Content Studio

Content Studio has two tabs: Board, a Kanban view of every post by status, and Configs, where you set up what gets written and how.

Knowledge

A knowledge entry (internally, a BrandProfile) is scoped to one account and holds a name, a voice/tone description, and a free-text description of facts the model should know. This is what keeps drafts from reading like generic AI copy for every account: it's grounding, not a prompt template.

Topics

Topics are shared across every account. Each one has a name, an angle (the specific take, not just the subject), and an optional live news flag. A live topic pulls in current news via web search before drafting, so the post can reference something that actually happened this week instead of writing in a vacuum.

Configs: the generation recipe

A ContentConfig is what actually runs. It ties together:

• One account: whose voice and posting limits apply.
• One knowledge profile (optional): if you skip it, generation still works, just without grounding.
• One or more topics: rotated by weight, so a topic you've marked more important comes up more often. Weights get multiplied by what the loop has learned about each topic's real performance.
• News keywords (optional): narrows what live-news topics search for.
• A mode: Autopilot posts immediately once a draft passes its checks; Review First holds it in the review queue for a human decision.

What happens when you press Compose

Generation picks a topic (by weight), optionally pulls live news for it, assembles a prompt from the knowledge profile plus recent post history plus the account's own performance data, and asks the model for a draft. If the draft is too similar to something already posted, or breaks the platform's length limit, it's silently retried: you only see the version that passed. If nothing passes after a few tries, the attempt fails rather than shipping something duplicate or broken.