Bounded Campaigns page aggregate for the signed-in workspace

Pure read: campaign list with latest run and counters, the selected campaign's summary (pick-up rate, disposition mix, spend) and one contact page, and the wizard reference data (agents, numbers, contact lists, retry presets, default windows, plan concurrency, and launch_gates: the workspace's record state, voice channel, destination allowlist, the sanctioned-country and blocked-purpose sets, the consent bases, whether every launch is reviewed, and the effective velocity caps -- what the wizard's review step previews a launch against; launch remains the authority). With preview_agent_id (an AI agent in the session workspace) the page also carries agent_preview: what that AI agent will say, rendered by the same preview the Contacts list detail serves (its DTO unchanged) for the first contact of preview_list_id by import order in the wizard's preview_language (a contact's own language still wins, as on the call), or state no_contact without a list and no_agent for an AI agent outside the workspace. The wizard sends one such request per AI agent selection its agent step shows. agent_preview is null when no AI agent is named. The same request carries agent_opening, how that AI agent's calls open, read from the publication a call freezes (its own settings while unpublished): first_speaker (callee or agent), silence_wait_seconds, kind (literal, generated or workflow), line (the written opening for literal, else empty), published, language, and waits_without_line, true when the AI agent waits for the callee and has no opening line, so a call where the person also waits can stay silent. A campaign has no opening of its own. agent_opening is null when no AI agent is named or it is outside the workspace. The selected campaign's summary carries the latest run's failure_breakdown, over the calls its dispatcher placed: failure_breakdown counts the labelled calls (classified_calls), those with a failure (failed_calls), and the failures by layer (telephony, media, agent; only layers with a failure appear) and within each layer by category, most frequent first; a breakdown with no failure has an empty layers list. One grouped read; labels are never decided on this read. Fixed query budget; never a caller-selected tenant.