Kyoo/scanner/README.md
2025-04-09 00:30:42 +02:00

70 lines
2.4 KiB
Markdown

# Scanner
## Workflow (for v5, not current)
In order of action:
- Scanner gets `/videos` & scan file system to list all new videos
- Scanner guesses as much as possible from filename/path ALONE (no external database query).
- Format should be:
```json5
{
path: string,
version: number,
part: number | null,
rendering: sha(path except version & part),
guess: {
kind: movie | episode | trailer | interview | ...,
name: string,
year: number | null,
season?: number,
episode?: number,
...
},
}
```
- Apply remaps from lists (AnimeList + thexem). Format is now:
```json5
{
path: string,
version: number,
part: number | null,
rendering: sha(path except version & part),
guess: {
kind: movie | episode | trailer | interview | ...,
name: string,
year: number | null,
season?: number,
episodes?: number[],
absolutes?: number[],
externalId: Record<string, {showId, season, number}[]>,
remap: {
from: "thexem",
oldSeason: number,
oldEpisodes: number[],
},
...
},
}
```
- If kind is episode, try to find the serie's id on kyoo (using the previously fetched data from `/videos`):
- if another video in the list of already registered videos has the same `kind`, `name` & `year`, assume it's the same
- if a match is found, add to the video's json:
```json5
{
entries: (uuid | slug | {
show: uuid | slug,
season: number,
episode: number,
externalId?: Record<string, {showId, season, number}> // takes priority over season/episode for matching if we have one
})[],
}
```
- Scanner pushes everything to the api in a single post `/videos` call
- Api registers every video in the database
- For each video without an associated entry, the guess data + the video's id is sent to the Matcher via a queue.
- Matcher retrieves metadata from the movie/serie + ALL episodes/seasons (from an external provider)
- Matcher pushes every metadata to the api (if there are 1000 episodes but only 1 video, still push the 1000 episodes)