After training on my DB, I'm not getting answers related to it.

[AbdoDabbas]

First of all I'm very new to this AI thing so please bare with me and consider that I may missed installing something.

I'm using MSSQL to train the model, and chromadb as vector db and llama3.1 (4.7GB size) as a local ai model.

I followed exactly the sample in this page https://vanna.ai/docs/mssql-ollama-chromadb/ , but I faced two issues:

1. When the code reaches the vn.train method, it stops, I don't know if training is done or not, but I tries to log some info on cmd after calling the train but it doesn't reach it.
2. I thought the training is done, and all, now I commented the train line in the code and called the ask function, it shows some data and it's trying to call the LLM model to ask it the question, but the model response like it doesn't know what I'm asking for, even when I tried to run the Flask app, and ask the question about how many of X (table name) do we have ? expecting to response with SELECT Count(*) FROM X but I get something like "I'm a large language model and I don't have this info ...etc".

BTW: If I used mistral as a model instead of llama3.1, it starts to understand and make sense of the answers !

Here's my python code for a reference:

from vanna.vannadb.vannadb_vector import VannaDB_VectorStore
from vanna.chromadb.chromadb_vector import ChromaDB_VectorStore
from vanna.ollama import Ollama
import logging

logging.basicConfig(level=logging.INFO,
                    format='%(asctime)s %(levelname)s %(message)s',
                    datefmt='%Y%m%d %H%M%S')
                    
# The Ollama class already knows how to connect to a local ollama service
# The ChromaDB_VectorStore will store data. It will only need to be populated once and persists it.
class MyVanna(ChromaDB_VectorStore, Ollama):
    def __init__(self, config=None):
        ChromaDB_VectorStore.__init__(self, config=config)
        Ollama.__init__(self, config=config)

vn = MyVanna(config={'model': 'llama3.1'})
vn.connect_to_mssql('Driver=SQL Server;Server=ServerName;DATABASE=DBName;')

# The information schema query may need some tweaking depending on your database. This is a good starting point.
df_information_schema = vn.run_sql("SELECT * FROM INFORMATION_SCHEMA.COLUMNS")

# This will break up the information schema into bite-sized chunks that can be referenced by the LLM
plan = vn.get_training_plan_generic(df_information_schema)

vn.train(plan=plan)
 ## Code doesn't reach here, but if I commented the previous row, it continues and runs the flask app.
logging.info('Done training.')


# To run vanna Flask app 
logging.info('running vannaFlaskApp')
from vanna.flask import VannaFlaskApp
app = VannaFlaskApp(vn, allow_llm_to_see_data=True)
app.run()

[RodolfoMontenegro]

Have you tried adding the crhromadb path?