tubecreate/tubecli/media-library

Media Library

A shared bag of raw material — images, GIFs and videos — that every extension can draw from. The user gathers material once; extensions point at a collection by id and the machine picks a file.

Why it exists

Each extension used to grow its own private store, so the same picture had to be uploaded several times and fixing it in one place left the other copies stale. One library, many consumers.

Concepts

• Collection: a named bag. It has an id (ASCII, immutable, used as the

folder name) and a display name that can be renamed freely — renaming never touches the folder.

• File: an image, GIF or video inside a collection. Order is stable.
• Picking: random (any file), cycle (walks the whole collection before

repeating, so two consecutive uses do not collide), ai (the caller supplies the filename a model chose).

Endpoints

Prefix /api/v1/media. Page at /media-library.

Using it from another extension

from tubecli.extensions import media_library
for c in media_library.collections():
    print(c["id"], c["name"], c["count"])
path, why = media_library.pick_media("kho_avatar", mode="cycle", kind="image")
if path:
    ...        # dùng file
else:
    log.warning("kho rỗng: %s", why)

Import it lazily and tolerate its absence: an older TubeCLI may not have it, and a missing library should degrade one layer, not break the whole job.

When an agent uses this

Ask for a collection by name, then pick. Pictures come from the bag; text does not — write the words from the actual content, and let the machine draw them with real fonts. That split is the whole point: material is reused, wording is written fresh for each video.

Thumbnail Studio consumes this directly: a store layer declaring "store": "lib:<collection_id>" draws from here instead of its own folder.