diff --git a/docs-developers/architecture/cross-repo-contract.md b/docs-developers/architecture/cross-repo-contract.md
index 12d75cf..fa03486 100644
--- a/docs-developers/architecture/cross-repo-contract.md
+++ b/docs-developers/architecture/cross-repo-contract.md
@@ -265,6 +265,27 @@ handler **and** `serverapi`, with tests on both. Changes to `pkg/match` or
 against them before calling it done. See
 [manager server integration](../manager/server-integration.md).
 
+## 14. Community metadata: a three-repo seam
+
+**What couples:** enriched book metadata (description, characters, recaps, "more
+in this series") originates in a fourth repo, `audiosilo-meta`, and flows through
+the server to the player.
+
+**Upstream (`audiosilo-meta`):** `metaserve` serves the community metadata
+read-only (`GET /lookup`, `/works/{id}`, `/series/{id}`) - see the
+[metadata database developer pages](../meta/overview.md).
+**Server:** `internal/meta` resolves a book's ASIN/ISBN against `metaserve` and
+composes an enrichment envelope, returned at `GET /libraries/{id}/meta` behind an
+admin off-switch and a bounded cache; the `metadata` capability on `GET /server`
+reflects whether the lookup is live.
+**Frontend:** the `BookMeta` envelope (hand-mirrored in `src/api/types.ts`) is
+fetched by `client.bookMeta` and rendered capability-gated on the book screen.
+
+**A change requires:** because the server consumes `metaserve`'s response shapes,
+a change to those shapes ripples audiosilo-meta -> the server's `internal/meta` ->
+the player (only if the server's outward `/meta` envelope changes). Keep it
+additive, same as every other seam here.
+
 ## The wire-change checklist
 
 Every change to the wire format follows the same shape (the worked example in
diff --git a/docs-developers/contributing/documentation.md b/docs-developers/contributing/documentation.md
index f42d838..f74ef98 100644
--- a/docs-developers/contributing/documentation.md
+++ b/docs-developers/contributing/documentation.md
@@ -60,6 +60,7 @@ mappings:
 | Scanner, detection, metadata | [server/scanner.md](../server/scanner.md) + `/users/getting-started/organizing-your-library` |
 | Auth, invites, shares | [server/auth-and-security.md](../server/auth-and-security.md) + `/users/admin/users-and-invites`, `/users/admin/sharing` |
 | Manager features | `/users/manager/*` + [manager developer pages](../manager/overview.md) + `manager/` screenshots |
+| Meta schemas, `metaserve` API, or intake tooling | [meta developer pages](../meta/overview.md) (data model, API, contributing) - and the [cross-repo contract](../architecture/cross-repo-contract.md) when the server's `/meta` envelope is affected |
 | Build / release / distribution | [release-pipeline.md](../architecture/release-pipeline.md) + [releasing.md](./releasing.md) |
 | A capability flag | [server/api/index.md](../server/api/index.md) + the feature's user page |
 
diff --git a/docs-developers/contributing/workspace.md b/docs-developers/contributing/workspace.md
index f3d69a8..f232a62 100644
--- a/docs-developers/contributing/workspace.md
+++ b/docs-developers/contributing/workspace.md
@@ -3,7 +3,7 @@ title: Workspace setup
 description: "How the multi-repo workspace is laid out, the toolchain you need, and how to run the whole stack locally."
 ---
 
-AudioSilo is one product across four repositories, and they are developed
+AudioSilo is one product across several repositories, and they are developed
 together from a single **workspace folder**. Get the layout right first - some of
 the build machinery assumes it.
 
@@ -22,6 +22,9 @@ it, gitignored by the meta repo:
 ├── audiosilo-server/      github.com/KodeStar/audiosilo-server    (Go)
 ├── audiosilo-frontend/    github.com/KodeStar/audiosilo-frontend  (Expo / React Native)
 ├── audiosilo-manager/     github.com/KodeStar/audiosilo-manager   (Wails desktop)
+├── audiosilo-meta/        github.com/KodeStar/audiosilo-meta      (community metadata DB)
+├── audiosilo-sidecars/    github.com/KodeStar/audiosilo-sidecars  (contributor extraction tool)
+├── audiosilo-site/        github.com/KodeStar/audiosilo-site      (marketing site)
 └── audiosilo-docs/        this documentation site (Docusaurus)
 ```
 
diff --git a/docs-developers/index.md b/docs-developers/index.md
index d75b2f9..34c9643 100644
--- a/docs-developers/index.md
+++ b/docs-developers/index.md
@@ -137,6 +137,10 @@ The server's priorities, in order - when they conflict, the earlier one wins:
 - **[Desktop manager](manager/overview.md)** - the Wails app:
   [server integration](manager/server-integration.md), the
   [Audible backup pipeline](manager/audible.md), and [transfers](manager/transfers.md).
+- **[Metadata database](meta/overview.md)** - `audiosilo-meta`, the
+  community audiobook metadata database behind meta.audiosilo.app that the server
+  and player enrich books from: the [data model](meta/data-model.md), the
+  [`metaserve` HTTP API](meta/api.md), and [contributing data](meta/contributing-data.md).
 - **[Contributing](contributing/workspace.md)** - the
   [workspace setup](contributing/workspace.md), [per-repo gates and CI](contributing/gates-and-ci.md),
   [how to land a cross-repo change](contributing/cross-repo-changes.md),
diff --git a/docs-developers/meta/api.md b/docs-developers/meta/api.md
new file mode 100644
index 0000000..9c238d9
--- /dev/null
+++ b/docs-developers/meta/api.md
@@ -0,0 +1,210 @@
+---
+title: Meta HTTP API
+description: "The metaserve read-only JSON API reference: every /api/v1 route, the Audiobookshelf provider at /abs/search, the production release webhook, CORS behavior, and how the server refreshes its artifact from GitHub Releases."
+---
+
+`metaserve` (`cmd/metaserve` over `internal/serve`) is a **read-only** JSON API
+over the compiled SQLite artifact. All data is public, so there is **no auth**;
+every `/api/v1` route (and `/abs/search`) responds with permissive CORS
+(`Access-Control-Allow-Origin: *`, `Vary: Origin`), and responses are
+gzip-compressed. Cross-origin `GET`s work from any browser; there is no
+preflight handling, so requests must stay CORS-simple (no custom headers). It
+can also serve a static site at `/` (`--site`) and hot-swaps a newer release
+artifact without a restart.
+
+Errors are a JSON envelope `{"error": "..."}` with the matching HTTP status. All
+routes are `GET`.
+
+## `/healthz`
+
+Liveness plus a cheap freshness signal. Always 200 while a snapshot is loaded:
+
+```json
+{ "status": "ok", "built_at": "2026-07-15T...", "works": 1234 }
+```
+
+## `/api/v1/stats`
+
+Catalogue totals, precomputed once per loaded snapshot:
+
+```json
+{ "works": 0, "recordings": 0, "people": 0, "series": 0,
+  "total_runtime_min": 0, "total_chapters": 0, "built_at": "..." }
+```
+
+## `/api/v1/search?q=&limit=`
+
+Full-text search over works, people, and series. `q` is **required** (400 `q is
+required` when empty). `limit` defaults to 20, clamped to `[1, 50]`. Returns
+`{"results": [...]}`, best-ranked first; each result is one of three shapes
+distinguished by `kind`:
+
+- **work**: `{kind, id, title, authors[], series, cover_url, added_at, narrators[]}`
+- **person**: `{kind, id, name}`
+- **series**: `{kind, id, name, works}` (`works` = member count)
+
+FTS input is escaped defensively (every token quoted, the final token
+prefixed with `*`), so no user input can break the underlying `MATCH`.
+
+## `/api/v1/works/latest?limit=`
+
+The newest works, for the site's landing grid. `limit` defaults to 12, clamped to
+`[1, 50]`. Returns `{"works": [workCard...]}` ordered by `added_at` descending
+(then title), with at most two works from any one series so a bulk import sharing
+one date can't fill the grid. A **workCard** is the compact shape reused across
+lists and lookups: `{id, title, authors[], series, cover_url, added_at}`.
+
+## `/api/v1/works/{id}`
+
+The full work document, or 404 `work not found`. It carries the work's
+identifiers, its `authors[]` and `series[]`, every `recordings[]` entry (with
+narrators, ASINs, ISBNs, and a `chapter_count`), and - when the loaded artifact is
+new enough and the work has them - the inline expressive layer:
+
+- `characters[]` - `{id, name, aliases?, role?, reveal:{chapter}, description?, xref?}`
+- `recaps[]` - `{through:{chapter}, scope?, text}`
+- `recap_summary` - `{in_short?, ending?}`
+
+All three are `omitempty` and gated on the artifact `schema_version`
+(characters/recaps at 2, recap summary at 3), so an older artifact simply omits
+them (see [the data model](data-model.md#the-compiled-artifact-and-schema-versioning)).
+
+## `/api/v1/works/{id}/recordings/{rid}/chapters`
+
+The chapter list for one recording of a work: `{"chapters": [{title, start_ms,
+length_ms}]}`, ordered by chapter index. An unknown work/recording yields an empty
+list, not a 404.
+
+## `/api/v1/people/{id}`
+
+A person plus their works, or 404 `person not found`:
+
+```json
+{ "id": "...", "name": "...", "sort_name": "...",
+  "authored": [workCard...],
+  "narrated": [{ "work": workCard, "recording_id": "..." }] }
+```
+
+## `/api/v1/series/{id}`
+
+A series with its ordered member works, or 404 `series not found`:
+
+```json
+{ "id": "...", "name": "...", "authors": [personRef...],
+  "works": [{ "position": "2.5", "work": workCard }] }
+```
+
+`works` is sorted by the numeric start of each `position` string (so `"1-3.5"`
+sorts by 1).
