Roadmap
Reqloom is pre-1.0. This page says what actually works, because “in progress” on its own isn’t useful when you’re deciding whether to adopt something.
The numbers come from an internal audit that checks every requirement against source with file-level evidence, rather than against intent.
Where it stands
Section titled “Where it stands”852 tests passing. Against the 92 functional requirements Reqloom is specified against:
| Status | Count |
|---|---|
| Done | 60 |
| Partial | 10 |
| Not done | 20 |
| Deferred (post-MVP by design) | 2 |
Weighted completion, counting partial as half: 71%.
By area
Section titled “By area”| Area | Completion | Notes |
|---|---|---|
| Response viewer | 100% | |
| Request editor | 100% | |
| Dependency graph | 100% | Shipped ahead of its phase |
| CLI | 92% | |
| Execution engine | 91% | The core is essentially complete |
| HTTP client | 89% | Only WebSocket/SSE missing, deliberately deferred |
The engine and CLI are the mature parts. That’s deliberate — they’re what the desktop app and CI both depend on.
What works today
Section titled “What works today”Everything documented on this site is implemented. The load-bearing pieces:
- Dependency resolution — implicit edges from references, explicit
depends_on, Kahn’s sort with a deterministic tie-break, cycle detection at load time - Eleven auth strategies plus per-operation inline auth, session caching, and token refresh
- Polling,
for_eachfan-out, response assertions, JS pre/post hooks, file uploads - Three output formats — text, JSON, JUnit — for CI
- Direct importers for seven formats, and the AI importer for unstructured input
- Secrets in the OS keychain, never on disk
- Desktop app — project explorer, request editor, response viewer, dependency graph
Known gaps
Section titled “Known gaps”Honest list of what you’ll notice missing:
| Gap | Impact |
|---|---|
| WebSocket / SSE | No streaming protocol support. Deferred by design |
No reqloom migrate | The error message for an unsupported version: suggests it; it doesn’t exist yet |
| No CLI secret management | Secrets go in via the desktop app or your OS keychain tool |
No --version flag | Check the release you downloaded |
| No strict schema mode | Unknown keys are silently ignored — see pitfalls |
| Unsigned builds | macOS and Windows warn on first launch. Needs a paid certificate |
| Desktop UI still maturing | Functional, less polished than the CLI |
Toward 1.0
Section titled “Toward 1.0”Roughly in priority order, not on a date:
- Signing and notarization so downloads stop looking broken
- Package managers — Homebrew and Scoop
- Closing the partial requirements — the 10 above
- Desktop polish and auto-update
- Strict schema mode, so misspelled keys become errors instead of silence
Post-1.0
Section titled “Post-1.0”Listed because people ask, not because they’re scheduled:
- Mock server generated from a schema
- Team workspace sync (a paid component — the engine, CLI, schema, and desktop stay Apache 2.0)
- Hosted runs
Stability commitments
Section titled “Stability commitments”Two things you can build on before 1.0:
- The schema is versioned.
version:is required and the loader accepts 1–3. A future breaking change gets a new version number, not a silent reinterpretation. - Error codes are stable.
E_*strings are a contract shared by the CLI, desktop, and JSON/JUnit output, and safe to assert on in CI.
Flags may gain options. Nothing documented here is planned for removal.
Following along
Section titled “Following along”- Releases — builds and changelogs
- Issues — bugs and requests
- Contributing — if you’d like to close one of the gaps