**How to Store Memory in Sequential Autogen Agent Workflow to Maintain Context in a Single Chat Session?**

[Priya-Issar]

Hi everyone,
I’m working on building a Text-to-SQL LLM chatbot using Autogen that takes user questions from the frontend and generates SQL queries as output.

Architecture(Sequential Autogen Flow):
Here’s the detailed workflow I’m using:-

1. User Proxy Agent: Takes the user question from the frontend.
2. Assistant: Understands the use case and prepares a task plan to solve the query.
3. Assistant: Retrieves the DDL from the vector database based on the user query (text question from the User Proxy Agent).
4. Assistant: Uses the retrieved DDL to generate the SQL query.

Goal:
I want to store memory to maintain the context throughout a single chat session, ensuring continuity and relevance in the responses.
I’ve already explored concepts like Teachability, Mem0, and MemGPT, but I’m unclear on how to implement these effectively in the above use case.

Questions:

1. What’s the best way to integrate memory storage into this workflow to maintain context?
2. Are there specific strategies for efficiently storing and retrieving chat session data for such a use case?
3. If anyone has sample code or repositories, could you kindly share them for reference?

Looking forward to your suggestions!