eval

type Client struct {
	// BaseURL is the plugin base URL, e.g. "http://localhost:3770".
	BaseURL string
	// HTTP is the underlying client; set to &http.Client{} if nil.
	HTTP *http.Client
	// SourceAccountID is forwarded as X-Nself-Source-Account-Id header.
	SourceAccountID string
}

type EvalRunResult struct {
	// SchemaVer is the schema version of this response; checked against expectedSchemaVer.
	SchemaVer int `json:"schema_ver"`
	// ID is the unique run identifier.
	ID string `json:"id"`
	// SuiteSlug is the evaluated suite's slug.
	SuiteSlug string `json:"suite_slug"`
	// Status is one of: queued, running, passed, failed.
	Status string `json:"status"`
	// PassRate is the fraction of tasks that passed (0.0–1.0).
	PassRate float64 `json:"pass_rate"`
	// SuiteScore is the weighted mean score across all tasks (0.0–1.0).
	SuiteScore float64 `json:"suite_score"`
	// Passed indicates whether the run met the configured threshold.
	Passed bool `json:"passed"`
	// Tasks holds per-task result details.
	Tasks []EvalTaskResult `json:"tasks"`
	// PreconditionFailed is true when a dependency (BGE-M3, gateway) was unavailable.
	PreconditionFailed bool `json:"precondition_failed"`
	// ErrorMessage is populated when Status is "failed" with a human-readable cause.
	ErrorMessage string `json:"error_message,omitempty"`
}

type EvalTaskResult struct {
	// TaskID is the unique task identifier.
	TaskID string `json:"task_id"`
	// Input is the query string used for this task.
	Input string `json:"input"`
	// Output is the system output evaluated.
	Output string `json:"output"`
	// Score is the numeric score (0.0–1.0).
	Score float64 `json:"score"`
	// Passed indicates whether this individual task passed.
	Passed bool `json:"passed"`
	// ScoringMode is one of: exact, semantic, rubric.
	ScoringMode string `json:"scoring_mode"`
	// Rationale is provided for rubric-scored tasks.
	Rationale string `json:"rationale,omitempty"`
}

type GateStatus struct {
	// Tier is the autonomy tier checked: supervised, semi-auto, full-auto.
	Tier string `json:"tier"`
	// Cleared is true when all required suites pass their thresholds.
	Cleared bool `json:"cleared"`
	// BlockingSuites lists suite slugs that are failing or have no runs.
	BlockingSuites []string `json:"blocking_suites"`
	// Enforced mirrors the threshold's enforced flag.
	Enforced bool `json:"enforced"`
}

type RunQueued struct {
	// RunID is the identifier to poll at GET /eval/runs/{id}.
	RunID string `json:"run_id"`
	// Status is always "queued" on the initial response.
	Status string `json:"status"`
}

type RunRequest struct {
	// SuiteSlug is the slug of the suite to run.
	SuiteSlug string `json:"suite_slug"`
}

type ValidationError struct {
	// Field is the JSON-path to the failing field.
	Field string `json:"field"`
	// Message is a human-readable description of the failure.
	Message string `json:"message"`
}

type ValidationResult struct {
	// Valid is true when the submitted YAML passes schema validation.
	Valid bool `json:"valid"`
	// Errors lists field-level validation errors.
	Errors []ValidationError `json:"errors,omitempty"`
}