Unable to train using interface

[usmanbinimran-octopus]

System Info

Windows 10
Python 3.9
PandasAi 2.2.5

🐛 Describe the bug

Hi PandasAI Team,

I am unable to see training QA or docs at: https://www.pandabi.ai/auth/sign-in
Did you guys removed it from front end?

Also unable to train using code!

Kr,
Usman Bin Imran

[dosubot]

To assist you with training using the PandasAI interface, please ensure you follow these steps:

1. Set the PandasAI API Key:
   Make sure you have set your PandasAI API key as an environment variable. This is crucial for the training process.

   import os
   
   os.environ["PANDASAI_API_KEY"] = "YOUR_PANDASAI_API_KEY"

2. Instructions Training:
   Use the train method on the Agent to provide specific instructions.

   from pandasai import Agent
   
   agent = Agent("examples/data/Loan payments data.csv")
   agent.train(docs="Only return loans information about the past 10 years")
   
   response = agent.chat("How many loans were paid off?")
   print(response)

3. Q/A Training:
   Use the train method on the Agent to provide specific Q/A pairs.

   from pandasai import Agent
   
   agent = Agent("examples/data/Loan payments data.csv")
   
   # Train the model
   query = "How many loans were paid off?"
   code = """
   import pandas as pd
   
   df = dfs[0]
   df['loan_status'].value_counts()
   """
   agent.train(queries=[query], codes=[code])
   
   response = agent.chat(query)
   print(response)

4. Troubleshooting:
   If you encounter the error No vector store provided. Please provide a vector store to train the agent, ensure that you have set the API key correctly. You can also instantiate the vector store directly with the API key.

   from pandasai import Agent
   from pandasai.vectorstores import BambooVectorStore
   
   # Instantiate the vector store with the API key
   vector_store = BambooVectorStore(api_key="YOUR_PANDASAI_API_KEY")
   
   # Instantiate the agent with the custom vector store
   agent = Agent("examples/data/Loan payments data.csv", vectorstore=vector_store)

5. Clearing Previous Conversations:
   If you need to start a new training session, you can clear the previous conversation history.

   agent.start_new_conversation()

6. Explanation of Code Execution:
   If you need to understand how the code reached a solution, you can use the explain method.

   explanation = agent.explain()
   print(explanation)

By following these steps, you should be able to train the PandasAI interface successfully ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[usmanbinimran-octopus]

Dear Team and @gventuri,

Hope this message finds you well. While testing the Agent class, it came to our notice that Agent unexpectedly stopped using trained data. I am attaching the screenshot of console:

We are also unable to train Agent using interactive screen provided by PandasAI at: https://www.pandabi.ai/admin/

We are looking forward to your prompt response, as this is critical for our proof of concept development.

Kind regards,
Usman Bin Imran
+92-310-480-160-1