mlx-bun library API
    Preparing search index...

    A finished evaluation run, stored beside (never inside) the source tree.

    interface EvalHistoryEntry {
        capability:
            | {
                components: Readonly<Record<string, number>>;
                excluded: readonly string[];
                score: number;
            }
            | null;
        complete: boolean;
        durationMs: number;
        enableThinking: boolean;
        exitCode: number;
        finishedAt: string
        | null;
        id: string;
        label: string;
        machine:
            | { chip: string; host: string; memoryBytes: number; os: string }
            | null;
        model: { path: string; repo: string | null; revision: string | null };
        problems: readonly string[];
        runDirectory: string;
        startedAt: string;
        tasks: readonly TaskSummary[];
    }
    Index
    capability:
        | {
            components: Readonly<Record<string, number>>;
            excluded: readonly string[];
            score: number;
        }
        | null

    The runner's mean over the capability tasks that completed; the rest are named in excluded.

    complete: boolean

    Whether the runner judged every planned task complete and every pin intact; score acceptance stays unreviewed.

    durationMs: number
    enableThinking: boolean
    exitCode: number
    finishedAt: string | null
    id: string
    label: string
    machine: { chip: string; host: string; memoryBytes: number; os: string } | null
    model: { path: string; repo: string | null; revision: string | null }
    problems: readonly string[]
    runDirectory: string

    The run's own directory: result.json (with provenance), samples.jsonl, report.md, the server's log.

    startedAt: string
    tasks: readonly TaskSummary[]