+
+## `/api/v1/lookup?asin=|isbn=`
+
+Resolve one identifier to a work. At least one of `asin` / `isbn` is **required**
+(400 `asin or isbn is required`); a miss is 404 `not found`. On a hit:
+
+```json
+{ "work": workCard, "recording_id": "..." }
+```
+
+ASIN resolves against recording ASINs; ISBN resolves against recording ISBNs and
+then falls back to a work's print ISBN (pointing at its first recording). This is
+the entry point the AudioSilo server's `internal/meta` uses to enrich a book by
+its `asin`/`isbn`.
+
+## Coverage endpoints
+
+These back the site's contribute page and stay small at any catalogue size.
+
+- **`/api/v1/coverage`** - the top-line totals only:
+  `{"totals": {works, with_characters?, with_recaps?, with_recap_summary?}}`. The
+  three sidecar counts are **omitted** (not zero) when the loaded artifact's
+  `schema_version` predates that dimension's table, so an unknowable count is
+  never reported as a misleading 0.
+- **`/api/v1/coverage/works?filter=&q=&limit=&offset=`** - the paginated,
+  searchable per-work browser. `filter` selects the dimension - `missing` (missing
+  any dimension) or `has_characters` / `has_recaps` / `has_recap_summary` - and an
+  unknown filter is 400 `unknown filter`. `q` matches title/author; `limit`
+  defaults to 25, clamped to `[1, 100]`; `offset` is a non-negative row offset. The
+  response carries a per-filter `available` flag that is false when the dimension
+  is not evaluable at the artifact's schema version.
+- **`/api/v1/coverage/series-gaps?q=&limit=&offset=`** - the paginated,
+  name-searchable list of series with interior position gaps (integer positions
+  absent between the lowest and highest present integer). No schema-version
+  dependency, so it is always available.
+
+## `GET /abs/search` (Audiobookshelf provider)
+
+`metaserve` doubles as an **Audiobookshelf custom metadata provider**. An ABS
+admin configures the base URL `https://meta.audiosilo.app/abs` (no auth; ABS
+v2.8.0+), and ABS appends `/search`. ABS sends `?mediaType=book&query=<title>`
+with optional `&author=` and `&isbn=`, and **never** an ASIN.
+
+- `query` is **required** (400 `query is required`); the endpoint **never 404s** -
+  a no-match is a 200 with an empty array.
+- Resolution: if an ISBN is present, an exact identifier lookup runs first (the
+  hyphens ABS sends are stripped to the bare stored form); otherwise, or on an
+  ISBN miss, an FTS work search runs, with works whose authors loosely match
+  `author` boosted ahead of the rest (a wrong author boosts rather than filters,
+  so it never empties results).
+- The response is `{"matches": [...]}`, **one entry per recording** (a recording is
+  what ABS matches a local audiobook against), capped at 10. Each match carries
+  `title` (the only required field) plus, when present, `subtitle`, `author`,
+  `narrator`, `publisher`, `publishedYear` (a string), `description`, `cover`,
+  `isbn`, `asin`, `series[]` (`{series, sequence}`), `language`, and `duration`
+  **in minutes**. `genres` and `tags` are **deliberately never returned** - the
+  data model does not carry publisher genres/tags.
+
+## Production release webhook (optional)
+
+When `METASERVE_WEBHOOK_SECRET` (at least 32 bytes) is set **and** `--poll` is
+enabled, metaserve registers `POST /hooks/github/release`, authenticated by the
+standard `X-Hub-Signature-256: sha256=...` HMAC header. `release.yml` calls it
+only after every release asset has finished uploading. The request body is only a
+**trigger**: metaserve re-queries GitHub and goes through the same verified
+refresh path as polling, never trusting or installing data from the request body.
+The endpoint is not registered when the secret is absent, and a missed delivery
+is non-fatal - the fallback poller still discovers the release.
+
+## Flags
+
+`cmd/metaserve` is flag wiring only; every knob maps onto `internal/serve.Config`:
+
+| Flag / env | Default | Purpose |
+|---|---|---|
+| `--addr` | `:8080` | listen address |
+| `--db` | (none) | a local `meta.sqlite` artifact to serve (dev) |
+| `--site` | (none) | a static site directory to serve at `/` |
+| `--poll` | `false` | fetch and hot-swap the newest data release from GitHub Releases |
+| `--repo` | `KodeStar/audiosilo-meta` | GitHub `owner/name` to poll |
+| `--interval` | `1h` | fallback poll interval |
+| `--cache` | `./cache` | directory for downloaded artifacts |
+| `GITHUB_TOKEN` (env) | (none) | raises the GitHub API rate limit |
+| `METASERVE_WEBHOOK_SECRET` (env) | (none) | enables the signed release webhook (requires `--poll`) |
+
+With `--poll` and no `--db`, metaserve fetches the newest data release on boot so
+it never starts empty. With both, the baked `--db` serves immediately and the
+poller still runs one refresh at startup.
+
+## Serving and refresh
+
+The current artifact lives behind an atomic pointer (a `snapshot`); readers load
+the pointer once per request. With `--poll`, a background loop plus the optional
+webhook keep it current:
+
+- It selects the newest **data** release (the selection rule is on
+  [the overview](overview.md#release-artifacts)) and fetches conditionally
+  (`If-None-Match` / 304).
+- On a new release it first tries a `--patch-from` binary delta against the
+  currently-loaded artifact (zstd, `--long=31` window), verifying the reconstructed
+  file byte-for-byte against `meta.sqlite.sha256` before installing it; it falls
+  back unconditionally to a full `meta.sqlite.gz` download (verified against
+  `meta.sqlite.gz.sha256`) whenever a patch is unavailable or fails. The first
+  refresh after boot is always full.
+- Either way it hot-swaps the pointer; in-flight requests finish on the old
+  handle (closed after a grace delay). A rejected patch never swaps, and a poll
+  failure only logs and retries - it never crashes the process.
+
+The startup refresh means a recreated production container catches up to the
+newest release within seconds instead of serving build-time data for a full
+`--interval`. The release asset contract these steps rely on is described on
+[the overview](overview.md#release-artifacts).
diff --git a/docs-developers/meta/contributing-data.md b/docs-developers/meta/contributing-data.md
new file mode 100644
index 0000000..3f61c01
--- /dev/null
+++ b/docs-developers/meta/contributing-data.md
@@ -0,0 +1,170 @@
+---
+title: Contributing data to Meta
+description: "How metadata enters audiosilo-meta: the six GitHub issue forms and the intake automation that turns them into validated bot pull requests, the advisory ai-verify layer, the OpenAudible/Libation importers, metascan, the in-browser site tools, and the authoring rules."
+---
+
+## The rules that govern everything
+
+Two non-negotiable rules apply to every contribution, whichever path it takes.
+They are enforced by tooling where possible and by review everywhere else (the
+full policy is in the repo's `LICENSING.md` and `GOVERNANCE.md`):
+
+- **Facts only, never fabricated.** Contributed data must be real and verifiable.
+  If a fact can't be verified, the (optional) field is **omitted rather than
+  guessed** - no publisher blurbs, no invented ASINs, no cover files (covers are
+  URLs). Every record carries a `sources[]` provenance entry.
+- **Own words, never copied.** Descriptions and the CC BY-SA characters/recaps
+  are community-authored and length-capped for the reference-guide tier; verbatim
+  or near-verbatim phrasing from a source is a separate publish-pipeline failure
+  (see [the extraction docs](#authoring-the-expressive-layer)).
+
+Because the GitHub repository is the database, **all writes go through GitHub** -
+there are no server-side accounts. A contribution is either a direct pull request
+editing `data/**`, or an issue form that the intake automation turns into one.
+
+## The six issue forms
+
+`.github/ISSUE_TEMPLATE/*.yml` are structured forms (machine-parseable field ids)
+so a non-programmer can contribute without touching JSON. Each carries a
+`data:<kind>` routing label that the intake workflow branches on:
+
+| Form | Routing label | For |
+|---|---|---|
+| Add a work (book) and its first recording | `data:add-work` | a new book plus its first narration |
+| Add a recording (narration) | `data:add-recording` | another narration of a work already in the database |
+| Correct data | `data:correction` | a single-field fix to an existing record |
+| Add characters (the cast) | `data:characters` | the per-work characters sidecar (CC BY-SA) |
+| Add recaps (story so far) | `data:recaps` | the per-work recaps sidecar (CC BY-SA) |
+| Import a library export | `data:import` | an OpenAudible / Libation / Audiobookshelf / metascan export to bulk-import |
+
+## Intake automation: issue form to bot pull request
+
+`.github/workflows/intake.yml` converts a submitted form into a validated bot
+pull request. On a data-labelled issue it runs `metaissue`, which parses the
+rendered form body into canonical records (or a single-field correction, or a
+placed sidecar), **deduplicates against the existing catalogue**, and emits a
+machine-readable verdict the workflow branches on:
+
+| Verdict | Meaning | Workflow action |
+|---|---|---|
+| `ok` | valid new/changed records produced | opens a PR on branch `intake/issue-<n>` |
+| `duplicate` | everything already exists (requires at least one skip) | labels + comments, no PR |
+| `needs-human` | ambiguous - e.g. an import that produced and deduped nothing | labels for maintainer attention |
+| `invalid` | the submission fails schema/validation | labels + comments with the errors |
+
+Two behaviors are worth knowing:
+
+- **Envelope sniffing.** For an import, `metaissue` sniffs a self-identifying
+  `audiosilo-books` envelope and routes it to that importer regardless of the
+  form's export-type dropdown - the file is trusted over the form.
+- **The `labeled` trigger is load-bearing.** The GitHub API silently drops labels
+  on issues opened by non-collaborators (the sibling `audiosilo-sidecars`
+  contributor tool creates intake issues over the API), so such an issue arrives
+  label-less and the `opened`/`edited` runs skip it. When a maintainer later
+  applies the routing label, the `labeled` trigger admits it. The job gate
+  excludes the workflow's own outcome labels (`data:invalid` / `data:needs-human`
+  / `data:duplicate`) so outcome-labeling can't re-fire intake.
