llama_index provides a function called create_llama_agent that acts as a light wrapper around the initialize_agent function in langchain. This function is used to load an agent executor given a Llama Toolkit and LLM (Language Model). The create_llama_agent function accepts an optional agent parameter, which is a string specifying the agent type to use. If both agent and agent_path are None, it will default to the zero-shot-react-description agent type.create_llama_agent:gpt_index.langchain_helpers.agents.create_llama_agent(
toolkit: LlamaToolkit,
llm: BaseLLM,
agent: Optional[AgentType] = None,
callback_manager: Optional[BaseCallbackManager] = None,
agent_path: Optional[str] = None,
agent_kwargs: Optional[dict] = None,
**kwargs: Any
) -> AgentExecutorllama_index handles default langchain prompts when using initialize_agent. For more details, you may want to refer to the Llama Index documentation (https://gpt-index.readthedocs.io/en/latest/) or the GitHub repository (https://github.com/jerryjliu/llama_index).Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:create_llama_agent wrapper function provided by llama_index. To do this, you can pass the custom agent type as a string to the agent parameter when calling the function. The available agent types are:create_llama_agent:gpt_index.langchain_helpers.agents.create_llama_agent(
toolkit: LlamaToolkit,
llm: BaseLLM,
agent: Optional[AgentType] = None,
callback_manager: Optional[BaseCallbackManager] = None,
agent_path: Optional[str] = None,
agent_kwargs: Optional[dict] = None,
**kwargs: Any
) -> AgentExecutoragent_path parameter.Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:create_llama_agent function. However, I can show you the function signature and parameters:gpt_index.langchain_helpers.agents.create_llama_agent(
toolkit: LlamaToolkit,
llm: BaseLLM,
agent: Optional[AgentType] = None,
callback_manager: Optional[BaseCallbackManager] = None,
agent_path: Optional[str] = None,
agent_kwargs: Optional[dict] = None,
**kwargs: Any
) -> AgentExecutorNote: you can ask me a follow up question by @ mentioning me again :speech_balloon:create_llama_chat_agent function from the provided information extracts:from gpt_index.langchain_helpers.agents import create_llama_chat_agent
from gpt_index.langchain_helpers.memory_wrapper import ConversationBufferMemory
from gpt_index.langchain_helpers.llm import OpenAI
memory = ConversationBufferMemory(memory_key="chat_history")
llm = OpenAI(temperature=0)
agent_chain = create_llama_chat_agent(
toolkit,
llm,
memory=memory,
verbose=True
)ConversationBufferMemory is created to store the chat history, and an OpenAI instance is created as the language model. The create_llama_chat_agent function is then called with the toolkit, llm, and memory parameters to create a Langchain chatbot agent.Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:LlamaToolkit using the IndexToolConfig and GraphToolConfig from the provided information extracts:from llama_index.langchain_helpers.agents import LlamaToolkit, IndexToolConfig, GraphToolConfig
# Define toolkit
index_configs = []
for y in range(2019, 2023):
tool_config = IndexToolConfig(
index=index_set[y],
name=f"Vector Index {y}",
description=f"useful for when you want to answer queries about the {y} SEC 10-K for Uber",
index_query_kwargs={"similarity_top_k": 3},
tool_kwargs={"return_direct": True}
)
index_configs.append(tool_config)
graph_config = GraphToolConfig(
graph=graph,
name=f"Graph Index",
description="useful for when you want to answer queries that require analyzing multiple SEC 10-K documents for Uber.",
query_configs=query_configs,
tool_kwargs={"return_direct": True}
)
toolkit = LlamaToolkit(
index_configs=index_configs,
graph_configs=[graph_config]
)IndexToolConfig objects are created for each index (corresponding to a given year), and a GraphToolConfig object is created for the graph. These configurations are then combined into a LlamaToolkit.Note: you can ask me a follow up question by @ mentioning me again :speech_balloon: