Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
import type { LlmProviderId } from '@clerum/llm-providers'
// ─── WorkflowRecipe Phase (13-state machine) ────────────────────────────
/**
* All 13 phases in the WorkflowRecipe lifecycle.
* Source of truth: WORKLOADRECIPE-SPEC.md §12.1
*
* Terminal states: "deprecated", "rollback-failed"
*/
export type RecipePhase =
| 'candidate'
| 'pending-approval'
| 'approved'
| 'pending'
| 'pending-operator-input'
| 'deploying'
| 'testing'
| 'active'
| 'degraded'
| 'rolling-back'
| 'failed'
| 'deprecated'
| 'rollback-failed'
/** Type guard for RecipePhase */
export function isRecipePhase(value: string): value is RecipePhase {
const phases: RecipePhase[] = [
'candidate',
'pending-approval',
'approved',
'pending',
'pending-operator-input',
'deploying',
'testing',
'active',
'degraded',
'rolling-back',
'failed',
'deprecated',
'rollback-failed',
]
return phases.includes(value as RecipePhase)
}
// ─── Workload Types ─────────────────────────────────────────────────────
/**
* Supported workload types.
* Source of truth: WORKLOADRECIPE-SPEC.md §3.1
*/
export type WorkloadType = 'deployment' | 'statefulset' | 'cronjob' | 'job' | 'daemonset'
/** Type guard for WorkloadType */
export function isWorkloadType(value: string): value is WorkloadType {
return ['deployment', 'statefulset', 'cronjob', 'job', 'daemonset'].includes(value)
}
// ─── Resource Types ─────────────────────────────────────────────────────
export type ResourceType = 'pvc' | 'secret' | 'configmap'
// ─── Security ───────────────────────────────────────────────────────────
/**
* Security isolation levels.
* Source of truth: WORKLOADRECIPE-SPEC.md §8
*/
export type SecurityIsolationLevel = 'minimal' | 'standard' | 'strict'
/**
* Per-workload security overrides.
* Allows specific workloads to run as a non-root UID while maintaining
* the recipe-level isolation model.
* Source of truth: GAP 15 (PHASE-9-PRE-PRODUCTION-HARDENING.md)
*/
export interface WorkloadSecurityOverrides {
runAsUser?: number
runAsGroup?: number
fsGroup?: number
addCapabilities?: string[]
/** Opt-in root init container for non-distroless images and storage backends that do not honor fsGroup. */
prepareVolumeOwnership?: boolean
}
// ─── Secret-Backed Environment Variables ─────────────────────────────────
/**
* Mapping from a Secret key to an environment variable name.
* Source of truth: WORKLOADRECIPE-SPEC.md §3.4.1
*/
export interface EnvSecretMapping {
secretKey: string
envVar: string
optional?: boolean
}
/**
* Batch secret mapping configuration.
* References a Kubernetes Secret and maps multiple keys to environment variables.
* Source of truth: WORKLOADRECIPE-SPEC.md §3.4.1
*/
export interface EnvSecretConfig {
name: string
keys: EnvSecretMapping[]
}
// ─── Workload Definitions (Discriminated Union) ─────────────────────────
export interface BaseWorkloadDef {
id: string
type: WorkloadType
image: string
imagePullPolicy?: 'Always' | 'IfNotPresent' | 'Never'
port?: number
replicas?: number
command?: string[]
args?: string[]
env?: Array<{ name: string; value?: string; valueFrom?: unknown }>
envSecret?: EnvSecretConfig
volumeMounts?: VolumeMount[]
resources?: ResourceRequirements
healthCheck?: HealthCheck
dependsOn?: string[]
includeWhen?: string
transport?: TransportConfig
egressBindings?: EgressBindingDef[]
ownerReferences?: false // Set to false to skip ownerReferences (e.g., for cross-namespace resources)
imagePullSecrets?: string[] // Docker registry secrets for private images
security?: WorkloadSecurityOverrides // Per-workload security overrides (GAP 15)
/**
* Path B — oauthClient ids (from spec.oauthClients[], must be backgroundAccess)
* this workload uses. Opting in mounts the recipe OAuth broker token and
* grants egress to the control-api broker route. Spec §9.2.
