AI agentsIntermediate

Persistent Semantic Memory for AI Assistants

Semantic memory lets an assistant retrieve relevant past context instead of treating every conversation as isolated.

AI agentsMemoryEmbeddingsSemantic search

Site connection

The personal assistant project uses embeddings and ChromaDB-style retrieval for preferences, files, and past conversations.

Visual model

Memory retrieval as ranked context

A memory system retrieves the past facts most relevant to the current request.

Interactive

Hybrid retrieval turns a vague study question into ranked evidence

Lecture: embeddingskeyword 0.34 / vector 0.94
0.87
Canvas calendar exportkeyword 0.76 / vector 0.42
0.79
Syllabus policieskeyword 0.38 / vector 0.62
0.72
Study guide draftkeyword 0.28 / vector 0.68
0.61

Memory as Retrieval

The assistant should not dump all past history into every prompt. It should retrieve a small set of relevant memory records.

That makes memory closer to search than to a giant transcript.

What Gets Stored

Useful records include durable preferences, facts the user explicitly wants remembered, summaries of projects, and indexed documents. Temporary chatter should usually expire or remain unpromoted.

Common Pitfalls

  • Saving everything forever.
  • Retrieving memories without timestamps or source context.
  • Letting stale preferences override newer user intent.

Quick check

Quiz

Why rank memories before using them?
  1. To keep only relevant context in the prompt
  2. To delete all old context
  3. To avoid user consent
  4. To make responses random

Ranked retrieval keeps the active context focused.

Sources and Further Reading

Related Explainers