+
+:::note Intake runs on `issues`, not fork code
+`intake.yml` triggers on the `issues` event, so there is **no fork code
+execution**. The only untrusted input is the issue body and any attachment: it is
+written to a file via an environment variable (never interpolated into a shell
+command), parsed by `metaissue`, and never executed. Attachments are fetched
+HTTPS-only from GitHub's user-attachment hosts with a size cap. The security
+posture is deliberate - see [gates and CI](../contributing/gates-and-ci.md) for
+the workspace-wide CI rules.
+:::
+
+## Two validation layers on a pull request
+
+Every pull request touching `data/**` is checked twice:
+
+- **`check.yml` (mechanical, blocking).** Runs `go build`/`vet`/`test`,
+  `metacheck` (schema, id/shard agreement, referential integrity, uniqueness,
+  chapter/series rules), and `metafmt --check` (canonical JSON). A red pull
+  request never merges. It uses the plain `pull_request` trigger, so fork pull
+  requests run with a read-only token and no secrets.
+- **`ai-verify.yml` (advisory, never blocking).** An AI judgement layer on top of
+  the mechanical check: it posts a `PASS` / `FLAG` comment and label but **never
+  blocks a merge**. It also uses the plain `pull_request` trigger by design (not
+  `pull_request_target`, which is forbidden here as the "pwn request" pattern), so
+  a fork pull request gets a neutral skip notice until a maintainer pushes its
+  branch to the repo or re-runs it - fork secrets are never reached. The diff is
+  passed to the model as untrusted data and never executed.
+
+## Bulk importers: metaimport
+
+`metaimport` ingests an external library export into `data/` as reviewable
+records, for contributors who already have a library manager's export:
+
+```sh
+go run ./cmd/metaimport openaudible <books.json>  [--dry-run] [--date YYYY-MM-DD]
+go run ./cmd/metaimport libation    <export.json> [--dry-run] [--date YYYY-MM-DD]
+```
+
+It imports **factual fields only** (dropping publisher copy, genres, ratings, and
+personal state), maps one export entry to a work + recording (+ people + series),
+and **deduplicates by ASIN** against the catalogue. `--dry-run` prints the plan
+without writing; a real run writes the files, then validates the whole tree and
+exits non-zero if that fails. The identity rules are careful: a person slug is the
+identity (name variants merge, no numbered duplicates), a work is (title slug +
+author set) with per-volume disambiguation so distinct series volumes never merge,
+a trailing `(Unabridged)`/`(Abridged)` marker is stripped before identity (and
+seeds the recording's tri-state `abridged` when the source didn't state it), and a
+same-work/same-narrator entry whose only new fact is another ASIN **merges that
+ASIN into the existing recording** (guarded by runtime and abridged checks) rather
+than minting a sibling.
+
+## Scanning local files: metascan
+
+`metascan` is the low-friction path when you have only audio files - no export.
+It walks a folder locally and **sends nothing anywhere**, emitting an import JSON
+the site's `/import` page accepts:
+
+```sh
+go run ./cmd/metascan /path/to/audiobooks -o scan.json
+```
+
+Per book it gathers embedded tags (via `dhowden/tag`), the folder structure
+treated as a first-class source (`Author/Book`, `Author/Series/Book`, and name
+patterns like `01 - Title` or `Jack Reacher 03 - Title`), an ASIN hunted in tag
+atoms and file/folder names, and - if `ffprobe` is on `PATH` - runtime and chapter
+counts. Every field records where it came from (`tag` / `path` / `filename`) in
+the book's `sources`, and unknown fields are omitted rather than guessed. Without
+`ffprobe` the scan still works; embedded ASIN/series extraction is just more
+limited (pass `-ffprobe ""` to skip it entirely).
+
+## The in-browser site tools
+
+meta.audiosilo.app hosts two client-side helpers so a contributor never has to run
+Go tooling. The Developer Docs only note them; the end-user walkthrough is the
+User Guide's [community metadata site page](/users/community/meta-site):
+
+- **[`/import`](https://meta.audiosilo.app/import)** - parses an OpenAudible,
+  Libation, or Audiobookshelf export, or a `metascan` folder scan, entirely in the
+  browser and **diffs it against the live catalogue**, so a contributor sees
+  exactly what is new before submitting.
+- **[`/build`](https://meta.audiosilo.app/build)** - a guided builder that walks a
+  contributor through writing the characters or recaps sidecar for a work already
+  in the catalogue.
+
+## Authoring the expressive layer
+
+The CC BY-SA characters/recaps layer has its own documented process and tooling,
+all at the root of the `audiosilo-meta` repository:
+
+- `AUTHORING.md` - the reusable authoring process for characters/recaps:
+  positions, the spoiler model, the copyright length caps, and the submission
+  checklist.
+- `EXTRACTION.md` - the epub source-to-sidecar pipeline (rolling fact pass,
+  notes-only synthesis, adversarial spoiler audit), supported by
+  `metaextract split` + `ngram`.
+- `EXTRACTION-AUDIO.md` - the audio-only variant (chapter-isolated ASR +
+  proper-noun verification), the process the `audiosilo-sidecars` tool automates.
+- `GOVERNANCE.md` - the merge policy and contributor trust tiers
+  (schema/tooling/`.github` changes always need maintainer review via
+  CODEOWNERS).
+
+Source material and transcripts **never enter the repository** - only the derived
+CC BY-SA sidecars are committed, so the near-verbatim `ngram` check is run locally
+against the source you hold, by design.
diff --git a/docs-developers/meta/data-model.md b/docs-developers/meta/data-model.md
new file mode 100644
index 0000000..2fef18e
--- /dev/null
+++ b/docs-developers/meta/data-model.md
@@ -0,0 +1,166 @@
+---
+title: Meta data model
+description: "The audiosilo-meta entities and their on-disk layout: path-is-identity slugs and sharding, works/recordings/people/series, the characters and recaps sidecars, the position model, provenance, and the JSON Schema contract."
+---
+
+## Path is identity
+
+Every entity is addressed by a **slug** matching `^[a-z0-9]+(-[a-z0-9]+)*$`
+(`common.schema.json` `$defs/slug`, max 100 chars), and its file lives in a
+**shard directory** named for the first two characters of the slug. Sharding
+keeps any one directory small as the catalogue grows. The slug is the identity -
+there is no numeric id anywhere in the data, and the file path encodes exactly
+where a record lives.
+
+The JSON Schemas in `schema/*.schema.json` (JSON Schema draft 2020-12, every
+object `additionalProperties: false`) are the **authoritative, public contract**.
+They are embedded into the tooling via `schema.go`, so a schema edit is a code
+change that ships with tests. The field lists below are drawn straight from those
+schemas.
+
+## The factual core (CC0)
+
+### work - `data/works/<shard>/<slug>/work.json`
+
+The abstract book, independent of any particular narration.
+
+| Field | Required | Notes |
+|---|---|---|
+| `id` | yes | slug |
+| `title` | yes | |
+| `subtitle` | no | |
+| `authors` | yes | array of person slugs, at least one |
+| `language` | yes | BCP-47-ish (`^[a-z]{2,3}(-[a-z0-9]{2,8})*$`) |
+| `first_published` | no | `YYYY` or `YYYY-MM-DD` |
+| `description` | no | community-written, never a publisher blurb |
+| `xref` | no | `wikidata` (`Q\d+`), `openlibrary` (`OL\d+W`), `goodreads`, print `isbn[]` |
+| `license` | yes | `CC0-1.0` |
+| `sources` | yes | provenance (below) |
+
+### recording - `data/works/<shard>/<slug>/recordings/<rec-slug>.json`
+
+A specific narration/production of a work. **One work, many recordings** - the
+canonical example is *Harry Potter and the Philosopher's Stone*, one work with a
+Stephen Fry recording and a Jim Dale recording, each carrying its own ASINs. The
+shard is the **parent work's** slug shard.
+
+| Field | Required | Notes |
+|---|---|---|
+| `id` | yes | slug |
+| `work` | yes | parent work slug |
+| `narrators` | yes | array of person slugs, at least one |
+| `abridged` | no | **tri-state**: absence means *unknown*, so importers omit it rather than guess |
+| `language` | yes | |
+| `runtime_min` | no | integer minutes, > 0 |
+| `release_date` | no | `YYYY`, `YYYY-MM`, or `YYYY-MM-DD` |
+| `publisher` | no | |
+| `asin` | no | array of `{region, asin}`; `region` is one of 11 storefronts (`us`, `uk`, `ca`, `au`, `de`, `fr`, `es`, `it`, `jp`, `in`, `br`); `asin` is `[A-Z0-9]{10}` |
+| `isbn` | no | array of bare 10/13-digit ISBNs |
+| `cover_url` | no | must be an `https://` URL |
+| `chapters` | no | array of `{title, start_ms, length_ms}` |
+| `license` | yes | `CC0-1.0` |
+| `sources` | yes | |
+
+### person - `data/people/<shard>/<slug>.json`
+
+One human, shared across roles: authors on works and narrators on recordings are
+the same entity type, and a person can be both. Fields: `id`, `name`, optional
+`sort_name`, optional `description`, optional `xref` (`wikidata`, `openlibrary`
+`OL\d+A`, `audible` ASIN), `license` (`CC0-1.0`), `sources`.
+
+### series - `data/series/<shard>/<slug>.json`
+
+A named, ordered set of works. Each entry is `{work, position}`, where
+**`position` is a string** so it can express decimals and omnibus ranges:
+`"1"`, `"2.5"`, or a range spanning several entries `"1-3.5"` (pattern
+`^\d+(\.\d+)?(-\d+(\.\d+)?)?$`). `metacheck` enforces that **no two works share a
+position** within a series. Fields: `id`, `name`, optional `authors`, `works`,
+optional `xref` (`wikidata`, `goodreads`), `license` (`CC0-1.0`), `sources`.
+
+## The expressive layer (CC BY-SA)
+
+Two **per-work sidecars** carry the community-authored, spoiler-tagged content.
