git.stella-ops.org/docs/11_DATA_SCHEMAS.md

Data Schemas & Persistence Contracts

Audience – backend developers, plug‑in authors, DB admins. Scope – describes Valkey, PostgreSQL, and on‑disk blob shapes that power Stella Ops.

---

## 0 Document Conventions

• CamelCase for JSON.
• All timestamps are RFC 3339 / ISO 8601 with Z (UTC).
• ⭑ = planned but not shipped yet (kept on Feature Matrix “To Do”).

---

## 1 SBOM Wrapper Envelope

Every SBOM blob (regardless of format) is stored on disk or in object storage with a sidecar JSON file that indexes it for the scanners.

#### 1.1 JSON Shape

{
  "id": "sha256:417f…",          // digest of the SBOM *file* itself
  "imageDigest": "sha256:e2b9…", // digest of the original container image
  "created": "2025-07-14T07:02:13Z",
  "format": "trivy-json-v2",     // NEW enum: trivy-json-v2 | spdx-json | cyclonedx-json
  "layers": [
    "sha256:d38b…",              // layer digests (ordered)
    "sha256:af45…"
  ],
  "partial": false,              // true => delta SBOM (only some layers)
  "provenanceId": "prov_0291"    // ⭑ link to SLSA attestation (Q1‑2026)
}

format NEW – added to support multiple SBOM formats.
partial NEW – true when generated via the delta SBOM flow (§1.3).

#### 1.2 File‑system Layout

blobs/
 ├─ 417f…                # digest prefix
 │   ├─ sbom.json        # payload (any format)
 │   └─ sbom.meta.json   # wrapper (shape above)

Note

– RustFS is the primary object store; S3/MinIO compatibility layer available for legacy deployments; driver plug‑ins support multiple backends.

#### 1.3 Delta SBOM Extension

When partial: true, only the missing layers have been scanned. Merging logic inside scanning module stitches new data onto the cached full SBOM in Valkey.

---

## 2 Valkey Keyspace

Valkey (Redis-compatible) provides cache, DPoP nonces, event streams, and queues for real-time messaging and rate limiting.

Key pattern	Type	TTL	Purpose
scan:<digest>	string	∞	Last scan JSON result (as returned by /scan)
layers:<digest>	set	90d	Layers already possessing SBOMs (delta cache)
policy:active	string	∞	YAML or Rego ruleset
quota:<token>	string	until next UTC midnight	Per‑token scan counter for Free tier ({{ quota_token }} scans).
policy:history	list	∞	Change audit IDs (see PostgreSQL)
feed:nvd:json	string	24h	Normalised feed snapshot
locator:<imageDigest>	string	30d	Maps image digest → sbomBlobId
dpop:<jti>	string	5m	DPoP nonce cache (RFC 9449) for sender-constrained tokens
events:*	stream	7d	Event streams for Scheduler/Notify (Valkey Streams)
queue:*	stream	—	Task queues (Scanner jobs, Notify deliveries)
metrics:…	various	—	Prom / OTLP runtime metrics

Delta SBOM uses layers:* to skip work in <20 ms. Quota enforcement increments quota:<token> atomically; when {{ quota_token }} the API returns 429. DPoP & Events: Valkey Streams support high-throughput, ordered event delivery for re-evaluation and notification triggers. Alternative: NATS JetStream can replace Valkey for queues (opt-in only; requires explicit configuration).

---

3 PostgreSQL Tables

PostgreSQL is the canonical persistent store for long-term audit and history.

Table	Shape (summary)	Indexes
sbom_history	Wrapper JSON + replace_ts on overwrite	(image_digest) (created)
policy_versions	{id, yaml, rego, author_id, created}	(created)
attestations ⭑	SLSA provenance doc + Rekor log pointer	(image_digest)
audit_log	Fully rendered RFC 5424 entries (UI & CLI actions)	(user_id) (ts)

Schema detail for policy_versions:

Samples live under samples/api/scheduler/ (e.g., schedule.json, run.json, impact-set.json, audit.json) and mirror the canonical serializer output shown below.

{
  "_id": "6619e90b8c5e1f76",
  "yaml": "version: 1.0\nrules:\n  - …",
  "rego": null,                    // filled when Rego uploaded
  "authorId": "u_1021",
  "created": "2025-07-14T08:15:04Z",
  "comment": "Imported via API"
}

3.1 Scheduler Sprints 16 Artifacts

Tables. schedules, runs, impact_snapshots, audit (module-local). All rows use the canonical JSON emitted by StellaOps.Scheduler.Models so agents and fixtures remain deterministic.

