Our Memory
Our Memory transforms interactions with Relational Agents by mapping key details from conversations into a dynamic Knowledge Graph. This system uses Relational Intelligence, our underlying system for handling context in interactions. It identifies interests, goals, and facts, then connects them to form patterns, making responses more relevant over time. With each message, the visual graph updates incrementally, accumulating context to support agent performance and adaptability.
How It Works
Our Memory builds its understanding layer by layer as the chat grows. Here’s a straightforward look at how it all comes together, powered by Relational Intelligence to create real, lasting connections:
Short-Term Memory
Every time the chat with the agent starts, Our Memory scans it for key details using smart language analysis. The AI breaks the input into categories and only keeps reliable information, enabling immediate, context-aware responses that lay the foundation for trust-building interactions.
Session Memory
At the end of a chat session (e.g., when the app closes), Our Memory reviews the whole conversation. It filters out low-quality notes, removes duplicates, and creates a quick summary like:
"This session covered travel plans and work stress; user seemed excited about Italy."
This keeps things fresh and relevant for the next session, enhancing continuity and loyalty through seamless recall.
User Memory
Across all sessions, Our Memory builds an overall profile, a living snapshot of interests, goals, and patterns that evolves with every chat. It combines summaries into a master view, like an evolving biography, powered by Relational Intelligence to deepen connections over time.
The Knowledge Graph: Connecting the Dots
Behind the magic is a visual map called the Knowledge Graph. It’s like a mind map where details are colorful nodes (dots) linked by relationships (arrows). This graph, at the heart of our Relational Intelligence, helps the AI "think" holistically, uncovering insights that build authentic connections at scale.
-
Yellow: Goals (e.g., "Plan a trip" or "Achieve work balance").
-
Green: Topics (e.g., "Travel" or "Finance").
-
Blue: Extractors (deeper patterns, like "User engages thoughtfully" or "Prefers concise replies").
-
Pink: User Facts (straight facts, like "Lives in Italy" or "Believes in slow travel").
The graph grows with connections, like "User’s interest in film → Relates to brand partnerships → Connects to career goals," enabling our Relational Intelligence to engineer personalized pathways for trust and engagement.
How Memory Influences Conversations
Memory isn’t just storage, it’s what turns conversations into meaningful connections, building genuine relationships by linking the dots in the Knowledge Graph.
-
Personalization: Remembers the style. If once it was mentioned liking long explanations, future responses get longer, tailored every time.
-
Continuity: Picks up where it was left off. "Last time we talked Italy, have you booked that slow-travel spot yet?"
-
Smarter Suggestions: Spots patterns through graph connections, offering tailored ideas that bridges goals, interests, and current topics for more relevant advice.
-
Deeper Engagement: Uses linked insights to ask better questions, deriving from related beliefs and interests to spark meaningful dialogue.
-
Relationship Building: The graph’s connections create holistic understanding, building trust and loyalty, like evolving from casual chat to a true relational partner.
-
Privacy-First: Data is collected and stored securely within the ProemthistAI system. The feature can be enabled or disabled at the agent level using the anonymous toggle, without losing the core interaction quality.