+They are structurally separated from the core: their `license` field accepts only
+`CC-BY-SA-3.0` (`$defs/license_content`). Authoring them is documented in the
+repo's `AUTHORING.md` (see [contributing data](contributing-data.md)).
+
+### characters - `data/works/<shard>/<slug>/characters.json`
+
+An array of character entries under a `work` slug. Each character has:
+
+- `id` - unique **within the file**, not globally (two works may each have a
+  `bilbo-baggins`);
+- `name`, optional `aliases[]`, optional `role` (`protagonist` / `antagonist` /
+  `supporting` / `minor`);
+- `reveal` - a [position](#the-position-model), the spoiler gate: a consumer only
+  shows the entry once the listener has passed it (the Kindle X-Ray model);
+- optional `description` - own-words, length-capped at 1500 chars (a card
+  without one simply has nothing to reveal);
+- optional `xref` (`wikidata`, `goodreads`) - a shared `wikidata` QID links a
+  recurring character across a series' per-work files.
+
+Recurring characters are **re-described per book**, so what a reader sees stays
+bounded by which book they are currently in.
+
+### recaps - `data/works/<shard>/<slug>/recaps.json`
+
+Position-keyed "story so far" summaries under a `work` slug, plus two optional
+whole-book summaries. Each recap entry has:
+
+- `through` - a [position](#the-position-model): the recap is safe to show once
+  the listener has finished that chapter. **No two recaps in a file share a
+  `through` chapter.**
+- optional `scope` (`book` / `series`) - a `chapter: 0` + `series` entry is the
+  "previously, in earlier books" recap;
+- `text` - own-words, length-capped at 3000 chars.
+
+The file also carries two optional whole-book fields for a reader who has
+finished the book: `in_short` (the whole arc in one paragraph, ending included,
+cap 1500) and `ending` (how the book closes, stated plainly, cap 2000 -
+deliberately tighter than a chaptered recap entry, a crisp sequel-handoff).
+
+## The position model
+
+Spoiler positions use one shape everywhere (`common.schema.json` `$defs/position`):
+
+```json
+{ "chapter": 3 }
+```
+
+`chapter` is a non-negative integer: the **logical, edition-independent** work
+chapter (1-based), where `0` means front matter or knowledge carried in from
+earlier books in a series. A consumer maps its own recording-chapter timeline
+onto these ordinals; text-to-audio alignment is a consumer concern, out of schema
+scope. The object shape is deliberately extensible - a later `paragraph` or
+`offset_ms` can be added without a breaking change.
+
+## Provenance on every entity
+
+Every record carries a `sources[]` array (`$defs/sources`, at least one entry).
+Each source is `{type, ref?, imported_at?}`, where `type` is one of a fixed enum
+(`user`, `openaudible-import`, `libation-import`, `audiosilo-books-import`,
+`audible-lookup`, `openlibrary`, `wikidata`, `inventaire`, `community`) and
+`imported_at` is `YYYY-MM-DD`. Because every fact records where it came from, a
+whole source can be audited or retracted.
+
+## The compiled artifact and schema versioning
+
+`metabuild` compiles the tree into a deterministic SQLite artifact (`internal/build`),
+inserting rows in sorted id order so identical data always produces an identical
+file. It stamps a `schema_version` into the artifact's `meta` table, and the
+expressive layer was added in later versions:
+
+| artifact `schema_version` | Adds |
+|---|---|
+| 1 | the factual core (works, recordings, people, series, FTS5 search index) |
+| 2 | the `characters`, `character_aliases`, and `recaps` tables |
+| 3 | the per-work `recap_summaries` table (the `in_short` / `ending` fields) |
+
+`metaserve` returns characters and recaps inline on `GET /works/{id}`
+(`characters` / `recaps` / `recap_summary`, all `omitempty`). The serve queries
+**degrade gracefully** when a newer binary briefly serves an older release: the
+characters/recaps queries no-op below `schema_version` 2 and the recap summary
+below 3, so a missing table reads as "no data", never a 500. The same versioning
+drives the [coverage endpoints](api.md#coverage-endpoints), which omit a
+dimension's count rather than report it as a misleading zero when the artifact
+predates its table.
diff --git a/docs-developers/meta/overview.md b/docs-developers/meta/overview.md
new file mode 100644
index 0000000..317d3cf
--- /dev/null
+++ b/docs-developers/meta/overview.md
@@ -0,0 +1,159 @@
+---
+title: Meta overview
+description: "Architecture of audiosilo-meta: the GitHub repo that IS the community audiobook metadata database, validated by Go tooling, compiled to a SQLite artifact, and served read-only by metaserve to the AudioSilo player and Audiobookshelf."
+---
+
+## What audiosilo-meta is
+
+`audiosilo-meta` is the community audiobook metadata database behind
+[meta.audiosilo.app](https://meta.audiosilo.app). Its defining idea: **the GitHub
+repository is the database.** There is no application database and no server-side
+accounts - metadata lives as plain JSON files under `data/`, one file per entity,
+edited by pull request or issue form, validated by Go tooling in CI, and compiled
+into a single SQLite artifact that consumers download and serve read-only.
+
+The reason it exists is a gap no other open database fills: none is both
+**audiobook-specific** and **community-editable**. Open Library, Wikidata, and
+BookBrainz are open but carry essentially no narrator, recording, or chapter
+structure; Audnexus is a read-only cache over Audible's own catalogue. So
+narrators, specific recordings, and chapters are **first-class data** here - the
+fields every other database lacks.
+
+Module path: `github.com/kodestar/audiosilo-meta`. The code is **AGPL-3.0**; the
+data is split across two licence layers (below).
+
+```
+  contributor
+      │  pull request  /  issue form
+      ▼
+  data/*.json  ──►  CI validation  ──►  meta.sqlite  ──►  metaserve (API + site)
+  (the database)    (metacheck +        (release          read-only, hot-swaps
+                     metafmt, Go)        artifact)         new releases
+```
+
+## The two licence layers
+
+Data lives in two layers, and the boundary is **enforced structurally by the JSON
+Schema**, not by convention (see `LICENSING.md` in the repo): the `license` field
+of a core record accepts only `CC0-1.0`, and a sidecar's only `CC-BY-SA-3.0`
+(`common.schema.json` `$defs/license` vs `$defs/license_content`).
+
+| Layer | Entities | Licence |
+|---|---|---|
+| **Factual core** | works, recordings, people, series | **CC0-1.0** public-domain dedication |
+| **Expressive layer** | per-work characters + recaps sidecars | **CC BY-SA 3.0** |
+
+Publisher blurbs and cover art are referenced, never copied: descriptions are
+community-written and covers are URLs. Every entity also carries a `sources[]`
+provenance array so any source can be audited or retracted wholesale. Full detail
+is in [the data model](data-model.md).
+
+## Package layout
+
+The tooling follows the same discipline as the rest of the workspace: thin CLIs
+under `cmd/`, all logic in reusable packages, and a clean split between the
+public `pkg/*` (consumed by the sibling `audiosilo-sidecars` module as ordinary
+dependencies) and the private `internal/*`.
+
+```
+data/          the database: works/, people/, series/ (sharded JSON) + per-work sidecars
+schema/        JSON Schemas (one per entity) - the public contract, embedded via schema.go
+cmd/           thin CLIs: metacheck, metafmt, metabuild, metaserve, metascan,
+               metaimport, metaissue, metaextract (flag wiring only)
+pkg/model      PUBLIC entity structs, slug/shard rules, location parsing
+pkg/canonical  PUBLIC canonical JSON (sorted keys, 2-space indent, trailing LF)
+pkg/check      PUBLIC schema validation + integrity/uniqueness/chapter/series rules
+pkg/extract    PUBLIC epub split + the word-shingle near-verbatim check
+pkg/scan       PUBLIC local folder scanner (tags + path/filename heuristics + ffprobe)
+internal/importer   OpenAudible / Libation export -> canonical records (ASIN dedup)
+internal/issueform  issue-form body -> canonical records + an ok/duplicate/needs-human/invalid verdict
+internal/build      the deterministic SQLite builder (FTS5, ASIN/ISBN indexes, added_at)
+internal/serve      the read-only HTTP API + ABS provider + GitHub-release poller/hot-swap
+Dockerfile     image: site build + metaserve + baked data
+.github/       issue forms + CI workflows (check, release, image, intake, ai-verify)
+```
+
+Dependency direction mirrors the server's "transport is logic-free" rule:
+`cmd/metaserve` is flag wiring only, and all business logic lives in
+`internal/serve`.
+
+## The CLI tool set at a glance
+
+Each command is a thin `cmd/*` wrapper over a package; run any of them with `go
+run ./cmd/<name>`.
+
+| Command | What it does |
+|---|---|
+| `metacheck` | Validates the whole `data/` tree - schema, id/shard agreement, referential integrity, uniqueness, chapter ordering, series positions. Prints one line per problem and exits 1 if any are found. |
+| `metafmt` | Enforces canonical JSON for `data/**/*.json` (sorted keys, 2-space indent, single trailing LF). `--check` lists non-canonical files and exits 1; `--write` rewrites them. |
+| `metabuild` | Compiles `data/` into the SQLite artifact (`-o meta.sqlite`). Runs the full validation first and refuses to build invalid data; `--added` dates each work from a git-history-derived list. |
+| `metaserve` | Serves the compiled artifact read-only over HTTP (and optionally the static site at `/`), hot-swapping newer GitHub releases. See [the HTTP API](api.md). |
+| `metascan` | Scans a local audiobook folder into an import JSON - see [contributing data](contributing-data.md#scanning-local-files-metascan). |
+| `metaimport` | Ingests an OpenAudible/Libation library export into `data/` - see [contributing data](contributing-data.md#bulk-importers-metaimport). |
+| `metaissue` | Issue-form body to canonical records + verdict, for the intake bot - see [contributing data](contributing-data.md#intake-automation-issue-form-to-bot-pull-request). |
+| `metaextract` | Supports the source-to-sidecar extraction pipeline: `split` (epub -> chapter text + manifest) and `ngram` (near-verbatim overlap check against the source text). |
+
+## Build, validate, and serve locally
+
+Requires **Go 1.25+** (pure Go, no cgo, no external services). The full gate,
+matching CI (`.github/workflows/check.yml`):
+
+```sh
+go build ./... && go vet ./... && go test -race ./... && golangci-lint run
+go run ./cmd/metacheck            # validate the data tree
+go run ./cmd/metafmt --check      # canonical formatting (--write to fix)
+```
+
+Build the artifact and serve it:
+
+```sh
+go run ./cmd/metabuild -o meta.sqlite
+go run ./cmd/metaserve --db meta.sqlite --addr :8080
+```
+
+## Release artifacts
+
+On merge to `main`, `.github/workflows/release.yml` publishes a **dated data
+release** tagged `data-vYYYY.MM.DD-<shortsha>` when data or schema changes land.
