Screens Reference
MAESTRO CARE ships a screen per lifecycle stage. The three you’ll use most are the chat surface, the Library, and the Evolution screen.
Primary screens
Section titled “Primary screens”| Screen | How to reach | Purpose |
|---|---|---|
| Chat | boot | Natural-language input, mode toggle, slash palette, artifact pill, header Library link, Production toolbar. |
| Chain DAG | Chat → Read full | Full box-and-arrow graph of a freshly-generated chain + clickable step list + CARL detail pane (see The chain DAG). |
| Artifacts | /artifacts | Current-chat artifacts (chain / stage / tool / dataset / synth output); save, copy, inspect, JSON/DAG toggle (see below). |
| Library | /library · header link | Saved chains — sortable columns, filter, tag pool, recency strip, mean cost, bulk import/export (see below). |
| Inspection | Library Enter | Saved-chain detail + run history + Integration pane. |
| Edit Agent | Library e | Manual edit — name / description / tags / task / change summary, save-as-new-version or promote-to-stable (see below). |
| Execution | Library r | Live CARL run + token streaming. |
| Evolution | /evolution · Library v/E | Run + watch a GA over a chain (see The Evolution screen). |
| Evolution Dashboard | /evolution | List of active + recent runs; Enter opens a run, c compares two. |
| Replay | Runs Enter | Step through a saved ReasoningResult. |
| Runs | /runs | Local run history; Enter opens the Replay sidecar. |
| Catalog | Ctrl+K | Browse installed capabilities (skills / MCP / tools / cards). |
| Cost Dashboard | /cost | Token + USD spend rollup by provider / chain / session. |
| Marketplace | /marketplace | Search shared agent_skill listings on Memory. |
| Logs | /logs | Tail the rolling app log; m toggles a module filter. |
| Sandbox Trust | /sandbox | Audit + revoke trusted AgentSkills (SHA-pinned trust store). |
| Profile | /profile | List credential profiles under ~/.config/care/profiles/. |
| Settings | /settings | Edit MAGE / Memory / Platform creds + theme + advanced knobs. |
| Help | /help · ? | Tutorial + every binding (filtered by active screen). |
| Welcome | boot | Boot splash; routes to Chat (returning) or Settings (first-run). |
| Task List | Ctrl+B | In-flight workers panel. |
Modals
Section titled “Modals”| Modal | Triggered from | Purpose |
|---|---|---|
| Command Palette | Ctrl+P | Fuzzy palette over commands + saved entities. |
| Confirm | destructive actions | OK / Cancel for bulk delete, accept-winner, … |
| Conflict | save name collision | Resolve a name collision on save. |
| Diff | Library D | Side-by-side compare two chains / individual vs parent. |
| Lineage | Library l | Walk a chain’s ancestry DAG. |
| Import / Export | Library i / x | Import a chain bundle / export entities to a tarball. |
| Export Chain | Evolution x | Export a single chain to disk — pick Markdown / JSON / Python with a folder Browse picker. |
| Evolution Launch | Library v / E | Budget / rubric / objectives picker before evolving. |
| Evolution Compare | Dashboard c | Side-by-side fitness curves for two runs. |
| Human Input | CARL human-input step | Block execution for a human-supplied answer. |
| Resume | /resume | Rehydrate a Production-mode transcript. |
| Run Context | Library / Execution r | Re-run form: task + context-file picker + tags. |
| Save Agent | post-generation | Tag + name a freshly-generated chain. |
| Save Report | after save-all | Post-mortem table of save-all outcomes. |
| Tag Editor | Library T | Edit tags (bulk) + optional title. |
| Use It Now | post-save | Copy-paste recipe (python / curl / cli) for the saved chain. |
The Library screen
Section titled “The Library screen”/library — or the Library link in the Chat header — lists every saved chain. Click
a column header to sort: Name, Last Run, and Runs (run count) are sortable;
the sort persists between visits, and favourites stay pinned to the top. Each row carries
a recency strip (last run + a success-rate/N badge) and a mean cost column drawn
from local run history. Press e to edit a row inline, r to re-run it, Enter to
inspect it; i/x import or export bundles.
The Edit Agent screen
Section titled “The Edit Agent screen”Library → e opens the manual editing screen for a saved chain. You can change the
display name, description, tags, task description, and a change
summary; a Content tab also lets you hand-edit the CARL JSON. Validation runs on every
keystroke. The footer offers Save (Ctrl+S — writes a new version),
Promote (advances the chain into the stable channel), and Back (Esc). The
Save button shows how many fields changed, and a structural edit auto-fills a change
summary so the save never blocks silently.
The Artifacts screen
Section titled “The Artifacts screen”/artifacts is a two-pane browser over everything the current chat session produced
(chains, stage payloads, tool output, datasets, synth answers). The header shows an
artifacts pill — [ N · M unsaved ] — that stays in sync with the chat surface’s
pill: saving a chain (s) decrements the unsaved count, dropping one (d) shrinks the
total. v toggles the detail pane between raw JSON and the DAG step graph;
Enter inspects a chain, c copies its payload, S saves all unsaved, D diffs two
selected rows.
The genetic-algorithm view and the full chain DAG each have their own page:
- The Evolution screen — live fitness chart, Pareto front, versions, and the Stop / Accept-winner controls.
- The chain DAG — the Read full box-and-arrow graph with clickable steps and a CARL detail pane.