3.1.1 Schedule (schedules)

{
  "id": "sch_20251018a",
  "tenantId": "tenant-alpha",
  "name": "Nightly Prod",
  "enabled": true,
  "cronExpression": "0 2 * * *",
  "timezone": "UTC",
  "mode": "analysis-only",
  "selection": {
    "scope": "by-namespace",
    "namespaces": ["team-a", "team-b"],
    "repositories": ["app/service-api"],
    "includeTags": ["canary", "prod"],
    "labels": [{"key": "env", "values": ["prod", "staging"]}],
    "resolvesTags": true
  },
  "onlyIf": {"lastReportOlderThanDays": 7, "policyRevision": "policy@42"},
  "notify": {"onNewFindings": true, "minSeverity": "high", "includeKev": true},
  "limits": {"maxJobs": 1000, "ratePerSecond": 25, "parallelism": 4},
  "subscribers": ["notify.ops"],
  "createdAt": "2025-10-18T22:00:00Z",
  "createdBy": "svc_scheduler",
  "updatedAt": "2025-10-18T22:00:00Z",
  "updatedBy": "svc_scheduler"
}

Constraints: arrays are alphabetically sorted; selection.tenantId is optional but when present must match tenantId. Cron expressions are validated for newline/length, timezones are validated via TimeZoneInfo.

#### 3.1.2 Run (runs)

{
  "_id": "run_20251018_0001",
  "tenantId": "tenant-alpha",
  "scheduleId": "sch_20251018a",
  "trigger": "conselier",
  "state": "running",
  "stats": {
    "candidates": 1280,
    "deduped": 910,
    "queued": 624,
    "completed": 310,
    "deltas": 42,
    "newCriticals": 7,
    "newHigh": 11,
    "newMedium": 18,
    "newLow": 6
  },
  "reason": {"conselierExportId": "exp-20251018-03"},
  "createdAt": "2025-10-18T22:03:14Z",
  "startedAt": "2025-10-18T22:03:20Z",
  "finishedAt": null,
  "error": null,
  "deltas": [
    {
      "imageDigest": "sha256:a1b2c3",
      "newFindings": 3,
      "newCriticals": 1,
      "newHigh": 1,
      "newMedium": 1,
      "newLow": 0,
      "kevHits": ["CVE-2025-0002"],
      "topFindings": [
        {
          "purl": "pkg:rpm/openssl@3.0.12-5.el9",
          "vulnerabilityId": "CVE-2025-0002",
          "severity": "critical",
          "link": "https://ui.internal/scans/sha256:a1b2c3"
        }
      ],
      "attestation": {"uuid": "rekor-314", "verified": true},
      "detectedAt": "2025-10-18T22:03:21Z"
    }
  ]
}

Counters are clamped to ≥0, timestamps are converted to UTC, and delta arrays are sorted (critical → info severity precedence, then vulnerability id). Missing deltas implies "no change" snapshots.

#### 3.1.3 Impact Snapshot (impact_snapshots)

{
  "selector": {
    "scope": "all-images",
    "tenantId": "tenant-alpha"
  },
  "images": [
    {
      "imageDigest": "sha256:f1e2d3",
      "registry": "registry.internal",
      "repository": "app/api",
      "namespaces": ["team-a"],
      "tags": ["prod"],
      "usedByEntrypoint": true,
      "labels": {"env": "prod"}
    }
  ],
  "usageOnly": true,
  "generatedAt": "2025-10-18T22:02:58Z",
  "total": 412,
  "snapshotId": "impact-20251018-1"
}

Images are deduplicated and sorted by digest. Label keys are normalised to lowercase to avoid case‑sensitive duplicates during reconciliation. snapshotId enables run planners to compare subsequent snapshots for drift.

#### 3.1.4 Audit (audit)

{
  "_id": "audit_169754",
  "tenantId": "tenant-alpha",
  "category": "scheduler",
  "action": "pause",
  "occurredAt": "2025-10-18T22:10:00Z",
  "actor": {"actorId": "user_admin", "displayName": "Cluster Admin", "kind": "user"},
  "scheduleId": "sch_20251018a",
  "correlationId": "corr-123",
  "metadata": {"details": "schedule paused", "reason": "maintenance"},
  "message": "Paused via API"
}

Metadata keys are lowercased, first‑writer wins (duplicates with different casing are ignored), and optional IDs (scheduleId, runId) are trimmed when empty. Use the canonical serializer when emitting events so audit digests remain reproducible.

