#### ##### # # # # # # # # # ### # # # # # # # #### ##### #
LastDB is in alpha. The product you install from Homebrew is a headless local daemon — apps talk to it over a Unix socket.
Data is never accessed as a free-for-all dump. Queries and mutations go through the node; first-party apps (Brain, Kanban, …) are the usual interface. Building a new app uses the same socket contract.
[Install LastDB →] — start there if you do not have a node yet.
1. INSTALL
Use the home-page one-liner (Last Stack installs LastDB + apps). Database only:
brew install edgevector/lastdb/lastdb brew services start lastdb
Full copy-paste path: Home → Install.
2. CONFIRM THE NODE
curl -s --unix-socket ~/.lastdb/data/folddb.sock http://localhost/health
# → {"status":"ok"}
lastdb status
lastdb --version
lastdbd --versionSocket: ~/.lastdb/data/folddb.sock. Health path is /health (not /api/health). TCP :9001 is retired.
3. USE AN APP (OR THE SOCKET)
brain init --grant-consent brain concept new local-search --title "Local search" --body "Embeddings stay on this machine." brain ask "what did I note about search?" # raw status from the daemon curl -s --unix-socket ~/.lastdb/data/folddb.sock http://localhost/api/status
Apps are the happy path. Building a new app: talk HTTP to the same socket (query / mutation / schemas).
INCLUDED
lastdbd — semantic daemon (store, index, app identity, sync when connected).
lastdb — tiny control CLI (status, connect, …).
Socket API for apps over ~/.lastdb/data/folddb.sock.
NOT INCLUDED
No desktop UI, no DMG, no built-in file-ingestion product surface. Older docs that mention lastdb ingest or /api/ingestion/* describe a retired full-node path — not the Homebrew install.
All examples use the owner Unix socket. Prefer first-party CLIs unless you are building a new app.
HEALTH
curl -s --unix-socket ~/.lastdb/data/folddb.sock \ http://localhost/health
Expect {"status":"ok"}.
STATUS
curl -s --unix-socket ~/.lastdb/data/folddb.sock \ http://localhost/api/status
Process / data-dir / sync summary.
SCHEMAS
curl -s --unix-socket ~/.lastdb/data/folddb.sock \ http://localhost/api/schemas
List schemas known to this node.
QUERY
curl -s --unix-socket ~/.lastdb/data/folddb.sock \
-X POST http://localhost/api/query \
-H "Content-Type: application/json" \
-d '{"schema_name":"concept","limit":5}'Structured query. Apps also use /api/mutation for writes (POST).
Exact request shapes evolve with the node; treat first-party apps and the TypeScript app SDK as the stable surface when in doubt. See Apps.
Each field can carry a value, a security label, a trust-distance policy (Wn Rm), and optional capability constraints. Queries are evaluated under an access context C = (user, τ, keys).
Query arrives with access context C = (user, τ, keys)
|
v
Check trust distance: τ ≤ m for each field's Wn Rm policy
|
v
Check capabilities: caller holds required key, quota > 0
|
v
Check payment: P(user, field) satisfied
|
v
Check security labels: ℓ_in ⊑ ℓ_out for all transforms
|
v
All pass? → Apply transforms → Return authorized projection
Any fail? → Return Nothing (no data, no error, no leakage)TRUST DISTANCE
// Each field has a Wn Rm policy
// W = max write distance, R = max read distance
{
"fields": {
"name": { "policy": "W0 R1" },
"diagnosis": { "policy": "W1 R1" },
"lab_results": { "policy": "W1 R3" }
}
}
// τ=0: owner · τ=1: doctor · τ=3: researcherTRANSFORMS
Deterministic functions on structures. Outputs are written back under policies of their own — so derived data is not a side door around access control. Details: the papers below.
LASTDB (Homebrew)
lastdb status # is the daemon up? lastdb --version lastdbd --version lastdb connect # optional: multi-device recovery phrase brew services start lastdb brew services restart lastdb
The control CLI is intentionally small. App workflows live in app CLIs.
APPS
brain init --grant-consent brain concept new <slug> --title "…" --body "…" brain ask "…" brain mcp kanban init kanban list kanban add <slug> --title "…" kanban mcp situations list situations init
Full catalog: Apps. Daily loop: How to use it.
THE PAPER
“Fold DB: Compute Without Exposure” — formal model and architecture.
fold_db_paper.pdf
ELI5 PAPER
Plain-language walkthrough of the same ideas.
fold_db_paper_eli5.pdf
HOMEBREW TAP
GITHUB
github.com/EdgeVector — public apps: brain, fkanban, situations, last-stack, …