MkDocs retires and the Astro app becomes the only public site, so the artifact is the product and the Markdown stays the source -- the split tools/newsgen already states for releases. tools/docsite walks the published set and writes docs/data/generated/docs.json: every page's directory-style url, title, section and raw Markdown body, plus the sidebar tree derived from the layout. Three things the artifact has to get right, each of which was wrong in a draft: - LINKS. The published pages carry relative .md links that MkDocs rewrote at build time and nothing else will. Each is rewritten to the target page's url using the same function that builds urls, so a link and its target cannot disagree; anchors and scheme-qualified links pass through; assets resolve by the same rule and none are referenced. - FRAGMENTS. A link can resolve to a page and still be broken, because a fragment matching no heading lands the reader at the top of a long document. Each is checked against the target's headings, by the slug rule a table of contents uses. - REFUSAL. A link whose target the set holds back, or whose fragment matches no heading, fails the generator naming the file, the line and the target. This found three citations from settled architecture pointing at two PRDs the blanket prds/* exclusion withheld -- links that were already 404s on the published site, so the exclusion broke citations rather than withholding drafts. Both targets are now published, for a reason distinct from the three ratified PRDs already published. The published set moves out of mkdocs.yml into the generator with its reasons: a renderer reading the artifact cannot see why a page is absent. docs/data/generated is accounted for by docsgen's generatedArtifacts, which now names this file beside contracts.json -- one ownership registry for the directory rather than one per writer, so neither tool reports the other's artifact as unaccounted. docs:artifact regenerates; docs:artifact:check reproduces into memory and fails on drift without touching the tree, wired into task check beside news:check. The existing docs:site target is MkDocs and was left alone -- a first attempt at that name silently ran MkDocs and produced nothing. Verified: task check exit 0, docs:artifact:check reporting 23 pages. The artifact carries 23 pages, no archive/ inspiration/ or news/ entry, and the five PRDs the set declares. Coverage: the published set and its count, url derivation, the nav tree, both refusal paths, and check's no-write guarantee.
15 KiB
Generated Documentation
Status: Generator scaffold in place; renderer and publishing removed Date: 2026-07-28 (revised 2026-09-12) Tracking issue: #73
Purpose
Archie's documentation-generation approach is:
- production Go code, registries, configuration, and CLI behaviour remain the source of truth for published contracts;
- a repository-local Go generator extracts and normalizes documentation data;
- the generator writes deterministic JSON/data files and protocol schemas;
- the rendering and publishing surface is an open decision — the VitePress site and its Pages deployment were removed on 2026-09-12 as an unnecessary build step (see "Rendering and publishing" below).
Hugo is not part of the target architecture.
The documentation system is rooted at:
tools/
go.mod
go.sum
docsgen/
main.go
docs/
data/generated/
public/schemas/
*.md
Rendering and publishing — OPEN
Superseded 2026-09-12. The repository previously carried a VitePress site
(docs/.vitepress/config.mts, docs/package.json, a 1620-line
docs/pnpm-lock.yaml, docs/pnpm-workspace.yaml) built and deployed to GitHub
Pages by .github/workflows/docs.yml. It was removed as an unnecessary build
step: the published site was never stood up, and the Pages target is not
configured.
docs/ is now repository documentation only — markdown read in the repository,
an editor, or a forge's file view. Nothing in the repository builds, renders, or
publishes it.
The renderer for the generated data/generated/*.json is therefore undecided.
Consequences that follow, and their current state:
| Concern | State after removal |
|---|---|
| Reference-page rendering | Undecided; data/generated/contracts.json is committed but unrendered |
| Internal link validation | Not enforced by any build; links are repository-relative (../architecture/...) and resolve on the filesystem |
| Generated-data drift detection | Unchanged — docsgen check compares committed output, independent of any renderer |
| Navigation/sidebar | The authoritative index is docs/prds/01-project-management.md |
| Local authoring server | None; read markdown directly |
A future renderer MUST NOT reintroduce a dead-link-blind build: VitePress previously validated internal links during its production build, and no replacement currently does.
Single source of truth
Every published capability has one authoritative machine-readable definition owned by its domain or application boundary. Generated documentation derives from those definitions.
Generated reference includes:
- domain commands and events;
- Messages and typed message content;
- Agent and Workflow contracts;
- WorkflowSteps and Workflow plugin capability contracts;
- policies and their typed inputs and outcomes;
- tools and capabilities;
- CLI commands, arguments, flags, defaults, and help;
- runtime settings, types, defaults, validation, and secret metadata;
- public worker and transport contracts;
- indexes of registered functionality;
- deprecation information.
Handwritten documents contain:
- architecture decisions and rationale;
- tutorials and operational guides;
- examples and walkthroughs;
- migration guidance;
- conceptual explanations.
Handwritten documents link to generated reference rather than copying authoritative tables.
Target generation architecture
tools/docsgen is Archie's documentation extractor and normalizer. It produces
structured documentation data and protocol specifications. It does not own
page layout and MUST NOT generate Markdown pages or fragments.
