# DrawBox Roadmap

This roadmap describes the direction of DrawBox over the next 12–18 months. It is intended to give users, contributors, and partners a clear view of where the library is headed, why, and how to get involved.

The roadmap is a living document. Dates are targets, not commitments — major shifts are reflected in GitHub Milestones and announced in release notes.

## Vision

DrawBox aims to be **the canonical drawing SDK for Compose Multiplatform** — a small, focused, accessible, and well-benchmarked canvas layer that any annotation, whiteboard, diagramming, or design tool can embed across Android, iOS, Web (Wasm), and Desktop.

We prioritize **SDK primitives over end-user products**: a strong foundation other tools depend on, rather than a feature-rich drawing application. The same primitive — a versioned, editable JSON scene format — is what makes DrawBox uniquely positioned to be the target format for LLM- and agent-driven drawing generation in the Kotlin ecosystem (see the natural-language drawing milestone below).

## Status

Latest published version: **2.1.0-alpha02** (KMP preview)

The 2.x line is a full rewrite on Kotlin Multiplatform with shared drawing logic and an MVI architecture. The `2.0` public surfaces (`DrawBoxController`, `Mode`, `Intent`, `Event`, `State`) are frozen and extended-only; `2.1.x` adds image, text, and pen-pressure surfaces incrementally.

## Near term — 2.0.0 stable (Q3 2026)

Goal: promote the frozen `2.0` surfaces to a stable `2.0.0` release and publish the accompanying migration path from 1.x.

- Public API freeze for `DrawBoxController`, `Mode`, `Intent`, `Event`, and `State` — **done**, extended only in `2.1.x`
- Migration guide from 1.x (Android-only) to 2.x (KMP) — **published in [CHANGELOG.md](CHANGELOG.md#migration-1x--2x)**
- Public WASM playground at `akshay2211.github.io/DrawBox/sample/` — **live**
- Snapshot-based regression tests across platforms — Roborazzi harness landed for JVM; extend to Android and iOS
- Dokka API reference published alongside the user docs

## Next — Accessibility & performance (Q4 2026)

Goal: make DrawBox the first KMP drawing SDK with serious accessibility and measurable cross-platform performance.

- **Accessibility pass**
  - Semantic descriptions for the canvas, tool controls, and exported drawings
  - Keyboard-only drawing and tool selection
  - Screen reader support on Android (TalkBack), iOS (VoiceOver), and Web
- **Performance benchmarks**
  - Frame time, gesture latency, and memory profile published per platform
  - Reproducible benchmark harness in `benchmarks/`
  - Public baselines so regressions are visible

## Mid term — Extensibility (Q1 2027)

Goal: let third parties extend DrawBox without forking it.

- Pluggable tool API — community-contributed `Mode`s as separate artifacts
- Pluggable exporters — beyond SVG/PNG/JSON (e.g. PDF, DXF)
- Stable extension contracts with semantic versioning
- A first example community plugin to validate the surface

## Mid-to-long term — Natural-language drawing (Q2–Q3 2027)

Goal: turn structured visual intent (natural language, LLM tool calls, agent output) into **editable** DrawBox scenes — not throwaway raster images.

This is a Kotlin ecosystem gap: Compose Multiplatform apps that want to embed LLM-driven diagram or sketch generation currently either (a) render server-side SVG and lose editability, or (b) hand-roll their own translation layer. DrawBox already has the primitive an AI layer needs: a versioned, human-readable JSON scene format that round-trips through the SDK.

- **`DrawBoxAi` optional module** — separate artifact so the core library stays offline and vendor-free.
- **`AiDrawingProvider` interface** — provider-neutral; host apps supply their own OpenAI / Anthropic / local-model implementation. Keys, billing, and rate limits never touch the core SDK.
- **Validated scene output** — models return `AiDrawingScene` (versioned JSON), which is schema-validated and clamped (element count, coordinates, text length) before insertion. Preview-before-insert is required; the whole generation is a single undo step.
- **Reference implementations** — sample OpenAI and Anthropic providers ship in `samples/`, not in the core artifact.
- **Safety defaults** — existing drawings are never sent to a provider without explicit user opt-in; no prompt / key / raw-response logging in the SDK.

See [STRATEGY.md](STRATEGY.md#recommended-ai-architecture) for the full design rationale.

## Long term — Collaboration & domain integrations (2027 H2)

Goal: unlock real-time and domain-specific use cases.

- CRDT-based collaborative drawing SDK primitive
- PDF annotation surface
- Reference integrations with adjacent Compose libraries (e.g. rich text, slides)

## Continuous

Work that runs in parallel across all phases:

- Triage and respond to issues within 7 days
- Track JetBrains Compose Multiplatform releases and bug-report blocker findings upstream
- Maintain `CONTRIBUTING.md`, `good first issue` triage, and quarterly release cadence
- Publish a release blog post for every minor version

## How to get involved

- **Try it**: the live WASM sample is at `https://akshay2211.github.io/DrawBox/sample/`
- **Report**: open issues on [GitHub](https://github.com/akshay2211/DrawBox/issues) — bug reports, feature requests, and accessibility findings are all welcome
- **Contribute**: see [CONTRIBUTING.md](CONTRIBUTING.md). Start with issues tagged `good first issue`
- **Propose**: larger changes go through a short RFC in `docs/rfcs/` before implementation
- **Sponsor**: see the repository sidebar for sponsorship links

## Out of scope

To keep the library small and focused, the following are explicitly *not* on the roadmap:

- A full vector editor UI — DrawBox stays an embeddable SDK, not an end-user product
- Server-side rendering of drawings
- Native rendering backends other than Compose

If your use case needs one of these, please open a discussion — we may be able to support it via the extensibility surface above.