> ## Documentation Index
> Fetch the complete documentation index at: https://handbook.polar.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# ADR-0001: Record architecture decisions

> We keep short, immutable ADRs for significant, cross-cutting decisions.

<Info>
  **Status**: Accepted

  **Area**: Cross-cutting

  **Date**: 2026-07-02
</Info>

## Context

We write [design documents](/engineering/design-documents/index),
and those are useful to show how a feature will work.
We have some rules in `AGENTS.md` files and lint scripts.
But most of the conventions in the code are not written.
New engineers and agents don't have all the context,
and the rules may be missed.

Design docs don't fill this gap: they're large, forward-looking, and go stale once a feature
ships. We need a home for short, durable *decision* records.

## Decision

We record conventions or hard-to-reverse decisions as **ADRs** under
`handbook/engineering/decisions/`, one sequentially numbered file each, from the
[template](/engineering/decisions/template). One numbered sequence covers the whole
monorepo, and the **Area** field scopes each ADR (Backend, Frontend, Infra, or Cross-cutting).

## Consequences

* One place to learn *why* the codebase is shaped the way it is, reviewed and published like
  design docs.
* Agents and reviewers treat ADRs as binding, cite them when flagging violations,
  and propose new ones for uncovered decisions.
* Design docs are unchanged; an ADR may summarize and link to one.

## Alternatives considered

* **Keep everything in `AGENTS.md` or lint only**: captures the *rule* but not the *why* or
  the alternatives, and mixes binding decisions with tips and setup.
* **Only write design docs**: those should be used for feature design, being heavy and providing details.
* **Per-package ADR logs**: fragment cross-tier decisions and duplicate numbering.

## References

* [ADR index and lifecycle](/engineering/decisions/index)
* [Design documents](/engineering/design-documents/index)
