mindinventory/mindsql

MindSQL: A Python Text-to-SQL RAG Library simplifying database interactions. Seamlessly integrates with PostgreSQL, MySQL, SQLite, Snowflake, and BigQuery. Powered by GPT-4 and Llama 2, it enables natural language queries. Supports ChromaDB and Faiss for context-aware responses.

mindsql's Issues

How does this differ from Vanna?

Hey there! This seems eerily similar to Vanna:

https://github.com/vanna-ai/vanna

I’m curious why our projects need to be different

Support for wikisql or open surce llms

Hi, good work. Are there plans to add support for any open source llms that does not require api keys?

minds.ask_db() : list index out of range

I am trying to run a local Llama 2 model against my SQL database on ubuntu. However, I'm encountering the following range error.

WARNING — 2024-03-07 03:27:38,715 — Minds Core — Function:ask_db — Line:498 — An unexpected error occurred: list index out of range
Res: {'error': 'list index out of range'}

Here is the code:

from mindsql.core.mindsql_core import MindSQLCore
from mindsql.llms import LlamaCpp
from mindsql.databases import MySql
from mindsql.vectorstores import Faiss

config = {
    "model_path": "/home/guitmonk/Downloads/llama-2-7b.Q3_K_L.gguf"
}

class MindSqlGenAI(Faiss, LlamaCpp, MySql):
    def __init__(self, config=config):
        Faiss.__init__(self, config=config)
        LlamaCpp.__init__(self, config=config)
        MySql.__init__(self, config=config)

minds = MindSqlGenAI(config=config)
connection = minds.create_connection(<SQL_URL>)

# Index DB Schmas
ddls = minds.get_all_ddls(connection=connection, database="test")

for ind in ddls.index:
    minds.index_ddl(ddls["DDL"][ind])

# Provide Example query-SQL pairs
minds.index(bulk=False, path="examples.json")

# ask questions
res = minds.ask_db(question="How many total clients are there?", connection=connection)
print("Res: ", res)

connection.close()

I've tried to debug it but I'm not sure where is this error originating from.

Adding custom configurations for Llama

The context window in Llama is set to 512 by default. This can cause the prompt size to exceed the context window size if the DB schema is large enough.

It would be great if custom configuration for the Llama function from LlamaCpp can be passed in the config object for better control.