Inside the engine.

This is how Altadore processes a message. Local classification, privacy scrubbing, structured memory. Type something and watch the real pipeline work.

Send a message through the pipeline
connecting to Pi...
message the gate classify internet think shape response local — deterministic cloud — probabilistic
pipeline — idle
pipeline —
stages —
total time —
george
How it thinks

It doesn't guess. It scores.

Every message hits a deterministic local gate before a single cloud API wakes up. The gate answers 22 binary questions to classify intent, score complexity, detect PII, and route the pipeline.

The split is fixed:

TIER 1 — 14 questions, local
Deterministic. Regex and keyword matching. Is this a greeting? Does it contain a date? Is it venting? Answered in <5ms, zero cost.
TIER 2 — 8 questions, cloud
Requires inference. What domain is this? How complex? What emotional tone? Resolved by the first cloud call alongside response planning.

Simple inputs — greetings, confirmations — get pattern-matched and never reach the cloud. Real questions get real compute. The system scales API usage to match cognitive complexity, not message length.

The gate score decides which pipeline fires:

SNAP — 0 calls
Pattern-matched responses for greetings, confirmations, one-word replies. Instant. Free.
QUICK — 1–2 calls
Fast cloud model. Quick answers, casual questions, low-stakes lookups.
DEEP_LITE — 2 calls
Skips the expensive reasoning call. The first call both classifies and drafts the response in one pass.
DEEP — 3 calls
Full pipeline: classify + plan (fast model), generate (reasoning model), enforce voice + format (fast model). External data pulled before the model thinks.
Privacy is structural, not promised. Every decision traces to a flag you can read.
Privacy architecture

Nothing leaves the building unless it has to.

Before any message reaches the cloud, a 3-layer PII scanner (word list, regex, NER) finds names, phone numbers, emails, addresses, and sensitive identifiers. Names become realistic pseudonyms — not bracket tokens. The cloud models see natural language they were trained on, not synthetic [PERSON_1] syntax. Real names never reach an API.

LOCAL (Pi / Desktop) ─────────────────── CLOUD (Cloud Models)

Real data stays here: ─── sanitized text ───▸ Cloud sees only:
Phil Henderson ──────────────────── Michael Chen
403-555-0192 ───────────────────── 403-555-0147
[email protected] ──────────────────── [EMAIL_1]

◄──────────────── RESTORE ────────────────
rehydrate pseudonyms back to real values

The token map lives in local process memory. Never serialized. Never sent to any API. The cloud generates a response using pseudonyms, then the restore pass swaps them back before the user sees it.

10-vector memory schema

Every fact is scored, not stored.

Each piece of information in Altadore carries ten numerical scores — weight, depth, domain, expiry, sensitivity, confidence, urgency, valence, feedback, scope. The system doesn't search a text file. It runs weighted scoring against a SQLite table and pulls exactly what matters.

Engine modules

What's inside

The engine is modular. Each piece does one thing. Green border means zero API cost — pure logic, math, and local ops. Accent border means cloud model calls.