Overview
Context Buckets are a per-domain reference library that feeds the AI your own source material during content creation. This release added rich media ingestion, so a bucket can now hold far more than text: documents, images, web pages, YouTube videos, and uploaded audio or video. Audio and video sources are transcribed automatically and turned into a usable report, so spoken material becomes searchable context.
Steps
- Open a domain and go to the Context Buckets tab.
- Open or create a bucket and add a source: a document (PDF, Word, text, or Markdown), an image, a URL to scrape, a YouTube watch link, or an uploaded audio or video file.
- Wait for the item to move through its status — Pending, then Scraping/Transcribing/Extracting, then Embedding, then Ready.
- Select the Ready items as context when you create content pages or social posts.
Details
- Each YouTube, audio, or video source produces two items: a Transcript and an AI Report (named "{name} - Transcript" and "{name} - Report"). Deleting the transcript also removes its report.
- A bucket whose source was removed can appear as Archived.
- Deleting is permanent — the Delete action removes an item or bucket along with its extracted text, files, and any generated report, with no soft-archive of your own items.
- Limits — roughly 50 MB per document or image, 500 MB per audio or video, and 100 items per bucket.
Check before you move on
- Only Ready items are usable as context, so allow time for transcription and embedding on large media.
- Because deletion is final, double-check before removing a source that produced a transcript and report.