# iknowyou

Communication personalization as a signal-processing problem. iknowyou wraps tidalDB's
signal ledger, preference vectors, and windowed aggregation with an observation pipeline
(LM-as-classifier), a briefing engine (query-to-profile), and a generation interface
(brief-to-prompt) — so a system can learn how each person prefers to be communicated with
and adapt continuously, without a training loop or feature store.

> **Design vs shipped.** The vision/architecture docs describe the *target* tidalDB-native
> design; what runs today is a Next.js + vLLM implementation. See the roadmap and dev setup
> for current state.

## Docs

| Doc | What it covers |
|-----|----------------|
| [vision.md](vision.md) | The problem and the product thesis (target design) |
| [architecture.md](architecture.md) | Domain model, pipelines, components (target design) |
| [ROADMAP.md](ROADMAP.md) | Milestones and current build status |
| [devsetup.md](devsetup.md) | Running the engine + Next.js app locally |
| [engine/README.md](engine/README.md) | The personalization engine service |

Part of the tidalDB workspace's `applications/` example consumers — see the
[repository root](../../README.md) for the database itself.