Have you ever walked into a client meeting trying to remember what you promised them three weeks ago, their preferred communication style, or the exact details of the last technical debrief? Remembering everything across dozens of clients is tough.To solve this, I built the AI Meeting Prep & Contact Memory Agent—a full-stack Python application that securely retains client context in isolated memory banks, automatically logs history to spreadsheets, and generates instant pre-meeting executive briefings using high-speed LLM inference.
🛠️ The Tech Stack
Memory Persistence: Vectorize Hindsight Cloud (hindsight-client) for context retention and semantic recall.
inference / LLM: Groq API using the openai/gpt-oss-120b model for rapid executive summary synthesis.
User Interface: Streamlit for a clean, reactive dashboard.
Data Logging: Pandas & local CSV spreadsheets (client_meeting_tracker.csv, client_reminders.csv) for persistent offline logging.
🏗️ Core Architecture & Features
Multi-Bank Isolated Client Memory
To ensure client data is strictly segregated, the agent dynamically generates a sanitized, client-isolated Hindsight memory bank ID for every contact:-
Instant Executive Briefings via GroqWhen preparing for an upcoming call, the app queries Hindsight to recall past notes and passes them into Groq to synthesize a structured brief covering communication preferences, commitments made, and key discussion topics. 3. Interactive Streamlit Dashboard
- Interactive Streamlit Dashboard The app features a multi-tab interface built with Streamlit
AI Meeting Prep & Contact Memory Agent
An intelligent agent built with Vectorize Hindsight for persistent client memory, Groq API for executive briefings, and a Streamlit user interface.
Features
-
Multi-Bank Architecture: Isolated Hindsight memory banks per contact (
bank_client_*). - Automated Logging: Tracks debriefs and reminders in CSV spreadsheets.
- Executive Briefings: Generates instant summaries on client communication preferences and commitments.
Top comments (0)