Core concepts

This overview explains how Aspect structures media, how features run, and how to monitor progress.

​

Indexes

Indexes group related assets (e.g., “marketing”, “support”). Each index can specify default_features that run on every new asset (e.g., embedding).

​

Assets

An asset is a video, image, or audio file. When created, the system can immediately queue feature tasks. You can upload from a direct file (multipart) or a URL.

​

Features

Features are AI jobs that produce derived data or representations:

• proxy/audio_proxy/model_proxy: optimized media for streaming and processing
• preview: thumbnails or short previews
• transcription: speech-to-text (video/audio)
• embedding: vector embeddings for semantic search

​

Tasks

Tasks track feature execution for an asset. You can create tasks explicitly or rely on automatic queueing from index defaults. Use the SDK’s tasks.waitForDone to poll until all user-facing features reach a final state.

​

Previews and proxies

Proxies enable smooth playback (e.g., .mp4, .m4a), while previews provide quick visuals. Download URLs are signed and time-limited.

​

Analyze (images)

Ask natural language questions about an image, or locate objects via point/box queries.

​

Webhooks

Receive notifications when assets are created and when tasks progress, so you do not need to poll in production.