Wrap a Stash Scraper

The Stash community has hundreds of YAML-defined site scrapers — battle-tested for the long tail of porn-site metadata. Prismedia runs them through a compatibility adapter so you don't have to rewrite them in TypeScript or Python.

How it fits

Stash YAML scraper definition
            │
            ▼
   synthesized stash-compat manifest ─┐
                               │
                               ▼
   Prismedia.Infrastructure/StashCompat (native engine)
                               │
                               ▼
        Plugin execution envelope (same as native plugins)
                               │
                               ▼
              NormalizedVideoResult / NormalizedFolderResult / ...

The adapter:

• Maps Prismedia action names to Stash action names (videoByURL → sceneByURL, videoByName → sceneByName, videoByFragment → sceneByFragment, etc.).
• Runs the YAML scraper's HTTP / regex / xpath pipeline.
• Normalizes the output to Prismedia's shape.

A stash-compat plugin

The plugin directory just contains a manifest pointing at the YAML definition:

my-stash-plugin/
├── manifest.yml
└── definitions/
    └── ExampleSite.yml          ← the original Stash YAML

manifest.yml

id: example-site
name: Example Site Scraper
version: 0.1.0
runtime: stash-compat
stashDefinition: definitions/ExampleSite.yml

isNsfw: true                       # most Stash scrapers are NSFW

capabilities:
  videoByURL: true
  videoByFragment: true
  performerByURL: true
  performerByName: true

You declare which Prismedia capabilities the wrapped scraper provides — capabilities the YAML doesn't actually implement will fail at execution time, so be honest.

What the adapter handles

The adapter speaks the standard Stash scraper YAML dialect:

• sceneByURL, sceneByName, sceneByFragment, sceneByQueryFragment
• performerByURL, performerByName, performerByFragment
• movieByURL, movieByName, movieByFragment
• galleryByURL, galleryByName, galleryByFragment

Action mapping (Prismedia → Stash):

Prismedia action	Stash action
videoByURL	sceneByURL
videoByName	sceneByName
videoByFragment	sceneByFragment
performerByURL	performerByURL
performerByName	performerByName
performerByFragment	performerByFragment
movieByURL	movieByURL
movieByName	movieByName
galleryByURL	galleryByURL
galleryByName	galleryByName

Other Prismedia actions (folderCascade, seriesCascade, audioByURL, etc.) are not part of the Stash protocol; declaring them on a stash-compat plugin will fail at runtime.

Auth and headers

Stash YAML supports HTTP headers and per-action driver options. The adapter wires these through transparently. If the scraper needs an API key or cookie, declare it as auth in the Prismedia manifest:

auth:
  - key: SITE_API_KEY
    label: Site API Key
    required: true
    url: https://example.site/account/api

capabilities:
  videoByURL: true

In your YAML, reference the key as you would any Stash scraper variable. The adapter exposes auth values to the YAML driver context.

What the adapter does NOT handle

Stash has features that don't translate cleanly:

• JavaScript-driven scrapers that depend on a Stash-specific JS runtime aren't supported. (Most YAML scrapers are pure HTTP + regex/xpath and work fine.)
• Cascade actions (folderCascade, seriesCascade) — the Stash protocol has no equivalent.
• Per-result tag deduplication policies specific to Stash — the Prismedia normalizer applies its own (case-insensitive for tags).
• Stash UI integrations (custom drawer behavior, in-Stash CSS) — Prismedia uses its own cascade drawer.

If your YAML scraper relies on any of these, it'll need a port to native TypeScript or Python.

Result mapping

Stash result fields are normalized to Prismedia's shapes:

Stash field	Prismedia field (NormalizedVideoResult)
Title	title
Date	date
Details	details
URL / URLs	urls[]
Studio.Name	studioName
Performers[].Name	performerNames[]
Tags[].Name	tagNames[]
Image	imageUrl
Code	code
Director	director

For performers, studios, and tags the adapter only carries names through; nested rich data (performer aliases, performer ethnicity, etc.) is dropped by the simple NormalizedVideoResult shape. If you need the richer data, use the dedicated performer/studio actions which return more detail.

Installing existing Stash scrapers

The Stash community publishes scrapers at github.com/stashapp/CommunityScrapers. To use one:

1. Pick the YAML you want.
2. Build a small directory with a manifest.yml referencing it (as above) and the YAML file.
3. Install it from the Plugins UI just like any other plugin.

Or — for the long tail — install from the Plugins → Stash Community tab, which mirrors the upstream repo and auto-generates the wrapper manifest. One-click install, no manual packaging.

Versioning

The Stash community doesn't strictly version each YAML scraper, so the version field on a stash-compat plugin tracks your wrapper, not upstream. When the upstream YAML changes, bump your wrapper version and republish so the registry's update detection fires.

Debugging

When a stash-compat plugin returns nothing or a bad result, the adapter writes the raw HTTP response body to the worker logs. docker compose logs prismedia | grep stash-adapter is a good first look. Most failures are:

• The site changed its HTML and the regex no longer matches.
• The site requires a session cookie or bot-protection bypass.
• The endpoint moved.

YAML changes are easy: edit the file, restart the worker, retry the identify.

When to drop down to native

Stash-compat is great for "this scraper exists, just use it." Native (TypeScript or Python) is the right move when:

• You need a richer result shape (cascade, candidate disambiguation, image candidates with metadata).
• You want batch support to deduplicate API calls.
• The provider has a real API (not screen-scraping) and you'd rather call JSON.
• You need behavior the YAML protocol doesn't express.

Most first-party Prismedia plugins (TMDB, TVDB, MusicBrainz, YouTube) are native because they call structured APIs and use cascade flows. Most long-tail provider integrations remain stash-compat because the YAML community has already done the work.