Fix: Gemini Embeddings and LLM (#3050)
This commit is contained in:
Dev Khant
@@ -1,7 +1,8 @@
|
||||
import os
|
||||
from typing import Literal, Optional
|
||||
|
||||
import google.genai as genai
|
||||
from google import genai
|
||||
from google.genai import types
|
||||
|
||||
from mem0.configs.embeddings.base import BaseEmbedderConfig
|
||||
from mem0.embeddings.base import EmbeddingBase
|
||||
@@ -16,24 +17,23 @@ class GoogleGenAIEmbedding(EmbeddingBase):
|
||||
|
||||
api_key = self.config.api_key or os.getenv("GOOGLE_API_KEY")
|
||||
|
||||
if api_key:
|
||||
self.client = genai.Client(api_key="api_key")
|
||||
else:
|
||||
self.client = genai.Client()
|
||||
self.client = genai.Client(api_key=api_key)
|
||||
|
||||
def embed(self, text, memory_action: Optional[Literal["add", "search", "update"]] = None):
|
||||
"""
|
||||
Get the embedding for the given text using Google Generative AI.
|
||||
Args:
|
||||
text (str): The text to embed.
|
||||
memory_action (optional): The type of embedding to use. (Currently not used by Gemini for task_type)
|
||||
memory_action (optional): The type of embedding to use. Must be one of "add", "search", or "update". Defaults to None.
|
||||
Returns:
|
||||
list: The embedding vector.
|
||||
"""
|
||||
text = text.replace("\n", " ")
|
||||
|
||||
response = self.client.models.embed_content(
|
||||
model=self.config.model, content=text, output_dimensionality=self.config.embedding_dims
|
||||
)
|
||||
# Create config for embedding parameters
|
||||
config = types.EmbedContentConfig(output_dimensionality=self.config.embedding_dims)
|
||||
|
||||
return response["embedding"]
|
||||
# Call the embed_content method with the correct parameters
|
||||
response = self.client.models.embed_content(model=self.config.model, contents=text, config=config)
|
||||
|
||||
return response.embeddings[0].values
|
||||
@@ -4,11 +4,8 @@ from typing import Dict, List, Optional
|
||||
try:
|
||||
from google import genai
|
||||
from google.genai import types
|
||||
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"The 'google-generativeai' library is required. Please install it using 'pip install google-generativeai'."
|
||||
)
|
||||
raise ImportError("The 'google-genai' library is required. Please install it using 'pip install google-genai'.")
|
||||
|
||||
from mem0.configs.llms.base import BaseLlmConfig
|
||||
from mem0.llms.base import LLMBase
|
||||
@@ -19,70 +16,79 @@ class GeminiLLM(LLMBase):
|
||||
super().__init__(config)
|
||||
|
||||
if not self.config.model:
|
||||
self.config.model = "gemini-1.5-flash-latest"
|
||||
self.config.model = "gemini-2.0-flash"
|
||||
|
||||
api_key = self.config.api_key or os.getenv("GEMINI_API_KEY")
|
||||
self.client_gemini = genai.Client(
|
||||
api_key=api_key,
|
||||
)
|
||||
api_key = self.config.api_key or os.getenv("GOOGLE_API_KEY")
|
||||
self.client = genai.Client(api_key=api_key)
|
||||
|
||||
def _parse_response(self, response, tools):
|
||||
"""
|
||||
Process the response based on whether tools are used or not.
|
||||
|
||||
Args:
|
||||
response: The raw response from the API.
|
||||
response: The raw response from API.
|
||||
tools: The list of tools provided in the request.
|
||||
|
||||
Returns:
|
||||
str or dict: The processed response.
|
||||
"""
|
||||
candidate = response.candidates[0]
|
||||
content = candidate.content.parts[0].text if candidate.content.parts else None
|
||||
|
||||
if tools:
|
||||
processed_response = {
|
||||
"content": content,
|
||||
"content": None,
|
||||
"tool_calls": [],
|
||||
}
|
||||
|
||||
for part in candidate.content.parts:
|
||||
fn = getattr(part, "function_call", None)
|
||||
if fn:
|
||||
# Extract content from the first candidate
|
||||
if response.candidates and response.candidates[0].content.parts:
|
||||
for part in response.candidates[0].content.parts:
|
||||
if hasattr(part, "text") and part.text:
|
||||
processed_response["content"] = part.text
|
||||
break
|
||||
|
||||
# Extract function calls
|
||||
if response.candidates and response.candidates[0].content.parts:
|
||||
for part in response.candidates[0].content.parts:
|
||||
if hasattr(part, "function_call") and part.function_call:
|
||||
fn = part.function_call
|
||||
processed_response["tool_calls"].append(
|
||||
{
|
||||
"name": fn.name,
|
||||
"arguments": fn.args,
|
||||
"arguments": dict(fn.args) if fn.args else {},
|
||||
}
|
||||
)
|
||||
|
||||
return processed_response
|
||||
else:
|
||||
if response.candidates and response.candidates[0].content.parts:
|
||||
for part in response.candidates[0].content.parts:
|
||||
if hasattr(part, "text") and part.text:
|
||||
return part.text
|
||||
return ""
|
||||
|
||||
return content
|
||||
|
||||
def _reformat_messages(self, messages: List[Dict[str, str]]) -> List[types.Content]:
|
||||
def _reformat_messages(self, messages: List[Dict[str, str]]):
|
||||
"""
|
||||
Reformat messages for Gemini using google.genai.types.
|
||||
Reformat messages for Gemini.
|
||||
|
||||
Args:
|
||||
messages: The list of messages provided in the request.
|
||||
|
||||
Returns:
|
||||
list: A list of types.Content objects with proper role and parts.
