[Python] Trying to get my agent_executor for a sql agent in langchain to accept Prompts and...

The langchain, agent_executor, SQLDatabaseToolkit all working but I want to prompt it and keep chat history for follow up questions. For example place a prompt somewhere to push to use table x when asked about storage.

It looks like this at the moment :

toolkit = SQLDatabaseToolkit(db=db, llm=llm)
memory = ConversationBufferMemory(memory_key="chat_history")

agent_executor = create_sql_agent(
llm=llm,
toolkit=toolkit,
verbose=True,
memory=memory
)

agent_executor.run("How many products bought in Chicago")


Can i do that with this agent, toolkit or I need another method.?

At the moment running the above, I can't prompt it and it does not keep follow on conversation.

Like say you ask first question, how many products bought from Chicago, then follow up with a question - what were these.

Continue reading...