedDict):
+    fields: Field
+
+
+class FunctionResponse(TypedDict):
+    name: str
+    response: FunctionCallArgs
+
+
+class FunctionCall(TypedDict):
+    name: str
+    args: FunctionCallArgs
+
+
+class FileDataType(TypedDict):
+    mime_type: str
+    file_uri: str  # the cloud storage uri of storing this file
+
+
+class BlobType(TypedDict):
+    mime_type: Required[str]
+    data: Required[bytes]
+
+
+class PartType(TypedDict, total=False):
+    text: str
+    inline_data: BlobType
+    file_data: FileDataType
+    function_call: FunctionCall
+    function_response: FunctionResponse
+
+
+class ContentType(TypedDict, total=False):
+    role: Literal["user", "model"]
+    parts: Required[List[PartType]]