+The asset contract:
+
+- `meta.sqlite.gz` + `meta.sqlite.gz.sha256` - the universal anchor every
+  consumer verifies against.
+- `meta.sqlite.sha256` - the raw-file digest, used to verify a patched artifact.
+- `meta.sqlite.patch.from-<PREV_TAG>.zst` - a **best-effort** zstd `--patch-from`
+  binary delta against the previous data release (`--long=31`; a stock zstd CLI
+  consumer must also pass `--long=31` at decompression time).
+
+The repo also cuts code/image `v*` releases with no data assets, so consumers
+select the newest **data** release by asset presence: the non-draft,
+non-prerelease release carrying `meta.sqlite.gz` with the maximum `published_at`
+(GitHub's release list order is not publish-chronological, and its "latest" can be
+either kind). The [`metaserve` refresh loop](api.md#serving-and-refresh) applies
+the same rule.
+
+## How it connects to the rest of AudioSilo
+
+audiosilo-meta is the **upstream** of a three-repo metadata seam. `metaserve`
+serves the community data; the AudioSilo server composes a book's enrichment from
+it; the player renders that enrichment.
+
+```mermaid
+flowchart LR
+    meta["audiosilo-meta<br/>metaserve (this repo)"]
+    srv["audiosilo-server<br/>internal/meta"]
+    player["audiosilo-frontend<br/>book screen"]
+    meta -->|"GET /lookup, /works/{id}, /series/{id}"| srv
+    srv -->|"GET /libraries/{id}/meta<br/>(composed envelope)"| player
+```
+
+The seam of record is
+[the cross-repo contract's community-metadata section](../architecture/cross-repo-contract.md#14-community-metadata-a-three-repo-seam) -
+what couples, which files sit on each side, and what a `metaserve` response-shape
+change ripples into. The consumer sides are documented on
+[server configuration](../server/configuration.md), the server's
+[HTTP API reference](../server/api/reference.md), and the frontend's
+[state and data](../frontend/state-and-data.md).
+
+`metaserve` additionally doubles as an **Audiobookshelf custom metadata
+provider** (`GET /abs/search`) for that competitor's users - see
+[the HTTP API](api.md#get-abssearch-audiobookshelf-provider).
diff --git a/docs-users/community/meta-site.md b/docs-users/community/meta-site.md
new file mode 100644
index 0000000..5009828
--- /dev/null
+++ b/docs-users/community/meta-site.md
@@ -0,0 +1,118 @@
+---
+title: The community metadata site
+description: Browse, check and contribute to AudioSilo Meta - the free, community-built audiobook database that powers the "About this book" extras in the player.
+---
+
+**AudioSilo Meta** at [meta.audiosilo.app](https://meta.audiosilo.app) is an
+open, community-built database of audiobooks: the books themselves, each
+narration (recording), the narrators and authors behind them, and the series
+they belong to. It is the catalogue that powers the
+[About this book](../listening/browsing.md#about-this-book) extras in the
+player - and it is free for anyone to browse, with no account and no sign-up.
+
+The database treats the details other catalogues skip as first-class
+information: every recording lists its **narrators**, and one book can hold many
+recordings (for example a Stephen Fry reading and a Jim Dale reading), each with
+its own runtime, publisher, chapters and store identifiers.
+
+![The AudioSilo Meta home page](/img/screenshots/meta/home.png)
+
+## Browsing the database
+
+The home page has a single search box. It searches **books, people (authors and
+narrators), and series** at once, and if you paste an ASIN or ISBN it jumps
+straight to the exact recording. Below the search box you'll find the latest
+additions and a running count of what the community has catalogued so far.
+
+![Search results for a series query](/img/screenshots/meta/search.png)
+
+- A **book page** lists every catalogued narration of that title. Each recording
+  card shows its narrators, runtime, release date, publisher, any region-specific
+  ASINs and ISBNs (tap one to copy it), and an expandable chapter list. If the
+  book belongs to a series, the page links to the series and offers previous and
+  next buttons plus a "more in this series" row of covers.
+- A **series page** lists the member books in reading order, with their positions
+  (including half-numbers like 2.5).
+- A **person page** collects everything a narrator or author is credited on.
+
+![A work page with its recordings](/img/screenshots/meta/work.png)
+
+![A series page in reading order](/img/screenshots/meta/series.png)
+
+## Characters and story-so-far recaps
+
+On a book page, two extra tabs appear once the community has written them - the
+same material the player shows under [About this book](../listening/browsing.md#about-this-book):
+
+- **Characters** - a card for each person in the book, written by readers in
+  their own words. A card's name, role and aliases are always visible, along with
+  the chapter where the character first appears; the description stays hidden
+  behind the card until you open it, so you choose when to read on.
+- **Story so far** - short "what's happened up to here" recaps, each labelled
+  with the chapter it is safe to read after. Every recap stays closed until you
+  open it, so you can catch up to exactly where you've listened without spoiling
+  what's ahead. Some books also carry a whole-book summary and an ending recap
+  for readers who have finished; those are marked as full spoilers.
+
+![Characters and story-so-far recaps](/img/screenshots/meta/characters.png)
+
+## Check your own library
+
+The [import page](https://meta.audiosilo.app/import) tells you which of your
+audiobooks are already in the database and which are new. Drop in a library
+export and it sorts your books into "In the database", "New - you can contribute
+these", and "Cannot auto-match".
+
+It accepts an export from **OpenAudible**, **Libation**, or **Audiobookshelf**,
+or a scan of a plain folder of audiobooks. The page explains how to produce each
+one.
+
+Your file is read **entirely in your browser** and never uploaded. Only the book
+identifiers (ASINs and ISBNs) are sent to the database to look for a match, plus
+the author names of books that didn't match, so it can tell a brand-new book from
+a new narration of a book that's already listed. Personal fields - purchase
+history, ratings, file paths - never leave your device.
+
+![Checking a library export against the database](/img/screenshots/meta/import.png)
+
+## Contributing
+
+Everything in AudioSilo Meta is added by its users, and you don't have to be a
+programmer to help. Contributions are made through **guided forms** that open a
+prefilled entry on the project's GitHub page; a bot then checks the submission
+and files it. You need a free GitHub account to submit one.
+
+There are forms to add a book or a narration, to write a book's characters or its
+story-so-far recaps, to send in a correction, and to submit a whole library at
+once. The import page above turns each new book it finds into a one-click "add
+this book" link.
+
+The [contribute page](https://meta.audiosilo.app/contribute) shows where help is
+most needed: which books still need characters or recaps, and which series are
+missing volumes. From there, a **guided builder** walks you through writing the
+characters and recaps for a book you know well, checking the spoiler positions
+and length limits as you go, and hands you a finished submission at the end.
+
+![The contribute coverage browser](/img/screenshots/meta/contribute.png)
+
+## How the data stays honest
+
+Two rules keep the catalogue clean, and they're worth knowing before you
+contribute:
+
+- **Facts only, in your own words.** The catalogue records verifiable facts and
+  community-written descriptions. It never copies a publisher's blurb or text
+  from another site, and it stores cover art as a link rather than a copy.
+- **Two open licences.** The factual catalogue - books, recordings, narrators,
+  series - is dedicated to the public domain (CC0), free to use in any app for
+  any purpose. The community-written characters and recaps are shared under
+  CC BY-SA (credit the source and share your own additions on the same terms).
+
+## Works with Audiobookshelf
+
+If you use Audiobookshelf, AudioSilo Meta can act as a **custom metadata
+provider** for it, so your Audiobookshelf library can pull narrators, recordings,
+series order and cover art from the community database. The
+[Audiobookshelf page](https://meta.audiosilo.app/audiobookshelf) has the setup
+steps, and also shows how to send your Audiobookshelf library back to help fill
+in the catalogue.
diff --git a/docs-users/faq.md b/docs-users/faq.md
index 1c761db..c021c9a 100644
--- a/docs-users/faq.md
+++ b/docs-users/faq.md
@@ -100,6 +100,15 @@ your files at any time - but the database also holds the things that
 favourites, shares, and folder overrides. That's why the data directory
 belongs in your backups.
 
+### Where does the extra "About this book" information come from?
+
+The description, characters, story-so-far recaps and series links that some
+books show in the player come from **AudioSilo Meta**, a free, community-built
+audiobook database at [meta.audiosilo.app](https://meta.audiosilo.app). It's
+optional (your admin switches the lookup on or off), and you can browse and add
+to the catalogue yourself - see
+[The community metadata site](./community/meta-site.md).
+
 ## Users & sharing
 
 ### How many users can I have, and can I limit what each one sees?
diff --git a/docs-users/listening/browsing.md b/docs-users/listening/browsing.md
index 5099d60..17f90ba 100644
--- a/docs-users/listening/browsing.md
+++ b/docs-users/listening/browsing.md
@@ -61,7 +61,7 @@ Some books show an extra section below the chapters with more about the title:
 - A **More in this series** row - other books in the same series, each opening its page on the metadata site so you can see what to read next.
