how can i use 2 sources of documents as tools to a langchain agent

Question

i have a use case where i have a csv and a text file .

the csv holds the raw data and the text file explains the business process that the csv represent.

i want to inject both sources as tools for a wrapper agent, that will answer the client questions. this is how i defined the tools:

tools = [
Tool(
    name='process_info',
     func=process_chain.run,
     description=desc_process
),
Tool(
    name='csv-chain',
     func=csv_agent.run,
     description=description)
]

then, i define the wrapper agent:

llm_chain = LLMChain(llm=llm, prompt=prompt)

tool_names = [tool.name for tool in tools]
agent = LLMSingleActionAgent(
           llm_chain=llm_chain, 
           output_parser=output_parser,
           stop=["\nObservation:"], 
           allowed_tools=tool_names
        )

agent_executor = AgentExecutor.from_agent_and_tools(agent=agent, tools=tools, 
                                                    verbose=True)

agent_executor.run("which test had the biggest drop in value")

the output indicates that it couldn't reach the actual csv data.

 Action: csv-chain
 Action Input: 
 - File path to the CSV file containing the value data
 - Calculate the difference between each test's value and the overall value
 - Sort the tests by the difference in descending order

 > Entering new AgentExecutor chain...
 Thought: I need to read the CSV file into a pandas dataframe and then calculate the 
 difference between each test's value and the overall value. Finally, I need to sort 
 the tests by the difference in descending order.
  Action: python_repl_ast
  Action Input:
   ```
  import pandas as pd

  # read the CSV file into a pandas dataframe
   df = pd.read_csv('file_path_to_csv')

  # calculate the overall value
   overall_value = df['value'].mean()

  #calculate the difference between each test's value and the overall value
   df['value_difference'] = df['value'] - overall_value

  # sort the tests by the difference in descending order
   df_sorted = df.sort_values(by='value_difference', ascending=False)
 ```
 Observation: FileNotFoundError: [Errno 2] No such file or directory: 
             'file_path_to_csv'

but when i use the actual tool as :

csv_agent.run("what is the average value")

it answers correctly:

> Entering new AgentExecutor chain...
Thought: Use the `mean()` function to calculate the average value
Action: python_repl_ast
Action Input: df['value'].mean()
Observation: 0.58463
Thought:The average value is 0.58463
Final Answer: 0.58463
> Finished chain.
0.58463

how can i use 2 sources of documents as tools to a langchain agent

0 Answers0