Skip to content

Latest commit

 

History

History
34 lines (25 loc) · 2.42 KB

roadmap.md

File metadata and controls

34 lines (25 loc) · 2.42 KB

Roadmap

Available

  • ✅ One self-hosted Rust executable, SQLite, and the bundled dashboard.
  • ✅ Boolean, string, integer, float, named enum, and flat JSON record results.
  • ✅ Everyone, percentage, custom-rule, and nobody rollouts; local evaluation in all SDKs.
  • ✅ Python, Rust, Go, and Node.js SDKs with Partly providers for the official OpenFeature SDKs.
  • ✅ Standard typed getters, named-type metadata, and optional native exact-type assertions. See typed results.
  • ✅ Environment-scoped shared flags, app assignments, API Key rotation/revocation, and usage reports.
  • ✅ Individual accounts, RBAC, transactional audits, HTTPS, backup/restore, and operational logging.
  • ✅ OpenAPI contract generated from Rust and checked against real responses.

Finish the alpha release

  • ⬜ Pass the full build, vulnerability, packaged-consumer, and release workflows on the final commit.
  • ⬜ Publish verified binaries and SDK archives to GitHub Releases.
  • ⬜ Complete registry distribution and verify first use on clean machines. See publishing.

Evaluate next

  • ⬜ Git integration: start with reviewed export/import and clear ownership of changes. Automatic two-way synchronization needs a separate decision.
  • ⬜ Flag lifecycle: expiry reminders and lightweight ownership. Usage reports cannot prove a flag is unused; never delete automatically.
  • ⬜ Review newer OpenFeature baselines and upstream Rust gaps. Do not claim full conformance. Coverage and limits.
  • ⬜ Protect stored API Key secrets before production use; document recovery and rotation.

Outside the current scope

  • ❌ HA, Redis, required external databases, SSO, billing, or premium activation.
  • ❌ Browser/mobile SDKs or additional server languages before demand justifies them.
  • ❌ Arrays, optional fields, nested records, generated application classes, arbitrary rule code, or weighted experiments.
  • ❌ Reusable audiences, stored user profiles, per-evaluation analytics, and automatic flag removal.

The typed-result design and limits are recorded in the implementation guide. Boolean-only applications retain policy v1; typed applications use v2. New SDKs accept both. Upgrade before assigning typed flags. SQLite remains development schema v1: incompatible databases are rejected and preserved, never reset automatically.