Skip to content

Your First Memory

This guide walks you through capturing, finding, and recalling your first atom. By the end you'll understand what Lattice actually does with what you tell it.

Step 1: Capture a fact

With the daemon running, open the web UI at localhost:7337 or use the terminal:

uv run lc "I want to read Thinking Fast and Slow by Kahneman"

You'll see a confirmation: captured: 1 new.

Step 2: Look at the atom on disk

ls ~/.lattice/

You'll find a .md file — something like a1b2c3d4.md. Open it:

cat ~/.lattice/a1b2c3d4.md

---
id: a1b2c3d4
kind: goal
subject: reading list
content: Wants to read "Thinking, Fast and Slow" by Daniel Kahneman.
observed_at: 2025-11-14T09:12:00Z
source_id: lc-cli
quality_score: 1.0
tier: stm
---
Wants to read "Thinking, Fast and Slow" by Daniel Kahneman.

Lattice didn't just save the raw text. It extracted:

  • kind — goal (a thing you want to do)
  • subject — reading list (what this atom is "about")
  • content — a clean restatement of the fact

Step 3: Capture a few more atoms

The recall system works best with a few atoms to link together.

uv run lc "I read about 20 pages a day on average"
uv run lc "I prefer non-fiction over fiction"
uv run lc "My current book is Meditations by Marcus Aurelius"

Step 4: Ask a question

Open the web UI and type:

What am I reading? What should I read next?

Lattice will:

  1. Search your atoms (BM25 + graph BFS)
  2. Find atoms about current book, reading list, reading pace, genre preference
  3. Synthesize a prose answer with numbered citations like [1], [2]

The answer should mention Meditations as your current book and suggest Thinking Fast and Slow as next, grounded in what you actually told it.

Step 5: Check the graph

The graph links all your reading-related atoms by subject:

uv run lattice-daemon status

Returns atom_count: 4 — four atoms, all linked to the reading list subject node in the graph.

What just happened

your text
    ↓  segmented (source type = lc-cli)
    ↓  LLM extraction → {kind, subject, content, observed_at}
    ↓  atom written to ~/.lattice/
    ↓  graph updated: atom_has_subject edge → subject:reading list
    ↓  BM25 index rebuilt

your question
    ↓  BM25 seeds: "reading list", "book", "non-fiction"
    ↓  graph BFS: subject:reading list → all 4 atoms
    ↓  LLM synthesis: prose answer with [1][2][3] citations
    ↓  web UI renders answer + source chips

Next steps