The generator is divided internally into reusable stages:
load registries and types
-> normalize documentation model
-> validate ownership and identifiers
-> serialize data and protocol artifacts
-> compare or write output
The normalized documentation model prevents each output renderer from independently interpreting Go types. It contains stable documentation IDs, owners, names, descriptions, source references, versions, deprecations, fields, defaults, validation, secret classifications, and relationships.
Initial generator targets:
docsgen data
docsgen asyncapi
docsgen all
docsgen check
datawrites the normalized JSON reference data.asyncapiwrites protocol schemas and envelopes.allgenerates every committed artifact.checkgenerates into a temporary directory and fails on drift.
The command may use subcommands or an equivalent explicit --target flag. It
MUST retain one normalization and validation path.
Target output
During adaptation, docs remains the staging site. Generated outputs live in
a visibly owned subtree:
docs/
data/
generated/
catalog.json
contracts.json
docs.json
cli.json
configuration.json
messaging.json
agents.json
workflows.json
policies.json
plugins.json
tools.json
public/
schemas/
asyncapi-header.yaml
archie.yaml
.tmp/
All files under data/generated/ and public/schemas/ carry generated
ownership metadata where their format permits it and MUST NOT be edited
manually. .tmp/ is generator-owned, ignored, and never committed.
catalog.json is the deterministic machine-readable index used for:
- reference navigation and route inputs;
- duplicate-ID and duplicate-route validation;
- coverage checks;
- source-to-data drift diagnostics;
- future external consumers.
These files are consumed at render time by whichever renderer is chosen (see
"Rendering and publishing"); the generator's contract is the committed JSON, not
any particular loader. A renderer reads data/generated/*.json directly — Archie
does not use createContentLoader-style Markdown collection scanning, because
the authoritative inputs are JSON.
The final move from docs/ to docs/ occurs only after Archie content
contains intentional Archie pages and the current architecture documents have an
intentional destination. The staging name MUST NOT leak into public URLs.
Generation steps
Step 1: maintain the Archie tools module
Keep tools/go.mod isolated from the runtime module:
module github.com/samcharles93/archie-core/tools
require github.com/samcharles93/archie-core v0.0.0-00010101000000-000000000000
replace github.com/samcharles93/archie-core => ../
Keep generator-only dependencies such as invopop/jsonschema and any future
protocol-schema serializer in the nested tools module. They MUST NOT enter the
root module.
docsgen imports Archie-owned contracts and writes normalized Archie data and
protocol artifacts.
Step 2: establish authoritative registries
The generator consumes domain-owned registries rather than scanning arbitrary packages for anything exported.
Each published definition supplies:
- stable documentation ID;
- owning domain or boundary;
- name and description;
- version when versioned;
- the Go type or declarative definition;
- lifecycle and deprecation metadata;
- source reference.
Registries remain executable application definitions used by production code. A documentation-only duplicate registry is prohibited.
Where the migration has not yet created a final domain registry, the generator may use a narrow compatibility adapter over the current registry. That adapter has explicit deletion criteria.
The initial agent-execution and messaging schema set is such an adapter. It is
deleted when those domains expose their production-owned documentation
registries; adding another hardcoded type list to docsgen is prohibited.
Step 3: generate AsyncAPI
Use Archie's messaging and worker wire registries as the authoritative inputs.
The static header continues to own:
- AsyncAPI version;
- service and protocol description;
- servers;
- channels;
- operations;
- message envelope placement.
The generator owns:
- component schemas;
- discriminated concrete message lists;
- generated envelopes;
- rewritten schema references;
- stable ordering.
The generated protocol artifact becomes:
docs/public/schemas/archie.yaml
The final owner of each wire registry is the domain or capability defining its
meaning. docsgen renders the contract; it does not own the message semantics.
Step 4: add generated documentation data
Add extractors for the missing reference categories one vertical slice at a time.
Recommended implementation order:
- CLI commands and flags;
- runtime configuration/settings;
- Messaging commands, events, and message types;
- Agent definitions and public commands;
- Workflow definitions, WorkflowSteps, lifecycle, and plugin contracts;
- policy definitions and typed outcomes;
- tools and capability contracts;
- worker and external wire contracts.
Each category extractor contributes:
- stable IDs, owners, routes, and display ordering;
- fields, types, defaults, and validation metadata;
- source references;
- version and deprecation information;
- relationships to commands, events, policies, and contracts.
The first vertical slice proves the normalized model and deterministic JSON serialization. Later categories reuse it rather than introducing category-specific ad hoc output formats.
Step 5: render data (pending renderer decision)
Blocks on "Rendering and publishing" above. Once a renderer is chosen, it reads
data/generated/*.json, validates it, and renders reference views.
Navigation and routes derive from catalog.json; they MUST NOT be duplicated by
hand in renderer configuration.
No renderer is chosen, so this step is unstarted.