*/
oauthClientRefs?: string[]
}
export interface DeploymentDef extends BaseWorkloadDef {
type: 'deployment'
}
export interface StatefulSetDef extends BaseWorkloadDef {
type: 'statefulset'
serviceName?: string
volumeClaimTemplates?: VolumeClaimTemplate[]
}
export interface CronJobDef extends BaseWorkloadDef {
type: 'cronjob'
schedule: string
timeZone?: string
}
export interface JobDef extends BaseWorkloadDef {
type: 'job'
backoffLimit?: number
}
export interface DaemonSetDef extends BaseWorkloadDef {
type: 'daemonset'
}
export type WorkloadDef = DeploymentDef | StatefulSetDef | CronJobDef | JobDef | DaemonSetDef
// ─── Resource Definitions (Discriminated Union) ─────────────────────────
export interface PvcResourceDef {
id: string
type: 'pvc'
storageClass?: string
size: string
accessMode?: string
}
export interface SecretResourceDef {
id: string
type: 'secret'
data?: Record<string, string>
generateKeys?: string[]
}
export interface ConfigMapResourceDef {
id: string
type: 'configmap'
data: Record<string, string>
}
export type ResourceDef = PvcResourceDef | SecretResourceDef | ConfigMapResourceDef
// ─── Binding Definition ─────────────────────────────────────────────────
export interface BindingDef {
from: string
to: string
port: number
protocol?: 'TCP' | 'UDP'
}
// ─── Egress Binding Definition ──────────────────────────────────────────
export interface EgressBindingDef {
egressClass?: 'exact-host' | 'public-web'
dns?: string
port?: number
protocol?: 'TCP' | 'UDP'
}
// ─── Supporting Types ───────────────────────────────────────────────────
export interface TransportConfig {
type: 'sse' | 'streamableHttp' | 'stdio'
path?: string
}
export interface VolumeMount {
name: string
mountPath: string
subPath?: string
readOnly?: boolean
}
/**
* VolumeClaimTemplate for StatefulSet workloads.
* Defines PVC templates that are created per replica.
* Source: Kubernetes StatefulSet volumeClaimTemplates
*/
export interface VolumeClaimTemplate {
name: string
storageClass: string
accessMode: 'ReadWriteOnce' | 'ReadWriteMany' | 'ReadOnlyMany'
size: string
}
export interface ResourceRequirements {
requests?: { cpu?: string; memory?: string }
limits?: { cpu?: string; memory?: string }
}
export interface HealthCheck {
type: 'http' | 'tcp' | 'exec'
path?: string
port?: number
command?: string[]
initialDelaySeconds?: number
periodSeconds?: number
timeoutSeconds?: number
failureThreshold?: number
}
export interface ComputedValue {
name: string
expression: string
}
export interface RecipeDependency {
name: string
namespace?: string
cascadeRollback?: boolean
maxWaitMinutes?: number
}
// ─── Webhooks (W1.1) ────────────────────────────────────────────────────
/**
* Signature-verification scheme for an inbound webhook request.
* One per webhook entry. Spec §7.
*/
export type WebhookVerificationScheme =
| 'hmac-sha256-body'
| 'hmac-sha256-timestamp-body'
| 'jwt-bearer-jwks'
| 'static-bearer'
/** Type guard for WebhookVerificationScheme. */
export function isWebhookVerificationScheme(value: string): value is WebhookVerificationScheme {
return [
'hmac-sha256-body',
'hmac-sha256-timestamp-body',
'jwt-bearer-jwks',
'static-bearer',
].includes(value)
}
/** Reference to a Secret in the recipe's sandbox-recipes namespace. */
export interface WebhookSecretRef {
name: string
key: string
}
/**
* Verification configuration. Field requirements are scheme-dependent
* (W7/W9) and enforced by the CRD CEL rules at admission and again by
* the WRC reconciler at reconcile time. The TypeScript shape uses
* optionals because the CRD does — we don't pre-narrow here.
*/
export interface WebhookVerification {
scheme: WebhookVerificationScheme
secretRef?: WebhookSecretRef
signatureHeader?: string
signaturePrefix?: string
signatureEncoding?: 'hex' | 'base64'
/**
* static-bearer only — HTTP header name to read the token from.