 - A **View on AudioSilo Meta** link to the book's full entry.
 
-This information comes from the **AudioSilo community metadata database** at [meta.audiosilo.app](https://meta.audiosilo.app), a free, community-run catalogue of audiobook details. It appears only when the book can be matched (it carries an ASIN or ISBN) and your server has the metadata lookup switched on. The **Characters** and **Story so far** sections are contributed by the community, so a matched book shows them only once someone has written them - many books will have the About and series details but not these yet. If a book has no extra section, it simply isn't matched or your admin has turned the feature off - nothing else on the page changes. A server admin can switch the lookup on or off at any time from the admin console's Overview section (see the [console tour](../admin/console-tour.md#community-metadata-lookup)); when it is off, no book shows this section.
+This information comes from the **AudioSilo community metadata database** at [meta.audiosilo.app](https://meta.audiosilo.app), a free, community-run catalogue of audiobook details that you can also browse and contribute to yourself (see [The community metadata site](../community/meta-site.md)). It appears only when the book can be matched (it carries an ASIN or ISBN) and your server has the metadata lookup switched on. The **Characters** and **Story so far** sections are contributed by the community, so a matched book shows them only once someone has written them - many books will have the About and series details but not these yet. If a book has no extra section, it simply isn't matched or your admin has turned the feature off - nothing else on the page changes. A server admin can switch the lookup on or off at any time from the admin console's Overview section (see the [console tour](../admin/console-tour.md#community-metadata-lookup)); when it is off, no book shows this section.
 
 ## Search
 
diff --git a/docusaurus.config.ts b/docusaurus.config.ts
index bb1e6be..c0e8a36 100644
--- a/docusaurus.config.ts
+++ b/docusaurus.config.ts
@@ -111,6 +111,7 @@ const config: Config = {
             {label: 'audiosilo-server', href: 'https://github.com/KodeStar/audiosilo-server'},
             {label: 'audiosilo-frontend', href: 'https://github.com/KodeStar/audiosilo-frontend'},
             {label: 'audiosilo-manager', href: 'https://github.com/KodeStar/audiosilo-manager'},
+            {label: 'audiosilo-meta', href: 'https://github.com/KodeStar/audiosilo-meta'},
             {label: 'audiosilo-docs', href: 'https://github.com/KodeStar/audiosilo-docs'},
           ],
         },
@@ -132,6 +133,7 @@ const config: Config = {
             {label: 'Server', href: 'https://github.com/KodeStar/audiosilo-server'},
             {label: 'Player app', href: 'https://github.com/KodeStar/audiosilo-frontend'},
             {label: 'Desktop manager', href: 'https://github.com/KodeStar/audiosilo-manager'},
+            {label: 'Metadata database', href: 'https://github.com/KodeStar/audiosilo-meta'},
           ],
         },
         {
diff --git a/screenshots/README.md b/screenshots/README.md
index d844b40..38e80e5 100644
--- a/screenshots/README.md
+++ b/screenshots/README.md
@@ -18,7 +18,10 @@ npm install && npx playwright install chromium   # first time only
 LibriVox library (cached in `.cache/library`; `MAX_FILES=3` chapter files per
 book keeps it ~100 MB), starts a demo-mode server on `:8790` serving the
 frontend's web export (plus a `--setup` instance on `:8791` for the wizard
-shot), then runs the Playwright captures and backfills placeholders.
+shot), builds the `audiosilo-meta` data artifact + site and starts a
+`metaserve` on `:8795` for the meta-site shots, then runs the Playwright
+captures and backfills placeholders. `SKIP_META=1` skips the meta stack (for a
+web/admin-only run).
 
 Every capture is optimized in place with **pngquant** (`brew install pngquant`)
 - a lossy-palette pass that shrinks the retina PNGs ~60% with no perceptible
@@ -56,6 +59,27 @@ semi-manually instead:
 Anything not replaced stays a labelled placeholder - visible in the docs as
 "regenerate me", never a broken image.
 
+## AudioSilo Meta site captures
+
+The `meta/` shots come from `capture-meta.mjs`, which drives the
+meta.audiosilo.app site (the sibling `audiosilo-meta` repo: an Astro static
+site served same-origin by the Go `metaserve` together with its read-only
+`/api/v1` JSON). `run.sh` handles it end-to-end: it builds the data artifact
+(`metabuild` into `.cache/meta.sqlite`), builds the site once if
+`audiosilo-meta/site/dist` is missing, and starts `metaserve` on `:8795`.
+Because the site build is reused while it exists, **delete
+`audiosilo-meta/site/dist` after changing the meta site's UI** so the next run
+rebuilds it - otherwise the shots silently show the old UI.
+
+Needs: Go 1.25+, and (only for the site build) yarn + Node 24. After the caches
+are warm (node_modules, Go modules, `site/dist`), a run only touches the network
+for remote cover images - and the capture waits on rendered content, not on
+covers, so missing cover art never fails a shot.
+
+`capture-meta.mjs` is standalone: point `META_BASE` at any `metaserve` (or the
+live site) and run `META_BASE=http://127.0.0.1:8795 node capture-meta.mjs` - it
+needs no other services (the /import fixture is vendored in `fixtures/`).
+
 ## The store/marketing pipeline is separate
 
 `~/dev/audiosilo/store/tools` + `SCREENSHOTS.md` produce the app-store and
diff --git a/screenshots/capture-admin.mjs b/screenshots/capture-admin.mjs
index 13b542a..d8d0b84 100644
--- a/screenshots/capture-admin.mjs
+++ b/screenshots/capture-admin.mjs
@@ -7,7 +7,7 @@
 // Before capturing it provisions a little demo state through the admin API
 // (a listener account, an invite, a share) so the console looks lived-in.
 import {chromium} from 'playwright';
-import {sleep, shoot, step} from './lib.mjs';
+import {sleep, shoot, step, DESKTOP_CONTEXT} from './lib.mjs';
 
 const ORIGIN = (process.env.AS_ORIGIN || 'http://127.0.0.1:8790').replace(/\/$/, '');
 const PASSWORD = process.env.ADMIN_PASSWORD;
@@ -77,12 +77,7 @@ await step('provision listener + share', async () => {
 
 // ── Capture ────────────────────────────────────────────────────────────────
 const browser = await chromium.launch();
-const ctx = await browser.newContext({
-  viewport: {width: 1440, height: 900},
-  deviceScaleFactor: 2,
-  colorScheme: 'dark',
-  reducedMotion: 'reduce',
-});
+const ctx = await browser.newContext(DESKTOP_CONTEXT);
 const page = await ctx.newPage();
 
 const nav = async (section) => {
diff --git a/screenshots/capture-meta.mjs b/screenshots/capture-meta.mjs
new file mode 100644
index 0000000..9877309
--- /dev/null
+++ b/screenshots/capture-meta.mjs
@@ -0,0 +1,157 @@
+// Captures the meta.audiosilo.app site documentation screenshots against a
+// locally-running metaserve (run.sh builds the data artifact + site and starts
+// it). Env:
+//   META_BASE       the site/API origin (default http://127.0.0.1:8795)
+//   IMPORT_FIXTURE  library export dropped on /import (default: the vendored
+//                   fixtures/openaudible-books.json, synthetic data)
+//
+// The site is an Astro static export whose React islands fetch /api/v1 on the
+// CLIENT, so every wait here is content-based (a rendered element the island
+// produced), never a bare load event. Remote cover images may be slow or 404;
+// steps wait on text/layout and give images a short bounded settle rather than
+// blocking on them, so a missing cover never fails a shot.
+//
+// Standalone: point META_BASE at any metaserve (or the live site) and run
+// `node capture-meta.mjs` - it needs no other services.
+import path from 'node:path';
+import {fileURLToPath} from 'node:url';
+import {chromium} from 'playwright';
+import {sleep, shoot, step, DESKTOP_CONTEXT} from './lib.mjs';
+
+const HERE = path.dirname(fileURLToPath(import.meta.url));
+const BASE = (process.env.META_BASE || 'http://127.0.0.1:8795').replace(/\/$/, '');
+
+// The work referenced by the developer docs: two recordings (Jim Dale + Stephen
+// Fry), and a fully-populated community layer (characters + a story-so-far
+// recap), so the same book anchors both the work and characters shots.
+const WORK_ID = 'harry-potter-and-the-philosophers-stone';
+const SERIES_ID = 'harry-potter';
+// An OpenAudible books.json sample (synthetic ASINs, vendored here) that drives
+// the /import diff to a populated results view without matching real catalogue
+// entries.
+const IMPORT_FIXTURE =
+  process.env.IMPORT_FIXTURE || path.resolve(HERE, 'fixtures/openaudible-books.json');
+
+const browser = await chromium.launch();
+const ctx = await browser.newContext(DESKTOP_CONTEXT);
+const page = await ctx.newPage();
+
+// domcontentloaded, not networkidle: a hanging remote cover request would time
+// the navigation out even though the page content rendered long before. Each
+// step's own content wait (+ a bounded settle for cover paint) gates the shot.
+const goto = (route) => page.goto(`${BASE}${route}`, {waitUntil: 'domcontentloaded', timeout: 45000});
+
+// ── Home: search hero + stats band + latest additions ───────────────────────
+await step('home', async () => {
+  await goto('/');
+  await page.waitForSelector('input[role="combobox"]', {timeout: 20000});
+  // The stats band and latest-additions grid are client:visible islands - they
+  // only hydrate + fetch once scrolled into view. A single jump to the bottom
+  // skips past them (the observer never sees the crossing), so step down one
+  // viewport at a time to arm each island, wait for the work-card links (proof
+  // the API resolved), then anchor back to the top so the shot leads with the
+  // search hero.
