Problem:
How we can allow talking to our database data using a human language.

Solution:

• Use a Langchain package.
• Upload your data to a database. Connect to it using a known schema.
• Form a right prompt that would be sent to a language model (eg. OpenAI or LLama-2). It is typically needed on top of a human question.
• Some models may require pretraining and quantization. Then such a model can be served to client right from a client server or from a cloud.
• Figure 1 shows a human request and the LLM response which is then converted to a plot using Streamlit package for the human-to-model interface.