#### 3.1.5 Run Summary (run_summaries)

Materialized view powering the Scheduler UI dashboards. Stores the latest roll-up per schedule/tenant, enabling quick “last run” banners and sparkline counters without scanning the full runs collection.

{
  "tenantId": "tenant-alpha",
  "scheduleId": "sch_20251018a",
  "updatedAt": "2025-10-18T22:10:10Z",
  "lastRun": {
    "runId": "run_20251018_0001",
    "trigger": "conselier",
    "state": "completed",
    "createdAt": "2025-10-18T22:03:14Z",
    "startedAt": "2025-10-18T22:03:20Z",
    "finishedAt": "2025-10-18T22:08:45Z",
    "stats": {
      "candidates": 1280,
      "deduped": 910,
      "queued": 0,
      "completed": 910,
      "deltas": 42,
      "newCriticals": 7,
      "newHigh": 11,
      "newMedium": 18,
      "newLow": 6
    },
    "error": null
  },
  "recent": [
    {
      "runId": "run_20251018_0001",
      "trigger": "conselier",
      "state": "completed",
      "createdAt": "2025-10-18T22:03:14Z",
      "startedAt": "2025-10-18T22:03:20Z",
      "finishedAt": "2025-10-18T22:08:45Z",
      "stats": {
        "candidates": 1280,
        "deduped": 910,
        "queued": 0,
        "completed": 910,
        "deltas": 42,
        "newCriticals": 7,
        "newHigh": 11,
        "newMedium": 18,
        "newLow": 6
      },
      "error": null
    },
    {
      "runId": "run_20251017_0003",
      "trigger": "cron",
      "state": "error",
      "createdAt": "2025-10-17T22:01:02Z",
      "startedAt": "2025-10-17T22:01:08Z",
      "finishedAt": "2025-10-17T22:04:11Z",
      "stats": {
        "candidates": 1040,
        "deduped": 812,
        "queued": 0,
        "completed": 640,
        "deltas": 18,
        "newCriticals": 2,
        "newHigh": 4,
        "newMedium": 7,
        "newLow": 3
      },
      "error": "scanner timeout"
    }
  ],
  "counters": {
    "total": 3,
    "planning": 0,
    "queued": 0,
    "running": 0,
    "completed": 1,
    "error": 1,
    "cancelled": 1,
    "totalDeltas": 60,
    "totalNewCriticals": 9,
    "totalNewHigh": 15,
    "totalNewMedium": 25,
    "totalNewLow": 9
  }
}

• _id combines tenantId and scheduleId (tenant:schedule).
• recent contains the 20 most recent runs ordered by createdAt (UTC). Updates replace the existing entry for a run to respect state transitions.
• counters aggregate over the retained window (20 runs) for quick trend indicators. Totals are recomputed after every update.
• Schedulers should call the projection service after every run state change so the cache mirrors planner/runner progress.

Sample file: samples/api/scheduler/run-summary.json.

---

## 4 Policy Schema (YAML v1.0)

Minimal viable grammar (subset of OSV‑SCHEMA ideas).

version: "1.0"
rules:
  - name: Block Critical
    severity: [Critical]
    action: block
  - name: Ignore Low Dev
    severity: [Low, None]
    environments: [dev, staging]
    action: ignore
    expires: "2026-01-01"
  - name: Escalate RegionalFeed High
    sources: [NVD, CNNVD, CNVD, ENISA, JVN, BDU]
    severity: [High, Critical]
    action: escalate

Validation is performed by policy:mapping.yaml JSON‑Schema embedded in backend.

Canonical schema source: src/Policy/__Libraries/StellaOps.Policy/Schemas/policy-schema@1.json (embedded into StellaOps.Policy).
PolicyValidationCli (see src/Policy/__Libraries/StellaOps.Policy/PolicyValidationCli.cs) provides the reusable command handler that the main CLI wires up; in the interim it can be invoked from a short host like:

await new PolicyValidationCli().RunAsync(new PolicyValidationCliOptions
{
    Inputs = new[] { "policies/root.yaml" },
    Strict = true,
});

### 4.1 Rego Variant (Advanced – TODO)

Accepted but stored as‑is in rego field.
Evaluated via internal OPA side‑car once feature graduates from TODO list.

### 4.2 Policy Scoring Config (JSON)

Schema id. https://schemas.stella-ops.org/policy/policy-scoring-schema@1.json
Source. src/Policy/__Libraries/StellaOps.Policy/Schemas/policy-scoring-schema@1.json (embedded in StellaOps.Policy), default fixture at src/Policy/__Libraries/StellaOps.Policy/Schemas/policy-scoring-default.json.

