Skip to content

Roadmap

This is the single forward-looking roadmap for the Re-Shell monorepo (@re-shell/cli + @re-shell/ui + @re-shell/contracts + the dashboard app). It records honest delivery status, not aspiration — every “done” item is shipped in the current CLI (0.29.2) and backed by tests.

StatusMeaning
MVP-doneShipped in the current monorepo.
DONE+testedImplemented and verified by unit/integration tests.
SCAFFOLD/SPECCode structure + types present, but live-environment requirements (network, running cluster, LLM backend, Rust toolchain) make it env-limited; unit tests pass with controlled doubles.
post-MVP-plannedA real, scoped intention carried forward.
DROPPEDExplicitly removed from scope (speculative / mission-divergent).
  • Configuration — global config (~/.re-shell/config.yaml) and project config (.re-shell/config.yaml) with inheritance, cascading, templating, diff/merge, backup/restore, hot-reload, and schema validation.
  • Workspaces — declarative re-shell.workspaces.yaml, dependency-graph engine with cycle detection, topology health, state persistence, and change-impact analysis with content-hash file watching.
  • Plugins — registration/discovery, lifecycle, hooks API, dependency resolution, sandboxing, and the command-extension system.
  • Performance — sub-100ms startup (≈43ms achieved), lazy loading, tree shaking, and a startup cache.
  • Templates — a large, shipped backend/frontend registry (Node, Python, Rust, Java, .NET, PHP, Go, Ruby, and emerging languages), Docker/orchestration output, API-contract management, and database integration.
  • Health & analysisdoctor, analyze, and completion wired and reachable; command-group architecture (config / tools / workspace / templates / …).

Status as of Wave 9d. Each row states what was verified.

FeatureStatus
AI / NLP offline command interface (re-shell ai <prompt>)DONE+tested — offline intent parser resolves prompts to catalog-vetted argv with confidence scores; never auto-executes; spawns without a shell.
Cross-language service bridge (service bridge generate)DONE+tested — typed gRPC/REST/GraphQL client scaffolds, validated against installed tsc. Async transport (Kafka/Redis Streams, circuit breakers, tracing) is SCAFFOLD.
workspace.yaml v2 + JSON Schema + IDE autocompleteDONE+tested — canonical schema, AJV validation, VSCode/IntelliJ/Vim/Emacs config generation. Live $schema TLS resolution is SCAFFOLD.
K8s / Helm / GitOps generation (k8s generate, k8s helm generate, k8s gitops generate)DONE+tested — Deployment/Service/HPA/NetworkPolicy, Helm charts, and ArgoCD/Flux manifests; --out + --dry-run. Live cluster apply not run in CI.
Nx / Turbo monorepo importer (workspace migrate)DONE+tested — reads Nx and Turborepo configs, emits re-shell.workspaces.yaml v2.
Plugin marketplace / registryDONE+tested (CI-mocked) — real installer (npm/git/local) + registry client against the npm registry; live-network install is best-effort.
Policy packs + dependency drift (workspace policy check, workspace drift)DONE+tested — declarative zod-validated policy packs (recommended, baseline) + cross-workspace version-mismatch detection.
Template compatibility matrix + dry-run visual diff (templates matrix, create --dry-run)DONE+tested — full compat grid from the registry; throwaway-dir scaffold preview with per-file diff.
VS Code extension bridge (@re-shell/vscode)DONE+tested (compiled + unit) — pure core layer tested without the host; VS Code host launch is SCAFFOLD.
Hosted control plane (@re-shell/control-plane)SCAFFOLD/SPEC — typed tenant model, token auth, RBAC, allow-listed proxy as pure in-memory logic; no running server/DB/deploy.
Desktop / Tauri packagingSCAFFOLD/SPEC — src-tauri/ config + scaffold exist and the dashboard compiles; no signed binary produced.
  • A single shadcn-React component system in @re-shell/ui (the Web Components layer was retired).
  • @re-shell/contracts as the authoritative shared contract for CLI and UI.
  • The dashboard app + token-authed hub server: SSE /events, WS /jobs, 127.0.0.1 bind.
  • The re-shell ui launcher, design-system tokens, and theming.

Planned: WCAG 2.1 AA across all components, component performance budgets, component dev tooling (playground/codegen), and enterprise theming.

  • Expand emerging-language coverage (Deno, Bun, Kotlin, Scala, Crystal, Zig, Elixir, Nim) to parity.
  • Interactive terminal graph explorer scaling to thousands of nodes.
  • Optional, provider-abstracted LLM-assisted command translation / architecture analysis (no always-on telemetry, no mandatory cloud).
  • Unified package-manager abstraction and cross-language debugging/refactoring.
  • Multi-cloud deployment targets and compliance reporting.

Recorded so the decision is explicit and not silently re-introduced:

  • Quantum computing integration — DROPPED.
  • Blockchain / Web3 (smart-contract templates, dApp workflows) — DROPPED.
  • VR/AR / immersive development environments — DROPPED.
  • Voice/neural “natural-language-as-primary-interface” as a core pillar — DROPPED. (A narrow, optional, provider-abstracted AI assist remains — see ai.)

These are speculative and mission-divergent; they would dilute the platform’s actual value: a polyglot workspace + microfrontend toolkit with a typed CLI ↔ UI contract.