* Defaults to `Authorization` when omitted. Custom headers
* (e.g. `X-Telegram-Bot-Api-Secret-Token`) let providers that
* authenticate via a non-Authorization header fit the scheme.
*/
tokenHeader?: string
/**
* static-bearer only — string to strip from the front of the header
* value before constant-time comparison. Defaults to `Bearer `
* (with trailing space). Author-supplied empty string means "no
* prefix" — the entire header value IS the token (Telegram style).
*/
tokenPrefix?: string
jwksUrl?: string
issuer?: string
audience?: string
setupHandshake?: WebhookSetupHandshake
}
/**
* Known-shape setup-handshake strategies. Each strategy has an exact request
* shape that bypasses body-signature verification; anything else falls through
* to the normal verifier. See spec §7.5.
*/
export type WebhookSetupHandshakeStrategy =
| 'meta-hub-challenge'
| 'slack-url-verification'
| 'stripe-verify'
/** Type guard for WebhookSetupHandshakeStrategy. */
export function isWebhookSetupHandshakeStrategy(
value: string
): value is WebhookSetupHandshakeStrategy {
return ['meta-hub-challenge', 'slack-url-verification', 'stripe-verify'].includes(value)
}
/**
* Setup-handshake config (opt-in per webhook). The strategy's matching request
* shape is the only path that bypasses body-signature verification. Some
* strategies (currently `meta-hub-challenge`) carry their own credential via
* `secretRef` — the strategy's verify-token equivalent.
*/
export interface WebhookSetupHandshake {
strategy: WebhookSetupHandshakeStrategy
secretRef?: WebhookSecretRef
}
/** Replay-window config; required iff scheme is `hmac-sha256-timestamp-body` (W8). */
export interface WebhookReplay {
timestampHeader: string
toleranceSec: number
}
/**
* Browser CORS policy for a webhook. Consumed by the front-door
* webhook-proxy via control-api's registry endpoint; the per-recipe
* gateway never sees this — CORS is enforced before forwarding.
*/
export interface WebhookCors {
/** Exact origin strings (scheme + host + optional port, no trailing slash). */
allowedOrigins?: string[]
}
/** A single webhook entry under `spec.webhooks[]`. */
export interface WebhookDef {
id: string
workloadRef: string
path: string
methods?: Array<'POST' | 'GET'>
maxBodyBytes?: number
verification: WebhookVerification
replay?: WebhookReplay
/**
* When true, a missing or empty secretRef does NOT mark the recipe
* degraded. The webhook is kept dormant by the gateway (responds with
* 410 Gone + X-Clerum-Webhook-State: dormant) until the Secret/key is
* created; the Secret watcher then triggers a reconcile that transitions
* the entry to active. Default false (today's fail-closed behavior).
*/
optional?: boolean
cors?: WebhookCors
}
// ─── Sandbox UI ─────────────────────────────────────────────────────────
export interface SandboxUiInternalEgress {
workloadRef: string
port: number
}
/**
* External egress entry — a DNS hostname (FQDN) which the WRC reconciler
* resolves to one or more ipBlock rules at reconcile time and re-resolves
* on a periodic tick. Static CIDR is intentionally not accepted: raw IPs
* are unreviewable in PRs, freeze across vendor IP rotations, and can
* silently authorize cloud-metadata or sibling-tenant ranges that the
* RFC1918 carve-out doesn't trim (e.g. 169.254/16).
*/
export interface SandboxUiExternalEgress {
fqdn: string
port: number
reason?: string
}
export interface SandboxUiEgress {
internal?: SandboxUiInternalEgress[]
external?: SandboxUiExternalEgress[]
}
export interface SandboxUiSpec {
workloadRef: string
port: number
title?: string
icon?: string
defaultPath?: string
egress?: SandboxUiEgress
}
// ─── Recipe OAuth (spec.oauthClients[]) ────────────────────────────────
// Two consumer shapes: the sandbox UI embed (foreground end-user OAuth) and
// background Path B workloads (backgroundAccess: true, referenced from
// workloads[].oauthClientRefs on non-MCP workloads). Either consumer alone is
// enough to justify an oauthClient — spec.ui is not required.
/**
* Known-shape OAuth provider adapters baked into control-api. Adding a
* provider is a control-api change — recipes cannot speak OAuth to
* arbitrary endpoints. Spec §9.9 / Decision 20.
