Visual Agent

⬜ Proposed — partly built. Tier 2 — one dependency. The blog thumbnail slice lives in apps/backend/src/ai/agents/visual/ (prompt/thumbnail.ts + generate.ts); template selection, themes, GIF and guard-rail QC are still proposed. ✅ v1 live in apps/studio — reusable invoke API.
Full Marketing-OS kernel merge (tenant knowledge, jobs) still Tier 2.

Back to the agent roster · Marketing OS

What it does

Produces static creatives (image / thumbnail / infographic / educational / carousel) from a blog or brief, with hybrid layout, brand logo compose, moderation and fidelity review.

How to call it (anywhere)

Studio API:

GET /api/agents/visual — capability card
POST /api/agents/visual/invoke — structured generate

Backend helper: ai/agents/visual/invokeVisualAgentRemote when STUDIO_VISUAL_AGENT_URL is set.

Visual Studio UI remains a separate tab that uses the same agent.

Why an agent

Same reason as Compliance: one judgment+generation unit, many workflows — not a UI-only feature.

Roadmap source

Deliverable ① — Static / GIF / Video; guard rails — see roadmap_ref_extracted.md.

Still open

GIF / video end-state
In-process backend port (today: HTTP to studio for sharp/canvas)
Per-tenant logo packs in knowledge layer