MemBot AI uses JSON files for persistent memory

MemBot AI is a Streamlit-based chatbot that uses a Groq-hosted language model and a JSON-backed memory engine to persist customer data across sessions [DevTo]. The system has four components: a Streamlit front-end, a language model layer that calls Groq's LLM, a memory engine that indexes JSON blobs, and a GitHub repository for persistent storage [Groq Docs].

During a demo, the assistant remembered a prior refund delay and the user's preference for email notifications, then answered “Your refund is still processing; I’ll email you the update shortly.” The demo code, Dockerfile, and a sample dataset are available on GitHub.

MemBot AI's approach reduces friction by 30% in internal tests, as users no longer repeat the same issue in every ticket [DevTo]. The bot also personalizes responses by persisting preferences, such as communication channel and language. However, this approach raises compliance questions under GDPR and CCPA, as personal data is stored in plain JSON files. Teams that adopt this pattern must embed encryption, access controls, and audit logs to mitigate this risk [DevTo].