+  await page.evaluate(async () => {
+    const step = Math.round(window.innerHeight * 0.8);
+    for (let y = 0; y <= document.body.scrollHeight; y += step) {
+      window.scrollTo(0, y);
+      await new Promise((r) => setTimeout(r, 150));
+    }
+  });
+  await page.waitForSelector('a[href^="/work?id="]', {timeout: 20000});
+  await page.evaluate(() => window.scrollTo(0, 0));
+  await sleep(2000); // let the stats count-up settle + covers paint (bounded)
+  await shoot(page, 'meta/home.png');
+});
+
+// ── Search: the grouped dropdown for a seed-data query ───────────────────────
+await step('search', async () => {
+  // Continues on the home page; re-navigate if the home step failed partway.
+  if (page.url() !== `${BASE}/`) {
+    await goto('/');
+    await page.waitForSelector('input[role="combobox"]', {timeout: 20000});
+  }
+  const box = page.locator('input[role="combobox"]').first();
+  await box.click();
+  await box.fill('harry potter');
+  // Results render as role=option rows inside the listbox once the debounced
+  // fetch returns; wait for the first one rather than a fixed sleep.
+  await page.waitForSelector('[role="option"]', {timeout: 15000});
+  await sleep(1200); // covers in the option rows settle
+  await shoot(page, 'meta/search.png');
+});
+
+// ── Work detail: cover, metadata, recordings ─────────────────────────────────
+await step('work', async () => {
+  await goto(`/work?id=${WORK_ID}`);
+  await page.getByRole('heading', {level: 1}).first().waitFor({state: 'visible', timeout: 20000});
+  // Recordings are the client island's main payload; wait for a recording card
+  // (the cards only render inside the Recordings section).
+  await page.waitForSelector('article', {timeout: 20000});
+  await sleep(1800); // cover art settle (bounded)
+  await shoot(page, 'meta/work.png');
+});
+
+// ── Characters: the community layer with one card opened ─────────────────────
+// Continues on the work page the previous step loaded (same book by design);
+// re-navigates only if that step failed partway, so one transient failure
+// cannot cost both shots.
+await step('characters', async () => {
+  if (!page.url().includes(`/work?id=${WORK_ID}`)) {
+    await goto(`/work?id=${WORK_ID}`);
+    await page.getByRole('heading', {level: 1}).first().waitFor({state: 'visible', timeout: 20000});
+  }
+  // The Characters tab only exists when the work has a characters sidecar.
+  const tab = page.getByRole('tab', {name: /characters/i}).first();
+  await tab.waitFor({state: 'visible', timeout: 20000});
+  await tab.click();
+  await page.waitForSelector('#panel-characters article', {timeout: 15000});
+  // Open the first character card that has a description (its disclosure button
+  // starts collapsed) so the shot shows the spoiler-gated body revealed.
+  const disclosure = page.locator('#panel-characters button[aria-expanded="false"]').first();
+  await disclosure.waitFor({state: 'visible', timeout: 10000});
+  await disclosure.click();
+  await page.waitForSelector('#panel-characters p[id^="char-desc-"]', {timeout: 10000});
+  // Bring the section (intro + grid + opened card) to the top of the viewport.
+  await page.locator('#panel-characters').scrollIntoViewIfNeeded().catch(() => {});
+  await sleep(1200);
+  await shoot(page, 'meta/characters.png');
+});
+
+// ── Series: the ordered list of volumes ──────────────────────────────────────
+await step('series', async () => {
+  await goto(`/series?id=${SERIES_ID}`);
+  await page.getByRole('heading', {level: 1}).first().waitFor({state: 'visible', timeout: 20000});
+  // Each volume is a work link in an ordered list; wait for the list to fill.
+  await page.waitForSelector('ol li a[href^="/work?id="]', {timeout: 20000});
+  await sleep(1500); // thumbnail covers settle (bounded)
+  await shoot(page, 'meta/series.png');
+});
+
+// ── Contribute: the coverage browser ─────────────────────────────────────────
+await step('contribute', async () => {
+  await goto('/contribute');
+  // The coverage island renders stat cards + the browser heading once /coverage
+  // resolves; wait for the heading text rather than a load event.
+  await page.getByText('Browse books by coverage', {exact: false}).first().waitFor({timeout: 20000});
+  await sleep(1200);
+  await shoot(page, 'meta/contribute.png');
+});
+
+// ── Import: drop a library export, capture the diff results ───────────────────
+await step('import', async () => {
+  await goto('/import');
+  // The file input is visually hidden (sr-only); setInputFiles drives it
+  // directly, which triggers the in-browser diff against the API.
+  const input = page.locator('input[type="file"]').first();
+  await input.waitFor({state: 'attached', timeout: 20000});
+  await input.setInputFiles(IMPORT_FIXTURE);
+  // The diff finishes on the results stat tiles ("In the database" etc.).
+  await page.getByText('In the database', {exact: false}).first().waitFor({timeout: 30000});
+  await sleep(1000);
+  await shoot(page, 'meta/import.png');
+});
+
+await ctx.close();
+await browser.close();
+console.log('capture-meta: done.');
diff --git a/screenshots/fixtures/openaudible-books.json b/screenshots/fixtures/openaudible-books.json
new file mode 100644
index 0000000..0f2a402
--- /dev/null
+++ b/screenshots/fixtures/openaudible-books.json
@@ -0,0 +1,57 @@
+[
+  {
+    "asin": "B0SYNTH001",
+    "product_id": "ignored-123",
+    "title": "The Iron Ledger: Book One of the Ledger Wars",
+    "title_short": "The Iron Ledger",
+    "author": "Mara Quill, Devon Ashe",
+    "narrated_by": "Priya Lund, Tomas Reed",
+    "series_name": "The Ledger Wars",
+    "series_sequence": "1",
+    "duration": "12:03:40",
+    "seconds": 43420,
+    "release_date": "2025-04-15",
+    "publisher": "Vantage Audio",
+    "language": "english",
+    "region": "US",
+    "abridged": false,
+    "image_url": "https://covers.example.com/iron-ledger.jpg",
+    "description": "PUBLISHER BLURB THAT MUST NOT BE IMPORTED.",
+    "summary": "MORE PUBLISHER COPY.",
+    "genre": "Fantasy",
+    "rating_average": 4.6,
+    "chapters": [
+      { "start_offset_ms": 0, "length_ms": 60000, "start_offset_sec": 0, "title": "Prologue\t" },
+      { "start_offset_ms": 60000, "length_ms": 120000, "start_offset_sec": 60, "title": "  Chapter One  " },
+      { "start_offset_ms": 180000, "length_ms": 90000, "start_offset_sec": 180, "title": "Chapter Two" }
+    ]
+  },
+  {
+    "asin": "B0SYNTH002",
+    "title_short": "The Bronze Ledger",
+    "author": "Mara Quill, Devon Ashe",
+    "narrated_by": "Priya Lund, Tomas Reed",
+    "series_name": "The Ledger Wars",
+    "series_sequence": "2",
+    "seconds": 50000,
+    "release_date": "2025-09-01",
+    "publisher": "Vantage Audio",
+    "language": "english",
+    "region": "US",
+    "abridged": null,
+    "image_url": "https://covers.example.com/bronze-ledger.jpg"
+  },
+  {
+    "asin": "B0SYNTH003",
+    "title_short": "Grenzland",
+    "author": "Ingrid Falk",
+    "narrated_by": "Klaus Berger",
+    "series_name": "The Ledger Wars",
+    "series_sequence": "1-3.5",
+    "seconds": 72000,
+    "release_date": "2024-02-10",
+    "publisher": "Nord Hörbuch",
+    "language": "german",
+    "region": "DE"
+  }
+]
diff --git a/screenshots/lib.mjs b/screenshots/lib.mjs
index 84a2e63..cf0b160 100644
--- a/screenshots/lib.mjs
+++ b/screenshots/lib.mjs
@@ -12,6 +12,16 @@ export const CACHE = path.resolve(here, '.cache');
 
 export const sleep = (ms) => new Promise((r) => setTimeout(r, ms));
 
+// The standard desktop capture context (retina, dark, motion-reduced) shared by
+// the capture scripts. capture-web builds its own parameterized contexts (phone
+// profile, connect/demo variants) on purpose.
+export const DESKTOP_CONTEXT = {
+  viewport: {width: 1440, height: 900},
+  deviceScaleFactor: 2,
+  colorScheme: 'dark',
+  reducedMotion: 'reduce',
+};
+
 // Raw Playwright PNGs at deviceScaleFactor 2 are large (a full desktop shot is
 // ~0.5-1.3 MB); pngquant's lossy-palette pass shrinks them ~60% with no
 // perceptible loss on UI captures. It's a deterministic transform run on every
diff --git a/screenshots/manifest.mjs b/screenshots/manifest.mjs
index eb720cc..cfdfb64 100644
--- a/screenshots/manifest.mjs
+++ b/screenshots/manifest.mjs
@@ -9,6 +9,7 @@
 // `capture` says which script produces it:
 //   'web'     - capture-web.mjs (web player via a warmed demo session)
 //   'admin'   - capture-admin.mjs (admin console + public pages)
+//   'meta'    - capture-meta.mjs (the meta.audiosilo.app site via a local metaserve)
 //   'manager' - capture-manager.mjs (Wails dev server; falls back to placeholder)
 
 export const SHOTS = [
@@ -39,6 +40,15 @@ export const SHOTS = [
   {file: 'server/connect-page.png', capture: 'admin', title: 'Connect page', hint: 'public connect page (auth-code box / QR)'},
   {file: 'server/setup-wizard.png', capture: 'admin', title: 'Setup wizard', hint: 'first-run --setup wizard page'},
 
+  // ── AudioSilo Meta site (meta.audiosilo.app; desktop 1440x900, dark) ────
+  {file: 'meta/home.png', capture: 'meta', title: 'Meta - Home', hint: 'search hero, live stats band, latest additions grid'},
+  {file: 'meta/search.png', capture: 'meta', title: 'Meta - Search', hint: 'search dropdown with grouped work/person/series results'},
+  {file: 'meta/work.png', capture: 'meta', title: 'Meta - Work detail', hint: 'a work page with its cover, metadata and multiple recordings'},
+  {file: 'meta/characters.png', capture: 'meta', title: 'Meta - Characters', hint: 'the community characters section, one card opened'},
+  {file: 'meta/series.png', capture: 'meta', title: 'Meta - Series', hint: 'a series page listing its ordered volumes'},
+  {file: 'meta/contribute.png', capture: 'meta', title: 'Meta - Contribute', hint: 'the coverage browser (what still needs characters/recaps)'},
+  {file: 'meta/import.png', capture: 'meta', title: 'Meta - Import', hint: 'the in-browser library-export diff results'},
+
   // ── Desktop manager (Wails) ─────────────────────────────────────────────
   {file: 'manager/servers.png', capture: 'manager', title: 'Manager - Servers', hint: 'server list home'},
   {file: 'manager/server-detail.png', capture: 'manager', title: 'Manager - Server detail', hint: 'a connected server with its libraries'},
diff --git a/screenshots/run.sh b/screenshots/run.sh
index e85e25a..283f822 100755
--- a/screenshots/run.sh
+++ b/screenshots/run.sh
@@ -5,28 +5,36 @@
 #   2. seeds a small public-domain library (LibriVox; cached in .cache/library),
 #   3. starts a demo-mode server (port 8790) serving the frontend's web export,
 #      plus a second --setup instance (port 8791) for the wizard shot,
-#   4. runs the Playwright captures (web player + admin console + public pages),
