TextualOrm

This tool generates SQL queries from natural language and retrieves query results. This orm generates sql queries from your input and the specifics of your connected database and can also run them to retrieve records.

Find the code here

Currently, this ORM only supports retrieval queries. Queries that perform delete, create, or other CRUD (Create, Read, Update, Delete) actions will not be executed on your database..

The core of this application is the use of Large Language Models to generate the sql queries. This orm currently supports two LLMs:

1. SQL Generator LLM - (free and the default llm used).
2. OpenAI (requires subscription to OpenAI)

You can find more information about the default SQL Generator LLM by going to the link here. This has been fine-trained from the flan-t5-base model.

Requirements

• Python
• Postgres
• Redis Note: Support for additional databases such as MySQL will be added in future updates.

Installation

pip install textual_orm

Usage Instructions

1. Initialize the Orm

   from textual_orm import TextualOrm
   textual_orm = TextualOrm(connection_string="postgresql://user:password@host:port/db_name",
               llm_type=LLMType.DEFAULT, redis_host="localhost",
               redis_port=6379)
   

   To use the OpenAI implementation, add your api-key to the arguments:

   from textual_orm import TextualOrm
   textual_orm = TextualOrm(connection_string="postgresql://user:password@host:port/db_name",
               llm_type=LLMType.DEFAULT, redis_host="localhost",
               redis_port=6379, api-key="")
   
   

2. Call the setup method: await textual_orm.setup()

3. Generate the SQL query

   sql_query = await textual_orm.make_sql_request("List of settings", ["setting"])
   print(sql_query)
   

   This method takes three arguments:

   • question: Input question
   • tables: List of tables as reference
   • request_data: A boolean to indicate if it should query the database or return only the sql_query (default value is False)
   
   By default, the make_sql_request does not actually query the database. It returns back the generated sql query which you can look at and verify. To get an alternative sql query, please modify your input question.

   To query the database with the generated sql query. Call the method passing in request_data=True. This will return a response in this format:

   {
     'query': 'SELECT * FROM setting ORDER BY created_at DESC LIMIT 5;',
     'data': [<records>]
   }
   

   Where data is a list of records from the query response. data will be None if request_data=False as in the default case.

Note that first time run may take a little time.

For better performance, speed and caching, redis is required.

Additional Arguments

This orm uses a default postgres max pool of 10. You can modify it if needed by passing your value to the max_pool argument.

Below is a list of other supported arguments to Orm:

• min_size=1 minimum size of pool
• max_size=10 maxiumum size of pool
• api_key="" api key for the given llm