Files
bookhoard/docs/developer/api/sync/koreader-protocol.md
T
john-okeefe d429534b12
Release / build-and-push (push) Successful in 2m24s
docs(api): deleted-annotation history + KOReader deletion propagation
New media-items/deleted_annotations.md for the list/restore/purge
endpoints; endpoint index updated. The KOReader protocol page documents
deleted_highlights/deleted_bookmarks on the progress push and the
deletion-propagation contract: keys learned only from server pulls,
explicit arrays only (never absence), tombstone convergence via the
metadata fetch, no resurrection from stale replays, and the web history
as the restore path.
2026-08-22 13:16:54 -04:00

210 lines
7.9 KiB
Markdown

# KOReader Sync Protocol
KOReader uses a custom JSON-based sync protocol.
## KOReader Progress Sync
**Endpoint**: `POST /api/sync/koreader/progress`
**Auth**: Device token required
**Content-Type**: `application/json`
### Request Headers
| Header | Type | Required | Description |
| ------------- | ------ | -------- | ------------------- |
| Authorization | string | Yes | Bearer device token |
| Content-Type | string | Yes | application/json |
### Request Body
| Field | Type | Required | Description |
| ------------------ | ------- | -------- | ---------------------------------------------------- |
| library_id | string | No | Library UUID |
| books | array | Yes | Array of book sync data |
| books[].uuid | string | No\* | Book UUID (highest-confidence match; omitted on first sync of a newly downloaded book) |
| books[].sha256 | string | No\* | Full-file SHA-256 (64 hex chars); used to resolve the book when `uuid` is absent |
| books[].file_path | string | No | Device-local file path; used to create/look up a device file alias |
| books[].title | string | Yes | Book title |
| books[].authors | array | Yes | Array of author names |
| books[].progress | float | Yes | Progress percentage (0-1) |
| books[].percentage | float | Yes | Progress percentage (0-1) |
| books[].last_read | string | Yes | ISO 8601 timestamp |
| books[].chapter | integer | No | Current chapter |
| books[].epubcfi | string | No | EPUB CFI location |
| books[].character | integer | No | Character offset |
| books[].bookmarks | array | No | Array of bookmarks/highlights (shape, color mapping, and echo/dedup rules: see [Sync Bookmarks](../koreader/sync_bookmarks.md)) |
| books[].deleted_highlights | array | No | Highlights deleted on the device: `[{ "dedup_key": "..." }]` — keys previously served to this device (see [Deletion propagation](#deletion-propagation)) |
| books[].deleted_bookmarks | array | No | Bookmarks deleted on the device: `[{ "dedup_key": "..." }]` |
\* At least one of `uuid` or `sha256` should be present. The server resolves the
book through the shared `BookResolver` with this priority: `uuid``sha256`
`file_path` alias → `title`/`author`. SHA-256 matching is **format-aware**: it
checks `media_items.file_sha256` first, then `media_item_formats.file_sha256`, so
a converted file (e.g. KEPUB or PDF) downloaded via OPDS matches even though its
hash differs from the primary format's hash.
### Example Request
```json
{
"library_id": "optional-uuid",
"books": [
{
"uuid": "book-uuid",
"title": "Book Title",
"authors": ["Author Name"],
"progress": 0.45,
"percentage": 0.45,
"last_read": "2026-01-30T20:00:00Z",
"chapter": 3,
"epubcfi": "epubcfi(/6/4/2:15)",
"character": 15432,
"bookmarks": [
{
"chapter": 3,
"datetime": "2026-01-30T19:55:00Z",
"notes": "highlighted text",
"pos0": "epubcfi(/6/4/2:15)",
"pos1": "epubcfi(/6/4/2:20)",
"page": 45,
"text": "highlighted text excerpt",
"type": "highlight"
}
],
"highlights": [],
"notes": []
}
]
}
```
### Response (202 Accepted)
```json
{
"sync_status": "accepted",
"books_synced": 1,
"conflicts": [
{
"book_uuid": "book-uuid",
"conflict_type": "progress_mismatch",
"device_progress": 0.45,
"server_progress": 0.42,
"resolution": "device_wins"
}
]
}
```
## Book Resolution (UUID lookup)
**Endpoint**: `GET /api/sync/koreader/resolve?sha256={hash}`
**Auth**: Device token required
Read-only lookup mapping a file SHA-256 to the book's UUID (format-aware,
same `BookResolver` path as the progress push). Devices call this on the
first open of a newly downloaded book to learn the UUID **before** their
first pull. Full details: [Resolve Book](../koreader/resolve_book.md).
