Add the backfill census: changelog backfill inventory

docs/cli-schema.json

@@ -4433,7 +4433,96 @@
           ]
         }
       ],
-      "namespaces": []
+      "namespaces": [
+        {
+          "segment": "backfill",
+          "summary": "Backfill historical release-note bundles (docs-eng-team#656): census, planning, and guarded publication.",
+          "options": [],
+          "commands": [
+            {
+              "path": [
+                "changelog",
+                "backfill"
+              ],
+              "name": "inventory",
+              "summary": "Build the backfill census: an inventory document covering every release-notes product.",
+              "notes": "Enumerates every product in products.yml that participates in release notes (the\nrelease-notes feature defaults to enabled, so products.yml alone cannot say which\nproducts have release-note surfaces), merges in the hand-maintained census seed, and writes the\nversioned inventory document that backfill planning consumes. Products the seed does not cover\nstay visible as source-unresolved entries with a warning \u2014 an unresolved scope must never\nsilently produce empty bundles. Attributed repositories are checked against the link allowlist in\nthe local assembler.yml; planning re-validates against the deployed scrubber allowlist\nbefore any upload. This command only reads configuration and writes a local file: no S3 access,\nno writes to any remote system.",
+              "usage": "docs-builder changelog backfill inventory [options]",
+              "examples": [],
+              "parameters": [
+                {
+                  "role": "flag",
+                  "name": "sources",
+                  "type": "string",
+                  "required": false,
+                  "summary": "Path to the census seed YAML mapping products to their release-note sources. Without it, every release-notes product is reported as source-unresolved.",
+                  "validations": [
+                    {
+                      "kind": "rejectSymbolicLinks"
+                    },
+                    {
+                      "kind": "existing"
+                    },
+                    {
+                      "kind": "fileExtensions",
+                      "values": [
+                        "yml",
+                        "yaml"
+                      ]
+                    }
+                  ]
+                },
+                {
+                  "role": "flag",
+                  "name": "output",
+                  "type": "string",
+                  "required": false,
+                  "summary": "Where to write the inventory document JSON.",
+                  "defaultValue": "backfill-inventory.json"
+                },
+                {
+                  "role": "flag",
+                  "name": "log-level",
+                  "shortName": "l",
+                  "type": "enum",
+                  "required": false,
+                  "summary": "Minimum log level. Default: information",
+                  "enumValues": [
+                    "trace",
+                    "debug",
+                    "information",
+                    "warning",
+                    "error",
+                    "critical",
+                    "none"
+                  ]
+                },
+                {
+                  "role": "flag",
+                  "name": "config-source",
+                  "shortName": "c",
+                  "type": "enum",
+                  "required": false,
+                  "summary": "Override the configuration source: local, remote",
+                  "enumValues": [
+                    "local",
+                    "remote",
+                    "embedded"
+                  ]
+                },
+                {
+                  "role": "flag",
+                  "name": "skip-private-repositories",
+                  "type": "boolean",
+                  "required": false,
+                  "summary": "Skip cloning private repositories"
+                }
+              ]
+            }
+          ],
+          "namespaces": []
+        }
+      ]
     },
     {
       "segment": "codex",

docs/cli/changelog/backfill/cmd-inventory.md

@@ -0,0 +1,59 @@
+## Description
+
+Build the backfill census: an inventory document covering every release-notes product.
+
+`config/products.yml` cannot say which products have release-note surfaces on its own, because the `release-notes` feature defaults to enabled. This command enumerates every product that participates in release notes, merges in a hand-maintained census seed mapping products to their sources, and writes the versioned inventory document that backfill planning consumes.
+
+Products the seed does not cover stay visible as `source-unresolved` entries and produce a warning — "we looked and decided no" must always be distinguishable from "we never looked", and an unresolved scope can never silently produce empty bundles. Products can be deliberately deferred in the seed's `unmapped` list, each with a reason, which records the deferral without a warning.
+
+The command is read-only apart from the local output file: it reads configuration, no S3 access, no remote writes.
+
+### Defaults applied by the census
+
+- Sources whose products are all stack-versioned and whose scheme is `semver` get the epic's default cutoff of `9.0.0` when the seed does not specify one.
+- Unresolved products get a target scheme derived from their versioning system (`serverless`/project versioning → `date`, `ech` → `monthly`, everything else → `semver`), always paired with an unresolved note so a guess never reads as a confirmed fact.
+- Attributed repositories are checked against the link allowlist in the local `assembler.yml`. Planning re-validates against the **deployed** scrubber allowlist identity before any upload; the census status is advisory.
+
+## Seed format
+
+```yaml
+sources:
+  - repository: elastic/docs-content    # where the release-note content lives
+    git_ref: main
+    docset: docs-content                # optional
+    paths:
+      - release-notes/elasticsearch
+    products: [elasticsearch]           # products.yml ids
+    target_scheme: semver               # semver | date | monthly
+    cutoff:                             # optional; stack semver defaults to 9.0.0
+      kind: version                     # version | date
+      value: 9.0.0
+      notes: optional free text
+    substitutions: {}                   # docset variable expansions
+    link_mappings: {}                   # source link -> canonical destination
+    attributed_repositories:            # repos entries attribute changes to
+      - elastic/elasticsearch
+    default_repository: elastic/elasticsearch
+    bundle_filename_convention: "{repo}-{target}.yaml"
+    adoption: not-adopted               # not-adopted | partially-adopted | fully-adopted
+    classification: published-history-found
+    unresolved: []                      # open questions for a human
+unmapped:
+  - product: kibana
+    reason: Deferred to the stack family pass.
+```
+
+Valid classifications: `published-history-found`, `native-artifacts-found`, `hybrid-page`, `declared-no-history`, `outside-cutoff`, `already-live`. `source-unresolved` is deliberately not seedable — it is the census's own conclusion for products nobody mapped, never something an operator writes by hand.
+
+## Examples
+
+```sh
+# Census with a seed; writes the inventory document
+docs-builder changelog backfill inventory \
+  --sources config/backfill/inventory-sources.yml \
+  --output .artifacts/backfill-inventory.json
+
+# Without a seed: every release-notes product is reported source-unresolved,
+# useful to see the full census surface before mapping begins
+docs-builder changelog backfill inventory --output .artifacts/backfill-inventory.json
+```

docs/cli/changelog/backfill/index.md

@@ -0,0 +1,12 @@
+## Description
+
+Backfill historical release-note bundles ([docs-eng-team#656](https://github.com/elastic/docs-eng-team/issues/656)).
+
+The changelog/bundle pipeline only contains data produced since each repository adopted the live workflows. The backfill commands make the public data look as if the current system had been in use throughout the docs-builder era: they census the products with published release notes, plan exactly which resolved bundles to create, and publish them with create-only writes.
+
+The pipeline is staged, and every stage exchanges versioned, content-addressed JSON documents (see `src/services/Elastic.Changelog/Backfill/README.md`):
+
+1. `inventory` — the census: which products and release-note sources exist and what was decided about each.
+2. Planning, materialization, apply, and verification stages follow as they are implemented.
+
+No stage writes to S3 except the guarded apply stage; everything before it is read-only and reviewable.