Seven parts built the individual layers. This final part assembles them into a complete, deployable production system with a full reference architecture, infrastructure configuration, monitoring setup, cost model, and a decision framework for when to use each memory type.
Tag: agent memory
AI Agents with Memory Part 5: Memory Consolidation – Summarising and Compressing Long-Term History with Node.js Background Workers
Episodic memory grows indefinitely. Without consolidation, retrieval degrades and storage costs climb. This part builds a Node.js background worker that compresses episodic memory into semantic knowledge on a rolling schedule, keeping your agent sharp without ballooning your database.
AI Agents with Memory Part 3: Semantic Memory – Building a Long-Term Knowledge Layer with Qdrant and Python
Episodic memory records what happened. Semantic memory stores what the agent has learned. This part builds a production semantic memory layer using Qdrant and Python, with fact extraction, importance-weighted upserts, and similarity retrieval that lets agents build genuine knowledge about users and domains over time.
AI Agents with Memory Part 2: Episodic Memory – Storing and Retrieving Conversation History at Scale with PostgreSQL, pgvector, and Node.js
Episodic memory is what lets an agent remember what happened in past sessions. This part builds a complete production episodic memory system using PostgreSQL with pgvector, implementing hybrid time-based and semantic retrieval in Node.js so your agent never starts from zero again.
AI Agents with Memory: Why Single-Session Agents Fail in Enterprise and the Three Memory Types That Fix It
Most agent guides cover single-session work. Enterprise agents need persistent memory across sessions. This first part explains why stateless agents break down at enterprise scale, introduces the three memory types that solve it, and maps out the architecture this series will build.
