Pandasai was working with watsonx till a few days back... again started getting the error "pandasai.exceptions.NoResultFoundError: No result returned"

[raviisrani]

System Info

Python : 3.12.4
pandasai : 2.2.14
ibm_watsonx_ai : 0.2.6

🐛 Describe the bug

from pandasai import SmartDataframe
import pandas as pd
from pandasai.llm import IBMwatsonx

textwx.csv uploaded to githib for testing

#textwx.csv

#Create the prompt variable
First_prompt = "What is the average margin for 22?"
print ("\nPrompt first: ", First_prompt)

initialize watsonx llm

parameters1 = {
"decoding_method": "greedy",
"max_new_tokens": 1000,
"min_new_tokens": 1,
}

#model="ibm/granite-13b-chat-v2"
#model="mistralai/mixtral-8x7b-instruct-v01"
#model="codellama/codellama-34b-instruct-hf"
#model="google/flan-t5-xl"

llm = IBMwatsonx(
api_key = '0KX97ceUcTdzkUNGOxH_-RQgAydygoveSehxrLTHuqvd',
watsonx_url = "https://jp-tok.ml.cloud.ibm.com",
watsonx_project_id = '831054fa-285f-4d3d-a53f-ada70e8ac329',
model="meta-llama/llama-3-1-8b-instruct"
)

Supported wx models

#'codellama/codellama-34b-instruct-hf', 'elyza/elyza-japanese-llama-2-7b-instruct',

'google/flan-t5-xl', 'google/flan-t5-xxl', 'google/flan-ul2', 'ibm/granite-13b-chat-v2',

'ibm/granite-13b-instruct-v2', 'ibm/granite-20b-multilingual', 'ibm/granite-7b-lab',

'ibm/granite-8b-japanese', 'meta-llama/llama-3-1-70b-instruct', 'meta-llama/llama-3-1-8b-instruct',

'meta-llama/llama-3-70b-instruct', 'meta-llama/llama-3-8b-instruct',

'mistralai/mixtral-8x7b-instruct-v01', 'mncai/llama2-13b-dpo-v7'

Create a Smart Data Frame

df1 = pd.read_csv("textwx.csv")
df = SmartDataframe(df1, config={"llm": llm})
print("\nCreated smart data frame with PandasAI")
print("\nAverage margin : ",df.chat(First_prompt))

[dosubot]

Hey there, @raviisrani! I'm here to help you with any bugs or questions you have while waiting for a human maintainer. Let's squash those bugs together!

The pandasai.exceptions.NoResultFoundError is raised when no result is found in the response, which typically occurs when an operation or query is expected to return a result, but it does not ^[1].

In your case, this error might be related to the handling of LLM responses by the SmartDataframe class. The ResultValidation class is responsible for validating the output of the LLM, and a 'No result returned' error could occur if the LLM response is None ^[2].

Here are a few steps you can take to troubleshoot and potentially resolve the issue:

1. Check Model Compatibility: Ensure that the model you are using, meta-llama/llama-3-1-8b-instruct, is compatible with the current version of PandasAI and IBM Watsonx. There might be recent changes or updates that affect compatibility.

2. Verify API Credentials: Double-check your API key, Watsonx URL, and project ID to ensure they are correct and have not expired or been revoked.

3. Inspect the Prompt: Make sure that the prompt you are using is correctly formatted and that the data in textwx.csv is as expected. Sometimes, the issue might be with the input data or the way the prompt is structured.

4. Review Recent Changes: If the setup was working previously, consider any recent changes in your environment or updates to the libraries that might have introduced this issue.

5. Test with Different Models: Try using a different model from the list of supported models to see if the issue persists. This can help determine if the problem is specific to the model you are currently using.

If these steps do not resolve the issue, you might want to look into any recent updates or discussions related to similar issues in the PandasAI repository.

To continue talking to Dosu, mention @dosu.

[raviisrani]

My responses below..

Check Model Compatibility: Ensure that the model you are using, meta-llama/llama-3-1-8b-instruct, is compatible with the current version of PandasAI and IBM Watsonx. There might be recent changes or updates that affect compatibility.

Ravi - Tried testing with other models on watsxonx platform.. same results

Verify API Credentials: Double-check your API key, Watsonx URL, and project ID to ensure they are correct and have not expired or been revoked.

Ravi - The same credentials and API key is working on other projects

Inspect the Prompt: Make sure that the prompt you are using is correctly formatted and that the data in textwx.csv is as expected. Sometimes, the issue might be with the input data or the way the prompt is structured.

Ravi - Prompt works perfect with Gemini flash.. in fact same was working till a few days back

Review Recent Changes: If the setup was working previously, consider any recent changes in your environment or updates to the libraries that might have introduced this issue.

Ravi - No changes in my local conda environment.. have not updated the python libraries

Test with Different Models: Try using a different model from the list of supported models to see if the issue persists. This can help determine if the problem is specific to the model you are currently using.

Ravi - Tried the same code, data set and prompt on gemini flash code .. works perfect!

Is there a way to debug pandasai classes, like resultvalidation classes to see what is the cause of this bug?

Can you assign someone to fix it.. have been having this issue with pandasai library with mainly watsonx.. works great with gemini