{
  "version": "1.0",
  "severityWeights": {"Critical": 90, "High": 75, "Unknown": 60, "...": 0},
  "quietPenalty": 45,
  "warnPenalty": 15,
  "ignorePenalty": 35,
  "trustOverrides": {"vendor": 1.0, "distro": 0.85},
  "reachabilityBuckets": {"entrypoint": 1.0, "direct": 0.85, "runtime": 0.45, "unknown": 0.5},
  "unknownConfidence": {
    "initial": 0.8,
    "decayPerDay": 0.05,
    "floor": 0.2,
    "bands": [
      {"name": "high", "min": 0.65},
      {"name": "medium", "min": 0.35},
      {"name": "low", "min": 0.0}
    ]
  }
}

Validation occurs alongside policy binding (PolicyScoringConfigBinder), producing deterministic digests via PolicyScoringConfigDigest. Bands are ordered descending by min so consumers can resolve confidence tiers deterministically. Reachability buckets are case-insensitive keys (entrypoint, direct, indirect, runtime, unreachable, unknown) with numeric multipliers (default ≤1.0).

Runtime usage

• trustOverrides are matched against finding.tags (trust:<key>) first, then finding.source/finding.vendor; missing keys default to 1.0.
• reachabilityBuckets consume finding.tags with prefix reachability: (fallback usage: or unknown). Missing buckets fall back to unknown weight when present, otherwise 1.0.
• Policy verdicts expose scoring inputs (severityWeight, trustWeight, reachabilityWeight, baseScore, penalties) plus unknown-state metadata (unknownConfidence, unknownAgeDays, confidenceBand) for auditability. See samples/policy/policy-preview-unknown.json and samples/policy/policy-report-unknown.json for offline reference payloads validated against the published schemas below.

Validate the samples locally with Ajv before publishing changes:

# install once per checkout (offline-safe):
npm install --no-save ajv-cli@5 ajv-formats@2

npx ajv validate --spec=draft2020 -c ajv-formats \
  -s docs/schemas/policy-preview-sample@1.json \
  -d samples/policy/policy-preview-unknown.json

npx ajv validate --spec=draft2020 -c ajv-formats \
  -s docs/schemas/policy-report-sample@1.json \
  -d samples/policy/policy-report-unknown.json

• Unknown confidence derives from unknown-age-days: (preferred) or unknown-since: + observed-at: tags; with no hints the engine keeps initial confidence. Values decay by decayPerDay down to floor, then resolve to the first matching bands[].name.

---

## 5 SLSA Attestation Schema ⭑

Planned for Q1‑2026 (kept here for early plug‑in authors).

{
  "id": "prov_0291",
  "imageDigest": "sha256:e2b9…",
  "buildType": "https://slsa.dev/container/v1",
  "builder": {
    "id": "https://git.stella-ops.ru/ci/stella-runner@sha256:f7b7…"
  },
  "metadata": {
    "invocation": {
      "parameters": {"GIT_SHA": "f6a1…"},
      "buildStart": "2025-07-14T06:59:17Z",
      "buildEnd": "2025-07-14T07:01:22Z"
    },
    "completeness": {"parameters": true}
  },
  "materials": [
    {"uri": "git+https://git…", "digest": {"sha1": "f6a1…"}}
  ],
  "rekorLogIndex": 99817    // entry in local Rekor mirror
}

---

## 6 Notify Foundations (Rule · Channel · Event)

Sprint 15 target – canonically describe the Notify data shapes that UI, workers, and storage consume. JSON Schemas live under docs/modules/notify/resources/schemas/ and deterministic fixtures under docs/modules/notify/resources/samples/.

Artifact	Schema	Sample
Rule (catalogued routing logic)	docs/modules/notify/resources/schemas/notify-rule@1.json	docs/modules/notify/resources/samples/notify-rule@1.sample.json
Channel (delivery endpoint definition)	docs/modules/notify/resources/schemas/notify-channel@1.json	docs/modules/notify/resources/samples/notify-channel@1.sample.json
Template (rendering payload)	docs/modules/notify/resources/schemas/notify-template@1.json	docs/modules/notify/resources/samples/notify-template@1.sample.json
Event envelope (Notify ingest surface)	docs/modules/notify/resources/schemas/notify-event@1.json	docs/modules/notify/resources/samples/notify-event@1.sample.json

