How can I create a simple agent in Llama_index .. just like CrewAI?

[mraguth]

In CrewAI, I use this code to create a simple agent. However, I like to create a similar agent in llama_index without using CrewAI. Is that possible? Can you help us in giving sample code?

data_scientist = Agent(
    role='Senior Data Scientist',
    goal='Develop the best machine learning model according to intended goal',
    backstory=""""As a Senior Data Scientist, you are expert in Python pandas, you interpret complex data to provide strategic insights and support decision-making.
        You act on 'llamaindex query response data' only. You are not going to analyze anything as a first hand. Or you will wait for llamaindex query response first.
        You cleanse and organize the query response data and come up with your own findings.
        By applying statistical analysis and predictive modeling, you uncover trends, patterns, 
        and correlations that drive operational improvements and innovative solutions on llamaindex query response data.
        Your proficiency python pandas allows for effective database interactions, 
        Staying updated with the latest analytics techniques, you continuously refine your models for enhanced performance, significantly contributing to the organization's data-driven culture.
        Do not make up numbers and end the process after your final answer.
        """,
    verbose=True,
    allow_delegation=False,
    llm=cllm,
    tools=[query_tool]
)

[dosubot[bot]]

To create a simple agent in LlamaIndex similar to the one you created in CrewAI, you can follow these steps:

1. Set up your environment:

   • Create a new Python virtual environment using poetry or any other virtual environment manager.
   • Install the necessary dependencies:

   poetry init
   poetry shell
   pip install llama-index python-dotenv

2. Set up your OpenAI API key:

   • Create a .env file in the root of your project and add your OpenAI API key:

   OPENAI_API_KEY=sk-proj-xxxx

3. Import dependencies:

   from dotenv import load_dotenv
   
   load_dotenv()
   from llama_index.core.agent import ReActAgent
   from llama_index.llms.openai import OpenAI
   from llama_index.core.tools import FunctionTool

4. Create basic tools:

   def query_tool(query: str) -> str:
       """Simulate a query tool that processes a query and returns a response"""
       # Implement your query processing logic here
       return f"Processed query: {query}"
   
   query_tool_instance = FunctionTool.from_defaults(fn=query_tool)

5. Initialize the LLM:

   llm = OpenAI(model="gpt-3.5-turbo", temperature=0)

6. Initialize the agent:

   agent = ReActAgent.from_tools([query_tool_instance], llm=llm, verbose=True)

7. Define the agent's role and backstory:

   agent.role = 'Senior Data Scientist'
   agent.goal = 'Develop the best machine learning model according to intended goal'
   agent.backstory = (
       "As a Senior Data Scientist, you are expert in Python pandas, you interpret complex data to provide strategic insights and support decision-making. "
       "You act on 'llamaindex query response data' only. You are not going to analyze anything as a first hand. Or you will wait for llamaindex query response first. "
       "You cleanse and organize the query response data and come up with your own findings. "
       "By applying statistical analysis and predictive modeling, you uncover trends, patterns, and correlations that drive operational improvements and innovative solutions on llamaindex query response data. "
       "Your proficiency in Python pandas allows for effective database interactions. "
       "Staying updated with the latest analytics techniques, you continuously refine your models for enhanced performance, significantly contributing to the organization's data-driven culture. "
       "Do not make up numbers and end the process after your final answer."
   )

8. Ask a question:

   response = agent.chat("What trends can you identify from the latest query response data?")
   print(response)

This code sets up a simple agent using the ReActAgent class, initializes the agent with a query tool, and defines the agent's role and backstory. The agent will use the provided tool to process queries and respond accordingly ^[1][2][3][4].

To continue talking to Dosu, mention @dosu.

[mraguth]

@dosu What about I don't have any query processing logic and I like to get dynamically from llm? Can I leave that blank in your above response?

def query_tool(query: str) -> str:
    """Simulate a query tool that processes a query and returns a response"""
    # Implement your query processing logic here
    return f"Processed query: {query}"

[dosubot[bot]]

Yes, you can create an agent in LlamaIndex without any predefined query processing logic, allowing the LLM to handle it dynamically. Here’s how you can do it:

1. Set up your environment and install dependencies:

   poetry init
   poetry shell
   pip install llama-index python-dotenv

2. Create a .env file with your OpenAI API key:

   OPENAI_API_KEY=sk-proj-xxxx

3. Import necessary components and load environment variables:

   from dotenv import load_dotenv
   
   load_dotenv()
   from llama_index.core.bridge.langchain import BaseLLM
   from llama_index.core.langchain_helpers.agents.toolkits import LlamaToolkit
   from llama_index.core.langchain_helpers.agents.agents import create_llama_chat_agent
   from llama_index.llms.openai import OpenAI

4. Initialize the LLM:

   llm = OpenAI(model="gpt-3.5-turbo", temperature=0)

5. Initialize the agent without predefined tools:

   toolkit = LlamaToolkit(tools=[])
   agent = create_llama_chat_agent(toolkit, llm, agent_kwargs={"verbose": True})

6. Ask a question and let the LLM handle it dynamically:

   response = agent.chat("What is the capital of France?")
   print(response)

This setup creates an agent that relies entirely on the LLM to process queries dynamically without predefined tools ^[1][2][3][4].

To continue talking to Dosu, mention @dosu.

[mraguth]

@dosu But where I give the agent description as a 'Senior Data Scientist'?

[dosubot[bot]]

To include a detailed agent description such as "Senior Data Scientist" in your LlamaIndex agent, you can use the ReActAgent implementation. Here’s how you can do it:

1. Set up your environment and install dependencies:

   poetry init
   poetry shell
   pip install llama-index python-dotenv

2. Create a .env file with your OpenAI API key:

   OPENAI_API_KEY=sk-proj-xxxx

3. Import necessary components and load environment variables:

   from dotenv import load_dotenv
   
   load_dotenv()
   from llama_index.core.tools import FunctionTool
   from llama_index.llms.openai import OpenAI
   from llama_index.core.agent import ReActAgent
   from llama_index.core.tools import ToolMetadata

4. Define a sample tool (optional):

   def query_tool(query: str) -> str:
       """Simulate a query tool that processes a query and returns a response"""
       return f"Processed query: {query}"
   
   query_tool_instance = FunctionTool.from_defaults(fn=query_tool)

5. Initialize the LLM:

   llm = OpenAI(model="gpt-3.5-turbo", temperature=0)

6. Define the agent description:

   agent_description = "Senior Data Scientist"

7. Initialize the agent with the description:

   agent = ReActAgent.from_tools(
       [query_tool_instance],  # You can leave this empty if you don't have predefined tools
       llm=llm,
       verbose=True,
       metadata=ToolMetadata(name="senior_data_scientist", description=agent_description)
   )

8. Ask a question and let the LLM handle it dynamically:

   response = agent.chat("What trends can you identify from the latest query response data?")
   print(response)

This setup includes the agent description "Senior Data Scientist" and allows the LLM to handle query processing dynamically ^[1][2][3].

To continue talking to Dosu, mention @dosu.