Skip to content

ROADMAP.md

Latest commit

 

History

History
67 lines (53 loc) · 3.25 KB

ROADMAP.md

File metadata and controls

67 lines (53 loc) · 3.25 KB
Raw

Roadmap

A 12-month outlook for aphrody, broken down by quarter. These are targets, not promises — release dates slip, scope changes, and an unexpected CVE can reshuffle a quarter. We publish the plan anyway because the alternative is worse. For background on what aphrody is and why it exists, start at SOURCE_OF_TRUTH.md and COMPARISON.md.

Q2 2026 (current — May to June)

Stabilise v1.0.0-canary and execute the crates.io publish ladder. Publish order: base first (no_std primitives, smallest blast radius), then a2a-pb (the protobuf generated code, frozen API surface), then aphrody itself (the user-facing CLI). Show HN launch lands in D+15; the public read-out targets one thousand GitHub stars and the start of an organic contributor pipeline. Internal milestone: the YOLO grind loop runs unattended across overnight cycles without manual intervention (the mode is implemented as the /aphrody-yolo-grind skill under .claude/skills/).

Q3 2026 (July to September)

Ship v1.0.0 stable. Cut a release line, freeze the public API surface, and commit to backporting security fixes for at least two minor versions. Distribution surface widens: a Snap package for Ubuntu, an AUR entry for Arch, and a Homebrew tap for macOS users who want a best-effort build. The devcontainer image goes GA. Star target: ten thousand. Contributor target: fifty distinct authors with merged PRs. The stable line also unlocks the formal third-party audit window we have been deferring.

Q4 2026 (October to December)

Adopt the AGNTCY a2a/v0.5 spec once it lands, with backward compatibility for v0.4 peers during the migration window. WebGPU support arrives in aphrody-wasm via the wgpu crate, enabling browser-side render workloads driven from the CLI. The MRX (Monorepo Real-time X-platform mapper) daemon goes GA — aphrody mrx watch will run as a long-lived process emitting filesystem-change deltas over the A2A channel. Star target: fifty thousand. This is also the quarter we expect to be invited (or rejected) into the AGNTCY consortium.

Q1 2027 (January to March)

AGNTCY consortium membership formalised. Multi-language agent support lands behind a stable FFI: Python, Go, and TypeScript bindings expose the A2A protocol without requiring a Rust toolchain. The official MCP (Model Context Protocol) bridge ships, replacing the current third-party adapter we vendor. Star target: one hundred thousand. Quarter-end check: enough external usage and feedback to plan a v2.0 cycle with confidence.

Cross-cutting commitments

These hold across every quarter and are non-negotiable:

  • Linux Ubuntu 26.04 remains build target #1. A regression on Linux blocks the release; Windows or WASM regressions do not.
  • Zero telemetry. Period. See FAQ.md for the rationale.
  • Supply-chain hygiene: cargo deny check and cargo vet must stay green on main at all times. See SECURITY.md for the vulnerability disclosure policy.
  • All public docs follow the cross-link style established here — answers point to the source-of-truth file rather than duplicating prose.

Pull requests proposing changes to this roadmap are welcome — open an issue first to discuss.