I work on my initial version of this project, I opensource it and I post this in HN two weeks back and it bombed. But I still got good feedback from two people who saw my post in a different group. So I revamped the project with some features and separate website with docs and posting it now.

Here is how this framework works.

Problem: Build customer facing apps/chatbot where customers can ask questions in natural language and query the DB and answer it. Build internal slack bot where your customer success person or other team member can ask your slack bot a question and you need to query your db.

Solution: Text2SQL there are plenty of solutions. But maintaining the semantic layer is the hard part, the more you label your data and give your business definitions or glossary, the better accurate it is. You need to maintain it somewhere and then you need to create some some endpoints or functions to integrate them.

In DataNeuron, the cli guides on creating the context layer through a set of cli commands. Once the context is set, one could chat using cli to see how the LLM responds and correct the context layer. Once you feel it is ready, you can then start integrating within your python app using dataneuron sdk or through the API (which you can deploy to LAMBDA or to VPS). You can also chat with db per customer or tenant, and it would add the client/tenant filter to the resultant query -- this is still WIP, currentl recursive CTE is not addressed yet.

I have been working on this last two weeks, so still requires lot of testing and iteration. Would like some feedback to improve this project.