This matters for conflict avoidance: a device that pushes to bootstrap its
identity transmits its current (first-page) position, which the server
treats as a real progress update — overwriting/conflicting with genuine
mid-read progress from other sources. Resolve, then pull, then push.
### Example Request
```http
GET /api/sync/koreader/resolve?sha256=e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
Authorization: Bearer device-token
```
### Response (200 OK)
```json
{
"book_uuid": "book-uuid",
"sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"title": "Book Title",
"author": "Author Name"
}
```
404 when no book in the library matches the hash.
## KOReader Metadata Fetch
**Endpoint**: `GET /api/sync/koreader/metadata/{book_uuid}`
**Auth**: Device token required
### Example Request
```http
GET /api/sync/koreader/metadata/book-uuid
Authorization: Bearer device-token
```
### Response (200 OK)
```json
{
"uuid": "book-uuid",
"sha256": "ff3e4501bf9d72dea2ae28731a6cb5b83d7a7532c05b5d2dd083d0dbc9193ebf",
"title": "Book Title",
"authors": ["Author Name"],
"progress": {
"percentage": 0.42,
"character": 15432,
"epubcfi": "epubcfi(/6/4/2:15)",
"chapter": 3,
"chapter_progress": 0.234
},
"annotations": {
"highlights": [...],
"notes": [...],
"bookmarks": [...]
},
"last_sync": "2026-01-30T20:00:00Z"
}
```
`sha256` is the canonical primary-format hash of the book on the server. It is
returned so clients can cache it regardless of how the book was originally
obtained. The library list endpoint (`GET /api/sync/koreader/library`) includes
the same `sha256` field on each book.
## Deletion propagation
The progress push is upsert-only: absence of an annotation from
`highlights`/`notes`/`bookmarks` is **never** interpreted as a delete (a
client with a category disabled must not wipe the server). Deletions are
reported explicitly:
- Devices remember the `dedup_key` of every annotation the server served
them (persisted locally, e.g. KOReader's sidecar `bookhoard_known_keys`).
- When one of those annotations no longer exists locally, the next push
lists its key in `deleted_highlights` / `deleted_bookmarks`.
- The server tombstones the matching rows (`deleted = TRUE`, kept for the
retention window). Tombstones are served back to *other* devices via the
metadata fetch's `deleted_highlights` / `deleted_bookmarks` arrays so the
deletion converges everywhere.
- A stale replay pushing the annotation's content cannot resurrect the
tombstone: device pushes carry no modification timestamp, so the save is
treated as older than the delete.
- Restoring is possible from the web book page's deleted-annotation
history (`GET /api/media-items/:id/annotations/deleted`, restore/purge
endpoints) until the retention window lapses.
Because keys are only learned from server pulls, a device-native annotation
deleted locally is simply never pushed again — it can never be mis-flagged
as a server annotation deletion.
## Book identification
Every client/sync interface (KOReader, Kobo, OPDS, the device-link UI, and any
future mobile app) resolves books through a single shared service:
[`internal/services/book_resolver.go`](../../../internal/services/book_resolver.go).
The import-time SHA-256 (stored on `media_items.file_sha256`, plus a per-format
hash on `media_item_formats.file_sha256` for KEPUB/PDF) is the canonical shared
identifier. New clients should resolve by SHA-256 via `BookResolver` rather than
re-implementing their own matcher.