Catalog schema and reuse

The catalog is a UTF-8 JSON envelope with schemaVersion, revision, generatedAt, site, guidance, listings, taxonomies, changes and advisories. Schema version 1 permits additive fields. Stable listing and taxonomy IDs are lowercase slugs. Research candidates are separately presented under review and are not admitted catalog records.

Revisions and focused files

revision is a SHA-256 digest of public listings, taxonomies, advisories and editorial change history. It excludes build time. generatedAt identifies the build, not a source review. Focused JSON has catalogRevision, catalog, url, kind and the record/subset. All focused formats preserve the parent catalog’s source qualifications. Examples and MCP tools.

Listing fields

Records include identity, description, categories and category-fit notes; documented clouds and cloudEvidence; installation, architecture and operations; licensing and infrastructure classifications with source notes; publisher, kind and optional parent; source URLs and human review dates. installMethods uses the IDs in taxonomies.installMethods. Installation facets summarize the recorded methods and do not certify a topology for production.

A solution is a configured use of its parent platform. A named cloud identifies documented evidence, not exclusive compatibility or an endorsement. publisher: true marks FormKiQ entries.

Evidence classification

Cloud evidence contains title, url, clouds, notes, sourceType, reviewed and vendorSupport. Source types distinguish vendor documentation, provider guidance, marketplace offers, package-platform documentation, third-party packages, community templates, partner evidence/relationships, institution deployments, corporate disclosures and public procurement records. Read the notes rather than treating these as a single quality score.

Optional applicableEdition and applicableVersions qualify specific sources. Missing values do not establish a current-release match. vendorSupport: not-assessed explicitly means a current support commitment for that exact deployment has not been established by this directory.

Buyer diligence

Optional diligence records use field IDs defined in taxonomies.diligenceFields. They cover marketplace delivery, commitment eligibility, SSO, SCIM, compliance, security controls, air-gapped operation, deployment artifacts, footprint, license pricing, operating cost and releases.

A record includes status, summary, scope, source and reviewed:

Optional supportingSources provides additional titled evidence links. artifactUrl/artifactRef identify a deployment artifact and release or explicitly moving reference; a moving branch is not a pinned release. Marketplace facts can include offers, whose records retain cloud, id, publisher, delivery, url and optional version. A platform/package version does not replace the upstream application version, and an offer does not establish commitment eligibility. Release snapshots can include releaseTag and either releasePublishedAt (a source-supplied timestamp) or releasePublishedOn (a publication date, YYYY-MM-DD, when no time is supplied). A date-only source never implies a midnight timestamp. These fields do not establish support, compatibility, an upgrade path or future cadence.

securityControls describes documented safeguards; compliance requires assessed scope rather than hosted-service marketing. licensePricing records published software terms or a documented requirement for a quote; cost requires a dated workload/region/resource model. None implies another field. Cloud commitments depend on the exact offer and buyer agreement. A vendor’s SaaS certification is not automatically a certification of a customer-hosted installation.

Cost findings can include a requested scenario and a worked estimate. A scenario records the target provider, region, users, data and availability; it does not establish deployment support. Check the listing’s cloud evidence separately. An estimate identifies its currency, period, region, workload, cost rows and exclusions. Each row has an ID, quantity, rate and description; a blank rate is unknown, not zero. kind: partial-infrastructure prices only the stated components and remains inconclusive for total operating cost. Missing categories, licences, labor and excluded services must not be treated as free. Planning quantities are not vendor capacity benchmarks or availability guarantees.

Configured solutions do not inherit buyer facts automatically. Research questions and source-discovery notes remain outside the public catalog.

History and corrections

changes contains a tracking start date, its limitations, and editorial events with stable ID, date, type, summary, affected listing IDs and fields. Types are baseline, added, updated, removed, evidence-changed, security-advisory and lifecycle. Dates describe recorded editorial events, not inferred product launch dates. The public changelog and RSS use the same events.

Attribution and reuse

FormKiQ publishes and maintains this reference. When citing an entry, link its canonical listing URL, preserve the source and edition qualifications, include its review date, and disclose FormKiQ’s publisher relationship where relevant. Third-party documentation, product names and software licenses retain their respective terms.

The catalog does not currently grant a blanket open-data license for bulk republication of its original descriptions or classifications. Contact the publisher for bulk reuse permission. The four evaluation worksheets separately permit copying and adaptation for your own evaluations.

Evidence filters, provenance and discovery relationships

taxonomies.diligenceFilters defines supported-option filters. All selected requirements must match. Positive availability must be explicit; documented limitations and unresolved or unreviewed statements do not qualify. Security-control evidence identifies reviewed safeguards, not certification.

Research completeness counts documented and inconclusive fields separately across the declared buyer fields. It measures research coverage, not a feature score. Unknown fields remain available in HTML details, Markdown and JSON.

reviewStatus is omitted for existing editor-reviewed facts or set to editor-reviewed after checking a submission. Pending statements never satisfy filters or increase researched coverage. provenance.kind can record vendor-submission or partner-submission; it describes origin, separately from editorial verification.

alternativeTo records a target ID, vendor-positioning source, summary and review date. Relationships do not establish feature parity. Listing advisories and the top-level advisory index preserve review date, exact scope, conditions, action and sources. No flag does not mean safe or supported.

A cost fact can contain a narrowly scoped worked illustration with estimate.currency, period, resource rows and exclusions. A partial infrastructure subtotal must be labeled as partial, with dated public rates and no implied production-capacity guarantee. It is not a complete operating budget. Keep the cost field inconclusive until the complete workload/region/resource and operations model is established; a verified component rate alone does not close that gap.

Deployment boundaries and operating responsibilities

Provider navigation groups are recorded as clouds[].providerType and defined in taxonomies.providerTypes. They group the documented provider offering, not every service that provider sells. Application-specific paths take precedence: DigitalOcean App Platform and DigitalOcean Droplets have different boundaries.

Each listing adds modelLabel and deploymentPaths. A path contains cloud (the existing provider ID), boundary (see taxonomies.deploymentBoundaries), operator (see taxonomies.operationModels), controlPlane (vendor-hosted or not-assessed) and evidence (zero-based indexes into that same listing’s cloudEvidence array). Treat provider, boundary and operator as a single path when filtering; matching separate paths does not establish a valid combination. not-assessed means the specific boundary or control-plane location is unconfirmed.

operationReview, when present, supplies a dated, source-linked review of the operating split. Other customer-operated labels preserve the existing source-backed self-hosting classification and do not assert the absence of vendor services or support access.

The original model strings remain as a compatibility field in schema version 1. They are broad historical buckets, not a current ownership or operating guarantee. Use modelLabel, paths and the source qualifications for current answers. Existing model=BYOC shared links retain their historical connected-deployment selection; the new operator filters distinguish customer, vendor/partner and shared operations.

Question answers

The optional questions array contains public editorial answers. Each record has a stable id, number, group, question, Markdown answer, plain text, canonical url, json and markdown links, related links, and a source-review date. Every answer has its own focused HTML page (dedicatedPage: true), with a short answer followed by examples or evaluation checks and supporting guidance. Some records identify a related glossary term, a provider whose listing count is calculated from the catalog, related listingIds, or additional public sources. taxonomies.questionGroups defines the topic IDs. These fields do not certify products or change the scope of listing evidence.

All questions have an independent JSON index, and every question has focused JSON and Markdown files. The MCP catalog resource retains the same records; the listing tools continue to return listings. These editorial additions change the catalog revision. A rebuild alone does not.