### 6.1 Rule highlights (notify-rule@1)

• Keys are lower‑cased camelCase. schemaVersion (notify.rule@1), ruleId, tenantId, name, match, actions, createdAt, and updatedAt are mandatory.
• match.eventKinds, match.verdicts, and other array selectors are pre‑sorted and case‑normalized (e.g. scanner.report.ready).
• actions[].throttle serialises as ISO 8601 duration (PT5M), mirroring worker backoff guardrails.
• vex gates let operators exclude accepted/not‑affected justifications; omit the block to inherit default behaviour.
• Use StellaOps.Notify.Models.NotifySchemaMigration.UpgradeRule(JsonNode) when deserialising legacy payloads that might lack schemaVersion or retain older revisions.
• Soft deletions persist deletedAt in PostgreSQL (and disable the rule); repository queries automatically filter them.

### 6.2 Channel highlights (notify-channel@1)

• schemaVersion is pinned to notify.channel@1 and must accompany persisted documents.
• type matches plug‑in identifiers (slack, teams, email, webhook, custom).
• config.secretRef stores an external secret handle (Authority, Vault, K8s). Notify never persists raw credentials.
• Optional config.limits.timeout uses ISO 8601 durations identical to rule throttles; concurrency/RPM defaults apply when absent.
• StellaOps.Notify.Models.NotifySchemaMigration.UpgradeChannel(JsonNode) backfills the schema version when older documents omit it.
• Channels share the same soft-delete marker (deletedAt) so operators can restore prior configuration without purging history.

### 6.3 Event envelope (notify-event@1)

• Aligns with the platform event contract—eventId UUID, RFC 3339 ts, tenant isolation enforced.
• Enumerated kind covers the initial Notify surface (scanner.report.ready, scheduler.rescan.delta, zastava.admission, etc.).
• scope.labels/scope.attributes and top-level attributes mirror the metadata dictionaries workers surface for templating and audits.
• Notify workers use the same migration helper to wrap event payloads before template rendering, so schema additions remain additive.

### 6.4 Template highlights (notify-template@1)

• Carries the presentation key (channelType, key, locale) and the raw template body; schemaVersion is fixed to notify.template@1.
• renderMode enumerates supported engines (markdown, html, adaptiveCard, plainText, json) aligning with NotifyTemplateRenderMode.
• format signals downstream connector expectations (slack, teams, email, webhook, json).
• Upgrade legacy definitions with NotifySchemaMigration.UpgradeTemplate(JsonNode) to auto-apply the new schema version and ordering.
• Templates also record soft deletes via deletedAt; UI/API skip them by default while retaining revision history.

Validation loop:

# Validate Notify schemas and samples (matches Docs CI)
for schema in docs/modules/notify/resources/schemas/*.json; do
  npx ajv compile -c ajv-formats -s "$schema"
done

for sample in docs/modules/notify/resources/samples/*.sample.json; do
  schema="docs/modules/notify/resources/schemas/$(basename "${sample%.sample.json}").json"
  npx ajv validate -c ajv-formats -s "$schema" -d "$sample"
done

Integration tests can embed the sample fixtures to guarantee deterministic serialisation from the StellaOps.Notify.Models DTOs introduced in Sprint 15.

---

## 6 Validator Contracts

• For SBOM wrapper – ISbomValidator (DLL plug‑in) must return typed error list.
• For YAML policies – JSON‑Schema at /schemas/policy‑v1.json.
• For Rego – OPA opa eval --fail-defined under the hood.
• For Free‑tier quotas – IQuotaService integration tests ensure quota:<token> resets at UTC midnight and produces correct Retry‑After headers.

---

## 7 Migration Notes

1. Add format column to existing SBOM wrappers; default to trivy-json-v2.
2. Populate layers & partial via backfill script (ship with stellopsctl migrate wizard).
3. Policy YAML previously stored in Valkey → copy to PostgreSQL if persistence enabled.
4. Prepare attestations table (empty) – safe to create in advance.

---

## 8 Open Questions / Future Work

• How to de‑duplicate identical Rego policies differing only in whitespace?
• Embed GOST 34.11‑2018 digests when users enable Russian crypto suite?
• Should enterprise tiers share the same Redis quota keys or switch to JWT claim tier != Free bypass?
• Evaluate sliding‑window quota instead of strict daily reset.
• Consider rate‑limit for /layers/missing to avoid brute‑force enumeration.

---

## 9 Change Log

Date	Note
2025‑07‑14	Added: format, partial, delta cache keys, YAML policy schema v1.0.
2025‑07‑12	Initial public draft – SBOM wrapper, Redis keyspace, audit collections.

---