*
* R1-L2: this is the full control-api ADAPTER SET (shared with mcp-servers),
* NOT the recipe-admissible set. monday/clickup/vercel (U2, spec 06) are
* mcpserver-only: the WorkflowRecipe CRD `provider` enum deliberately stays at
* the five legacy providers, so a recipe declaring one of the three is rejected
* at admission (fail-closed). Do NOT treat membership in this union / a true
* `isOAuthProvider` result as "valid for a recipe" — the recipe CRD enum is the
* authoritative gate for recipe OAuth clients.
*/
export type OAuthProvider =
| 'salesforce'
| 'slack'
| 'notion'
| 'microsoft-graph'
| 'google'
| 'monday'
| 'clickup'
| 'vercel'
export function isOAuthProvider(value: string): value is OAuthProvider {
return [
'salesforce',
'slack',
'notion',
'microsoft-graph',
'google',
'monday',
'clickup',
'vercel',
].includes(value)
}
/** Reference to a Secret in the recipe's sandbox-recipes namespace. */
export interface OAuthSecretRef {
name: string
key: string
}
/**
* One OAuth provider the recipe needs to act-on-behalf-of the user. Tokens
* are minted by control-api after the user completes the auth-code flow on
* the platform-rooted callback URL, then delivered into the embed via the
* `embed.onOauthCompleted` IPC. Refresh tokens stay platform-side; access
* tokens are fetched at request time via the `/sandbox-ui/oauth/token`
* helper.
*/
export interface OAuthClientDef {
id: string
provider: OAuthProvider
clientIdRef: OAuthSecretRef
clientSecretRef: OAuthSecretRef
scopes?: string[]
/**
* When true, an operator may connect this client as a recipe-owned `service`
* grant so background workloads can obtain provider tokens via the broker.
* Path B, spec §7.
*/
backgroundAccess?: boolean
}
// ─── Plugin Workload SDK ────────────────────────────────────────────────
/**
* Capability declaration for the Plugin Workload SDK (spec §9, §13).
* Recipes opt plugin workloads into two controlled side-effect channels:
*
* - promptBridge: one-shot LLM call routed through the recipe's mcp-host.
* - clientNotifications: notification intent authorized by control-api,
* delivered by the Notification Service.
*
* The reconciler validates this block (see pluginWorkloadSdkValidator.ts)
* and persists the validated capability to status.pluginWorkloadSdk so
* control-api can read it at runtime. Wildcard `*` entries are rejected at
* the CRD level — only explicit admin grants may use wildcards.
*/
export interface PluginWorkloadSdkSpec {
promptBridge?: {
allowedModels?: string[]
/** Per-grant output ceiling. Enforced as `max_tokens` on API-key providers; codex-subscription cannot bind it on the ChatGPT wire and is bounded there by the contract limit of 16384 instead. */
maxOutputTokens?: number
/** @deprecated Accepted for compatibility; no longer enforced (issue #348). Platform per-minute rate limits apply. */
maxRequestsPerRun?: number
/** Max total bytes across messages[].content. */
maxRequestContentBytes?: number
/** Max parallel promptBridge calls per workload; default 5. */
maxConcurrentInvocations?: number
/** Rate limit per workload; platform default 600 when unset (ENV CONTROL_API_PLUGIN_SDK_PROMPTBRIDGE_PER_MIN, issue #348). */
maxInvocationsPerMinute?: number
/** How long invocation results are queryable; default 24. */
resultRetentionHours?: number
/** Max attachments per request; default 0 (v1: no attachments). */
maxAttachments?: number
/** Max total bytes across attachments; default 0. */
maxAttachmentBytes?: number
}
clientNotifications?: {
allowedEventTypes: string[]
allowedTargetRefs?: string[]
allowedUserRefs?: boolean
/** @deprecated Accepted for compatibility; no longer enforced (issue #348). Platform per-minute rate limits apply. */
maxNotificationsPerRun?: number
/** Rate limit per workload; platform default 750 when unset (ENV CONTROL_API_PLUGIN_SDK_NOTIFICATIONS_PER_MIN, issue #348). */
maxNotificationsPerMinute?: number
/** Default 256. */
maxTitleBytes?: number
/** Default 4096. */
maxBodyBytes?: number
/** How long notification records are queryable; default 72. */
resultRetentionHours?: number
}
/** Workload ids permitted to call; empty = all declared workloads. */
allowedCallers?: string[]
/** Regex for idempotency key format validation; default '^[a-zA-Z0-9_-]{1,128}$'. */
idempotencyKeyPattern?: string
}
/** Snippet-scoped subset of the SDK capability (steps[].run.capabilities). */
export interface PluginWorkloadSdkSnippetConfig {
promptBridge?: {
allowedModels?: string[]
/** Per-grant output ceiling. Enforced as `max_tokens` on API-key providers; codex-subscription cannot bind it on the ChatGPT wire and is bounded there by the contract limit of 16384 instead. */
maxOutputTokens?: number
}
clientNotifications?: {
allowedEventTypes: string[]
allowedTargetRefs?: string[]
allowedUserRefs?: boolean
}
}
/**
* Validated capability projection persisted to status.pluginWorkloadSdk.
