Jupyter Notebook walkthrough of sql-eval with sqlite

[shravankshenoy]

Not an issue, but I thought it would be helpful to have a notebook which shows how to setup and use sql-eval, so I created an Jupyter notebook which shows how to use sql-eval with a sqlite database. Hope this notebook is helpful for others

Jupyter Notebook link : https://colab.research.google.com/drive/1pejJn1GDPo4wzfWbmq5sE0ZlNNWNNYfb?usp=sharing (update: link no longer valid, use the link below)

Let me know if it would be useful for adding this notebook to the repository. If yes, I can create a PR. If not, I will close this issue.

Edit: The above link is no longer valid. To view the jupyter notebook please visit the link below
https://github.com/defog-ai/sql-eval/blob/5ba97bcbe6b2b3ed9b826365cbbfa59e28446518/examples/sqlite_example.ipynb

[wongjingping]

Thanks for sharing this! Would you like to add this into a new folder /examples/sqlite_example.ipynb?

[shravankshenoy]

Sure. I will raise a PR for this.

[shravankshenoy]

Many people have raised access to the Jupyter notebook link I had shared in the original comment. Kindly note that the link is no longer valid, and to access the jupyter notebook please refer the link below:

https://github.com/defog-ai/sql-eval/blob/5ba97bcbe6b2b3ed9b826365cbbfa59e28446518/examples/sqlite_example.ipynb

[andrewclayman]

This example would be useful. One addition that would make it even more valuable for text-to-SQL reliability is showing separate checks for execution result, schema linking, and ambiguous intent handling. In production, a single pass/fail score can hide whether the model chose the wrong table, used the wrong join, misunderstood the time window, or produced SQL that happened to return a plausible result.