Metadata-Version: 2.1
Name: asql
Version: 0.1.2
Summary: Query a database in natural language
Home-page: https://github.com/paulfitz/asql
Author: Paul Fitzpatrick
Author-email: paulfitz@alum.mit.edu
License: UNKNOWN
Description: # asql
        
        Query a database in natural language.  Uses docker images from 
        https://github.com/paulfitz/mlsql - make sure Docker RAM limits
        are not too restrictive (should be above 3GB).
         
        ## Installation
        
          * `pip install asql` to work with Sqlite or CSV files.
          * `pip install asql[postgres]` to work with PostgreSQL databases.
          * `pip install asql[mysql]` to work with MySQL databases.
          * For other databases, see [SQLAlchemy supported dialects](https://docs.sqlalchemy.org/en/13/dialects/mysql.html).
        
        ## Use
        
        Tell `asql` which model to use:
        
        ```
        # pick one of these
        asql --docker sqlova
        asql --docker valuenet
        ```
        
        I suggest you start with just one of these, although you can start both.
        The `sqlova` model works on single tables (e.g. a csv file) and can handle queries that take parameters.
        The `valuenet` model works on many tables, but is less strong at queries that take parameters.
        
        
        Tell `asql` which data to use:
        
        ```
        $ pick one of these
        asql --db your_data.csv
        asql --db your_data.sqlite
        asql --db postgres://user:password@host/database
        asql --db mysql://user:password@host/database
        ```
        
        Don't use on an enormous database just yet.
        
        Now, ask whatever questions you like in plain English, and see what happens:
        
        ```
        asql how many players are there?
        asql which is the longest bridge?
        ```
        
Platform: UNKNOWN
Description-Content-Type: text/markdown
Provides-Extra: postgres
Provides-Extra: mysql
