Episodic Memory with Vector Store - Interactive Visualization

Episodic memory allows agents to recall relevant past experiences when facing a new task. Each past episode (a completed task or conversation) is embedded into a high-dimensional vector and stored in a vector database. When a new query arrives, it is embedded and compared to stored episodes using cosine similarity. The top-K most similar episodes are retrieved and injected into the agent's context window, providing relevant prior experience without retraining. This visualization shows a bank of 8 past episodes, a 2D embedding projection, similarity scores, and token budget impact.

• 8 past episodes stored in the memory bank - real engineering tasks with timestamps and tags
• Select from 5 preset queries and watch the retrieval scores update across all episodes
• Adjust top-K (1–5) and similarity threshold to filter what gets injected into context
• Timeline view shows when memories were created and which are retrieved (purple dots)
• Optional 2D embedding projection shows query vector and memory clusters visually
• Token budget tracker shows how retrieved memories fill the context window

Part of the EngineersOfAI Interactive 3D - free interactive visualizations covering every major concept in machine learning and AI engineering. Hover any element for a plain-English explanation. No code required.