Step 6: implement deterministic drift checking
docsgen check:
- generates all artifacts into a temporary directory;
- normalizes line endings and file modes;
- compares the complete expected data and schema trees with committed outputs;
- reports missing, changed, and obsolete files;
- identifies the authoritative definition associated with each mismatch;
- exits non-zero without modifying the working tree.
Generation MUST exclude:
- wall-clock timestamps;
- absolute filesystem paths;
- map iteration order;
- environment-specific values;
- secrets and resolved credentials.
Two consecutive docsgen all runs MUST be byte-identical.
Step 7: renderer adaptation — blocked
Was "adapt the VitePress site". No renderer is chosen; this step is unstarted.
When a renderer is selected it MUST:
- read
data/generated/*.jsonrather than re-deriving from source; - validate internal links, with dead-link failures enabled (no blanket ignore list);
- state its own local preview and build commands;
- keep
node_modules/and any build output out of source control.
Step 8: developer commands
task docs:generate writes the committed artifact; task docs:check verifies
it. task check runs docs:check, so generated drift is gated rather than
advisory: a stale artifact fails the gate and names the authoritative Go
definition behind each mismatch.
task docs:generate
task docs:check
Both wrap the generator, which can still be run directly:
go -C tools run -mod=readonly ./docsgen --repo-root ..
go -C tools run -mod=readonly ./docsgen check --repo-root ..
docsgen all is still target-only: docsgen data and docsgen asyncapi do not
exist. Two artifacts are generated today: contracts.json from the Go
contract types (docsgen) and docs.json from the Markdown sources
(docsite, the published set a renderer reads) -- see task docs:artifact.
Step 9: CI
The original documentation workflow was deleted on 2026-09-12. deploy.yml still
does not reference docs/**.
.github/workflows/docs.yml (added 2026-09-18) builds the MkDocs site with
mkdocs build --strict on changes to docs/** or mkdocs.yml, so a dead
internal link fails before it reaches a deploy. It deliberately does no more
than that: it does not run docsgen, does not regenerate contracts.json, and
does not publish. Deployment is a separate path, run from the Ansible control
node by the vectislabs.automation docs_site role.
Generated-artifact drift is therefore still gated locally, by task check
running docs:check, and not in CI. That split is the point. A CI job that
regenerated output would be able to paper over uncommitted differences instead
of failing on them, which is the failure this section exists to prevent, so any
future documentation workflow MUST check generated output rather than produce
it.
Step 10: cutover — done, in reverse
The site cutover is moot: the site was removed rather than published. The
generator's outputs (docs/data/generated/, docs/public/schemas/) remain
committed and are the durable artifacts.
Still standing from the original cutover list:
- remove Markdown, examples, specs, generated schemas, and assets that do not describe Archie;
- adapt retained content and prove it against Archie code;
- update
docs.goand its tests to embed only intentional runtime documentation; - decide which current
docs/prdsremain published architecture pages and which remain repository-only planning documents — all of them are repository-only now, so the distinction is retired.
Documents MUST describe Archie. Publishing is a separate, undecided concern.
Hook and CI policy
Generated documentation is committed.
No pre-commit hook is installed in this repository, and no CI step runs the
generator. Step 3 below is enforced by task check; steps 1, 2, and 4 still
rely on the author, because nothing observes a commit:
- detects authoritative definition or generator changes;
- runs the generator;
- runs
docsgen check— enforced bytask check; - refuses the commit when generated changes are unstaged or validation fails.
A commit MUST NOT be created while generated documentation is stale.
Validation requirements
Generator tests cover:
- stable ordering;
- schema reference rewriting;
$defsflattening;- duplicate IDs and routes;
- registry coverage;
- comments and descriptions;
- secret redaction;
- obsolete-file removal;
- byte-identical repeated generation;
- check-mode diagnostics.
Repository-documentation validation covers:
- internal links resolve on the filesystem (repository-relative paths);
- absence of unrelated product names and URLs;
- absence of committed build artifacts and installed dependencies.
Renderer validation (frozen dependency installation, production build, expected output routes) is added with the renderer decision and is currently unenforced.
The root module remains free of generator-only dependencies.
Completion criteria
The documentation migration is complete when:
toolsbelongs to Archie and builds against the working tree;docsgenconsumes Archie-owned registries and types;- AsyncAPI output describes Archie;
- required reference categories are generated as committed deterministic data;
catalog.jsonmatches the generated definitions and rendered routes;- two consecutive generations are byte-identical;
docsgen checkdetects changed, missing, and obsolete artifacts;- build artifacts and installed dependencies are absent from source control;
task docs:generateandtask docs:checkwork from a clean checkout;- every document describes Archie and every internal link resolves.
Renderer-dependent criteria (rendered reference pages, local authoring server, deployable build, published site) are removed pending the renderer decision.
Implementation references
- GitHub-flavoured Markdown specification — the dialect repository documentation is written in
- pnpm — required only if a future renderer chooses a Node-based toolchain