A little practice.
A longer memory.

Get a feel for Lore before you install it. These hands-on examples use fictional memories to make the everyday Pi workflow tangible.

01 / A memory that stays

Pick up where you left off.

Save a preference with lore_save, leave the conversation, and retrieve it with lore_recall. The conversation changes; your saved memory doesn’t.

Use this in Pi

Interactive example · sample dataRuns in this page. Your Lore store is untouched.

1. Save a preference
2. Start a new session
3. Recall it

Pi Session 01

Ready to remember

What should Pi remember?

Edit this sample preference, then save it for the fictional demo/orchard project.

Nothing saved yet. Give this session something worth keeping.

Lore’s memory

0 notes

Your saved preference will appear here.

Local to this example

See the Pi tool call

No tool call yet.

02 / The right place

Some notes belong to a project.

Switch the active repository. In this example, project notes stay with their project, while a global preference can travel with you.

Read about scope

Interactive example · sample dataA simplified scope illustration, not the full ranking algorithm.

You’re working in

The example memory store

Why the selection changes

Repository scope is one part of retrieval. Lore also considers the query, memory type, limits, and configured cross-repository behavior. Transferable memories and explicit overrides aren’t modeled here.

03 / Words and meaning

A match without the same words.

Local embeddings can find relevant memories beyond keywords. Compare the example candidates, then see what remains if the local model becomes unavailable.

Set up local inference

Interactive example · prewritten resultsNo model runs here. These illustrate candidate sources.

For the query database, choose a setup Keyword only Local embeddings available Local model unavailable

Lexical candidates

The words match.

Additional semantic candidates

The meaning can match, too.

In Pi, explicit recall combines semantic matches with deterministic recall. Its recent-memory fallback and optional query expansion add behavior beyond this illustration.

Make it part of your day.

Install Lore in Pi and start building a memory of your own.

Start with Pi Using Copilot CLI?