One platform.
Evidence before scale.
This specification defines the scope, interfaces, operating rules and verification gaps of the current UNITIQX Mission Control. It is a working engineering foundation, not a claim of functional parity with Apple, NASA, Palantir or any frontier AI vendor.
1. System boundaries and inputs
The existing public HTTPS gateway hosts static approved applications; administrative services are not exposed through the Mission Control page. The collector reads approved public HTTP endpoints, model-name inventory, aggregate service/container counts and the existing scheduled security-receipt file.
| Component | Contract | Boundary |
|---|---|---|
| Dashboard | GET /mission-control/ | Read-only browser |
| Snapshot | GET /mission-control/snapshot.json | Curated public JSON |
| Registry refresh | systemd 15-minute timer | Internal; no public trigger |
| Incremental preflight | Private allowlisted CLI | Internal; no public shell |
| Security regression | Existing scheduled gate | Unaffected private tests |
2. Traceability and proof
The snapshot carries twelve uniquely identified requirements with acceptance criteria, verification-state metadata and specific evidence classes. A receipt only supports its own tested behavior. Signed build provenance, third-party assessments, human accessibility, disaster recovery, and model accuracy remain unverified where tests are missing.
Engineering patterns: NASA software requirements, NIST SSDF, SLSA.
3. Future operational ontology
Define Project, Capability, Model, Task, Requirement, Evidence, Artifact and Release as typed objects. Connect them with USES, DEPENDS_ON, SATISFIES, VERIFIES, PRODUCED_BY and SUPERSEDES relations. Future privileged actions require authenticated owner grants, approval receipts and rollback. This follows an original design direction informed by documented object/link/action modeling, not a Palantir integration or clone.
4. Accelerator design
Only deterministic syntax checks may be reused from a hash-and-runtime-keyed cache. Live availability and security-gate freshness are always checked. The first paired preflight observation took 309 ms cold and 65 ms warm (approximately 4.8× for that narrow preflight only). Broader system and AI throughput improvements require separate baseline and repeated benchmark measurements.
5. Quality targets
| Area | Target evidence |
|---|---|
| Interface | Responsive layout, keyboard, real iOS/Android and accessibility testing |
| Reliability | p50/p95 response latency, error rates, security negatives, restore drill |
| AI | Task-specific held-out accuracy, latency, failure analysis, cost |
| Builds | Source hash, dependency lock, reproducible build, artifact hash, approval |
| Operations | Alerts, least-privilege workers, rollback, offsite backup verification |
Related external guidance: Apple HIG and OpenTelemetry.
6. Explicit gaps
No independent evidence yet establishes secure high-availability hosting, frontier-model compute parity, customer production SLOs, end-to-end signed attestations, full penetration testing, audited mobile accessibility, comprehensive offsite restoration or autonomous multi-node execution. These are backlog items, not completed features.
7. Operating and recovery policy
Public pages never accept arbitrary commands or access private model completion endpoints. The internal inventory timer is constrained. Server operators can stop periodic refresh without taking down games. Any rollback must compare existing files with the saved pre-change gateway and retain concurrent modifications.
The private versioned specification, source manifests and build receipts remain in the VPS release archive. This public document contains only approved high-level interfaces and design commitments.