Skip to content

Memory

Jump to bottom
FischermanCH edited this page May 12, 2026 · 3 revisions

Memory

Updated: 2026-05-12

Purpose

Memory is ARIA's semantic knowledge store. It is separate from notes, logs, and raw runtime results.

ARIA uses Memory for:

  • stable facts about the user and environment
  • preferences
  • session context
  • longer-term rollups
  • document RAG
  • Experience Memory for safe learned action patterns

What deliberately does not go into memory automatically

  • every transient question
  • complete SSH/SMB/RSS snapshots
  • technical logs without lasting value
  • mutating action proposals without review

This reduces memory noise and prevents ARIA from turning random one-off events into durable assumptions.

Store types

Facts and preferences

Long-term knowledge ARIA may reuse later.

Session context

Working memory for active tasks and previous turns.

Rollups

Compressed weekly/monthly or work-context summaries. Rollups help without pulling every old chat detail into every prompt.

Document collections

RAG v1 for uploads under /memories. Supported formats are text, Markdown, and PDFs with embedded text. OCR/scan PDFs are not part of v1.

Experience Memory

Successful safe recipe/guardrail/action patterns can be stored as planner context. They help ARIA propose actions, but do not replace policy or guardrails.

Recall

Recall can combine:

  1. direct facts/preferences
  2. session context
  3. rollups
  4. document guides and matching chunks
  5. Experience Memory for action planning

Chat details show sources, collection, and chunk references when document recall was used.

UI

  • /memories for entries, search, editing, deletion, and JSON export
  • /memories/map for collections, document groups, rollups, and structure
  • /config/embeddings for embedding model and safety confirmation when memory already exists

Embedding fingerprint

Memory and document entries carry an embedding fingerprint. This prevents ARIA from silently mixing old and new embedding generations during recall or document routing.

Forgetting

Entries can be deleted directly in /memories. In chat, ARIA can recognize explicit forget requests, but destructive operations should ask for confirmation.

Why answers can feel thin

  • Memory is disabled or Qdrant is unreachable
  • embedding model changed and old entries no longer match
  • too few or wrong memories exist
  • the request is more of an action prompt and is routed to the Agentic Action Flow before RAG
  • Top-K or recall limits are too conservative

Test hints

  • use remember ... for explicit storage
  • ask about the same fact later
  • check /stats and chat details for recall sources
  • check /memories/map for collection/rollup structure

Clone this wiki locally