### Implementation of a testing environment (1 week) * The Spider dataset as well as the WikiSQL dataset have a testing framework. That can be used to evaluate our predictions. * [Link to evaluation framework of Spider dataset](https://github.com/taoyds/test-suite-sql-eval). * The Spider and WikiSQL use SQLite DB and our solution uses MySQL. The DB needs to be converted to MySQL before it can be tested. [SQLite to MySQL](https://github.com/techouse/sqlite3-to-mysql) * The predictions need to be made sequentially, and stored in a JSON format. ### Improving the current model (1 week) * [Link to the issue regarding the above topic](https://github.com/Samagra-Development/Text2SQL/issues/40). * Currently ways to improve the results by producing good extra hard prompts are being worked upon. ### Evaluate Implementation of RAT (1-2 weeks) * [Link to repository](https://github.com/Microsoft/rat-sql) ### Optimization of tokens (1-2 weeks) * The current model uses OpenAI API to generate responses and we would be most likely move forward with that methodology * New methods need to be explored, [Link](https://github.com/Samagra-Development/Text2SQL/issues/28)
Implementation of a testing environment (1 week)
Improving the current model (1 week)
Evaluate Implementation of RAT (1-2 weeks)
Optimization of tokens (1-2 weeks)