|
||||
tuple: (system_instruction, contents_list)
|
||||
"""
|
||||
new_messages = []
|
||||
system_instruction = None
|
||||
contents = []
|
||||
|
||||
for message in messages:
|
||||
if message["role"] == "system":
|
||||
content = "THIS IS A SYSTEM PROMPT. YOU MUST OBEY THIS: " + message["content"]
|
||||
system_instruction = message["content"]
|
||||
else:
|
||||
content = message["content"]
|
||||
|
||||
new_messages.append(
|
||||
types.Content(role="model" if message["role"] == "model" else "user", parts=[types.Part(text=content)])
|
||||
content = types.Content(
|
||||
parts=[types.Part(text=message["content"])],
|
||||
role=message["role"],
|
||||
)
|
||||
contents.append(content)
|
||||
|
||||
return new_messages
|
||||
return system_instruction, contents
|
||||
|
||||
def _reformat_tools(self, tools: Optional[List[Dict]]):
|
||||
"""
|
||||
@@ -97,7 +103,6 @@ class GeminiLLM(LLMBase):
|
||||
|
||||
def remove_additional_properties(data):
|
||||
"""Recursively removes 'additionalProperties' from nested dictionaries."""
|
||||
|
||||
if isinstance(data, dict):
|
||||
filtered_dict = {
|
||||
key: remove_additional_properties(value)
|
||||
@@ -108,16 +113,21 @@ class GeminiLLM(LLMBase):
|
||||
else:
|
||||
return data
|
||||
|
||||
new_tools = []
|
||||
if tools:
|
||||
function_declarations = []
|
||||
for tool in tools:
|
||||
func = tool["function"].copy()
|
||||
new_tools.append({"function_declarations": [remove_additional_properties(func)]})
|
||||
cleaned_func = remove_additional_properties(func)
|
||||
|
||||
# TODO: temporarily ignore it to pass tests, will come back to update according to standards later.
|
||||
# return content_types.to_function_library(new_tools)
|
||||
function_declaration = types.FunctionDeclaration(
|
||||
name=cleaned_func["name"],
|
||||
description=cleaned_func.get("description", ""),
|
||||
parameters=cleaned_func.get("parameters", {}),
|
||||
)
|
||||
function_declarations.append(function_declaration)
|
||||
|
||||
return new_tools
|
||||
tool_obj = types.Tool(function_declarations=function_declarations)
|
||||
return [tool_obj]
|
||||
else:
|
||||
return None
|
||||
|
||||
@@ -141,38 +151,53 @@ class GeminiLLM(LLMBase):
|
||||
str: The generated response.
|
||||
"""
|
||||
|
||||
params = {
|
||||
# Extract system instruction and reformat messages
|
||||
system_instruction, contents = self._reformat_messages(messages)
|
||||
|
||||
# Prepare generation config
|
||||
config_params = {
|
||||
"temperature": self.config.temperature,
|
||||
"max_output_tokens": self.config.max_tokens,
|
||||
"top_p": self.config.top_p,
|
||||
}
|
||||
|
||||
if response_format is not None and response_format["type"] == "json_object":
|
||||
params["response_mime_type"] = "application/json"
|
||||
if "schema" in response_format:
|
||||
params["response_schema"] = response_format["schema"]
|
||||
# Add system instruction to config if present
|
||||
if system_instruction:
|
||||
config_params["system_instruction"] = system_instruction
|
||||
|
||||
|
||||
if response_format is not None and response_format["type"] == "json_object":
|
||||
config_params["response_mime_type"] = "application/json"
|
||||
if "schema" in response_format:
|
||||
config_params["response_schema"] = response_format["schema"]
|
||||
|
||||
if tools:
|
||||
formatted_tools = self._reformat_tools(tools)
|
||||
config_params["tools"] = formatted_tools
|
||||
|
||||
|
||||
tool_config = None
|
||||
if tool_choice:
|
||||
if tool_choice == "auto":
|
||||
mode = types.FunctionCallingConfigMode.AUTO
|
||||
elif tool_choice == "any":
|
||||
mode = types.FunctionCallingConfigMode.ANY
|
||||
else:
|
||||
mode = types.FunctionCallingConfigMode.NONE
|
||||
|
||||
tool_config = types.ToolConfig(
|
||||
function_calling_config=types.FunctionCallingConfig(
|
||||
mode=tool_choice.upper(), # Assuming 'any' should become 'ANY', etc.
|
||||
allowed_function_names=[tool["function"]["name"] for tool in tools]
|
||||
if tool_choice == "any"
|
||||
else None,
|
||||
)
|
||||
)
|
||||
|
||||
response = self.client_gemini.models.generate_content(
|
||||
model=self.config.model,
|
||||
contents=self._reformat_messages(messages),
|
||||
config=types.GenerateContentConfig(
|
||||
temperature=self.config.temperature,
|
||||
max_output_tokens=self.config.max_tokens,
|
||||
top_p=self.config.top_p,
|
||||
tools=self._reformat_tools(tools),
|
||||
tool_config=tool_config,
|
||||
mode=mode,
|
||||
allowed_function_names=(
|
||||
[tool["function"]["name"] for tool in tools] if tool_choice == "any" else None
|
||||
),
|
||||
)
|
||||
)
|
||||
config_params["tool_config"] = tool_config
|
||||
|
||||
generation_config = types.GenerateContentConfig(**config_params)
|
||||
|
||||
response = self.client.models.generate_content(
|
||||
model=self.config.model, contents=contents, config=generation_config
|
||||
)
|
||||
|
||||
return self._parse_response(response, tools)
|
||||
|
||||
Reference in New Issue
Block a user