-#   5. backfills placeholders for anything not captured (e.g. the desktop
+#   4. builds the AudioSilo Meta data artifact + site and starts a local
+#      metaserve (port 8795) for the meta.audiosilo.app site captures,
+#   5. runs the Playwright captures (web player + admin console + public pages
+#      + the meta site),
+#   6. backfills placeholders for anything not captured (e.g. the desktop
 #      manager on a headless run - see README.md for manager captures).
 #
 # Prereqs: Go 1.25+, Node 24, ffmpeg/ffprobe, `npm install` +
 # `npx playwright install chromium` in this directory, and a web export at
 # ../audiosilo-frontend/dist (run audiosilo-server/scripts/build-web.sh once).
+# The meta section also needs yarn (for the sibling audiosilo-meta site build);
+# once its caches are warm, only remote cover images touch the network.
 #
 # Env knobs: MAX_FILES (chapter files per seeded book, default 3),
-#            SKIP_SEED=1 (reuse the cached library as-is).
+#            SKIP_SEED=1 (reuse the cached library as-is),
+#            SKIP_META=1 (skip the meta site stack + its captures).
 set -euo pipefail
 
 HERE="$(cd "$(dirname "$0")" && pwd)"
 WORKSPACE="$(cd "$HERE/../.." && pwd)"
 SERVER="$WORKSPACE/audiosilo-server"
 FRONTEND="$WORKSPACE/audiosilo-frontend"
+META="$WORKSPACE/audiosilo-meta"
 CACHE="$HERE/.cache"
 LIBRARY="$CACHE/library"
 DATA="$CACHE/data"
 SETUP_DATA="$CACHE/setup-data"
 PORT=8790
 SETUP_PORT=8791
+META_PORT=8795
 
 mkdir -p "$CACHE"
 
@@ -54,9 +62,18 @@ fi
 cleanup() {
   [ -n "${MAIN_PID:-}" ] && kill "$MAIN_PID" 2>/dev/null || true
   [ -n "${SETUP_PID:-}" ] && kill "$SETUP_PID" 2>/dev/null || true
+  [ -n "${META_PID:-}" ] && kill "$META_PID" 2>/dev/null || true
 }
 trap cleanup EXIT
 
+wait_healthy() { # <url> <logfile> <name>
+  for i in $(seq 1 60); do
+    curl -fsS "$1" >/dev/null 2>&1 && return 0
+    [ "$i" = 60 ] && { echo "$3 never became healthy; see $2"; exit 1; }
+    sleep 1
+  done
+}
+
 echo "==> starting demo server on :$PORT (fresh data dir)"
 rm -rf "$DATA" && mkdir -p "$DATA"
 cat > "$DATA/config.yaml" <<EOF
@@ -83,11 +100,7 @@ AUDIOSILO_BIND="127.0.0.1:$SETUP_PORT" AUDIOSILO_TLS_MODE=off \
 SETUP_PID=$!
 
 echo "==> waiting for the demo server"
-for i in $(seq 1 60); do
-  curl -fsS "http://127.0.0.1:$PORT/healthz" >/dev/null 2>&1 && break
-  [ "$i" = 60 ] && { echo "server never became healthy; see $CACHE/server.log"; exit 1; }
-  sleep 1
-done
+wait_healthy "http://127.0.0.1:$PORT/healthz" "$CACHE/server.log" "server"
 sleep 8   # let the startup scan index the seeded books
 
 ADMIN_PASSWORD="$(grep 'Admin password' "$CACHE/server.log" | awk -F': ' '{print $2}' | tr -d ' ')"
@@ -96,7 +109,28 @@ if [ -z "$ADMIN_PASSWORD" ]; then
 fi
 SETUP_URL="$(grep -o "http://[^ ]*/setup#token=[^ ]*" "$CACHE/setup.log" | head -1 || true)"
 
-# ── 4. Captures ─────────────────────────────────────────────────────────────
+# ── 4. AudioSilo Meta site (data artifact + site build + metaserve) ─────────
+if [ "${SKIP_META:-0}" != "1" ]; then
+  echo "==> building the meta data artifact"
+  (cd "$META" && go run ./cmd/metabuild -o "$CACHE/meta.sqlite")
+
+  if [ ! -f "$META/site/dist/index.html" ]; then
+    echo "==> no meta site build found; building (yarn, Node 24)"
+    (cd "$META/site" && yarn install --frozen-lockfile && yarn build)
+  fi
+
+  echo "==> starting metaserve on :$META_PORT"
+  (cd "$META" && go build -o bin/metaserve ./cmd/metaserve)
+  "$META/bin/metaserve" --db "$CACHE/meta.sqlite" \
+    --site "$META/site/dist" --addr "127.0.0.1:$META_PORT" \
+    > "$CACHE/metaserve.log" 2>&1 &
+  META_PID=$!
+
+  echo "==> waiting for metaserve"
+  wait_healthy "http://127.0.0.1:$META_PORT/healthz" "$CACHE/metaserve.log" "metaserve"
+fi
+
+# ── 5. Captures ─────────────────────────────────────────────────────────────
 cd "$HERE"
 echo "==> capturing web player"
 AS_BASE="http://127.0.0.1:$PORT/web/" node capture-web.mjs
@@ -105,7 +139,12 @@ echo "==> capturing admin console + public pages"
 AS_ORIGIN="http://127.0.0.1:$PORT" ADMIN_PASSWORD="$ADMIN_PASSWORD" \
   SETUP_URL="$SETUP_URL" node capture-admin.mjs
 
-# ── 5. Backfill placeholders for anything missing ──────────────────────────
+if [ "${SKIP_META:-0}" != "1" ]; then
+  echo "==> capturing the meta site"
+  META_BASE="http://127.0.0.1:$META_PORT" node capture-meta.mjs
+fi
+
+# ── 6. Backfill placeholders for anything missing ──────────────────────────
 echo "==> backfilling placeholders"
 node placeholders.mjs
 
diff --git a/sidebars-developers.ts b/sidebars-developers.ts
index 0717a74..9893bec 100644
--- a/sidebars-developers.ts
+++ b/sidebars-developers.ts
@@ -55,6 +55,16 @@ const sidebars: SidebarsConfig = {
         'manager/transfers',
       ],
     },
+    {
+      type: 'category',
+      label: 'Metadata database (audiosilo-meta)',
+      items: [
+        'meta/overview',
+        'meta/data-model',
+        'meta/api',
+        'meta/contributing-data',
+      ],
+    },
     {
       type: 'category',
       label: 'Contributing',
diff --git a/sidebars-users.ts b/sidebars-users.ts
index 8f0f3a2..c4b2c40 100644
--- a/sidebars-users.ts
+++ b/sidebars-users.ts
@@ -50,6 +50,11 @@ const sidebars: SidebarsConfig = {
         'manager/audible-backup',
       ],
     },
+    {
+      type: 'category',
+      label: 'Community metadata',
+      items: ['community/meta-site'],
+    },
     'demo',
     'troubleshooting',
     'faq',
diff --git a/static/img/screenshots/meta/characters.png b/static/img/screenshots/meta/characters.png
new file mode 100644
index 0000000..a226638
Binary files /dev/null and b/static/img/screenshots/meta/characters.png differ
diff --git a/static/img/screenshots/meta/contribute.png b/static/img/screenshots/meta/contribute.png
new file mode 100644
index 0000000..4e2c86d
Binary files /dev/null and b/static/img/screenshots/meta/contribute.png differ
diff --git a/static/img/screenshots/meta/home.png b/static/img/screenshots/meta/home.png
new file mode 100644
index 0000000..b4f9e06
Binary files /dev/null and b/static/img/screenshots/meta/home.png differ
diff --git a/static/img/screenshots/meta/import.png b/static/img/screenshots/meta/import.png
new file mode 100644
index 0000000..410430d
Binary files /dev/null and b/static/img/screenshots/meta/import.png differ
diff --git a/static/img/screenshots/meta/search.png b/static/img/screenshots/meta/search.png
new file mode 100644
index 0000000..f108f68
Binary files /dev/null and b/static/img/screenshots/meta/search.png differ
diff --git a/static/img/screenshots/meta/series.png b/static/img/screenshots/meta/series.png
new file mode 100644
index 0000000..73de745
Binary files /dev/null and b/static/img/screenshots/meta/series.png differ
diff --git a/static/img/screenshots/meta/work.png b/static/img/screenshots/meta/work.png
new file mode 100644
index 0000000..02d4dc3
Binary files /dev/null and b/static/img/screenshots/meta/work.png differ