* control-api reads this at runtime to know which SDK families a recipe
* actually has active (vs merely declared in spec).
*/
export interface PluginWorkloadSdkCapabilityStatus {
state: 'validated' | 'disabled' | 'degraded' | 'awaiting_policy'
promptBridge: boolean
clientNotifications: boolean
message?: string
// These fields are merge-patch clearable. A non-validated projection must
// explicitly send null so a previous validated record cannot survive a
// later degraded/awaiting-policy transition.
validatedAt?: string | null
bootstrapContractVersion?: 2 | 3 | null
bootstrapPodUid?: string | null
bootstrapProvider?: string | null
bootstrapModel?: string | null
policyRevision?: number | null
policyHash?: string | null
defaultTargetRef?: string | null
verifiedAt?: string | null
}
// ─── WorkflowRecipe CRD ─────────────────────────────────────────────────
export type WorkflowRecipeGfsScope =
| 'gfs.read'
| 'gfs.write'
| 'gfs.delete'
| 'gfs.manage_acl'
| 'gfs.share'
export interface WorkflowRecipeGfsIntentSpec {
publishTargets?: Array<{ drive: string; target: string }>
mounts?: Array<{ drive: string; target: string; scopes: WorkflowRecipeGfsScope[] }>
}
export interface WorkflowRecipeSpec {
description?: string
contextRef?: string
workloads?: WorkloadDef[]
webhooks?: WebhookDef[]
resources?: ResourceDef[]
bindings?: BindingDef[]
inputs?: Record<string, unknown>
inputContract?: Record<string, unknown>
activeProfile?: string
profiles?: Record<string, Record<string, unknown>>
computed?: ComputedValue[]
security?: {
isolationLevel?: SecurityIsolationLevel
}
runtimeEgress?: WorkflowRuntimeEgressSpec
dependencies?: RecipeDependency[]
dryRun?: boolean
ui?: SandboxUiSpec
oauthClients?: OAuthClientDef[]
pluginWorkloadSdk?: PluginWorkloadSdkSpec
gfs?: WorkflowRecipeGfsIntentSpec
// ─── Workflow Fields (Stage 1) ─────────────────────────────────────
coordinatorImage?: string
agent?: {
model: string
provider: LlmProviderId
/**
* Optional static-credential Secret. Omitted for oauth-broker providers
* such as `codex-subscription`; Control API remains the OAuth custodian.
*/
secretRef?: { name: string; namespace?: string }
soulRef?: {
storageRef: {
bucket: string
key: string
provider?: 's3' | 'gcs' | 'spaces' | 'minio'
region?: string
endpoint?: string
}
}
}
steps?: Array<{
id: string
instruction?: string
run?: WorkflowSnippetRunSpec
dependsOn?: string[]
timeoutSeconds?: number
backoffSeconds?: number
maxRetries?: number
agent?: {
model?: string
provider?: LlmProviderId
soul?: string
}
mcpServers?: string[]
allowedTools?: { include?: string[] }
toolChoice?: 'auto' | 'none' | 'required'
maxIterations?: number
requiresApproval?: {
target: { userId?: string; teamId?: string }
message: string
timeoutSeconds?: number
}
}>
mcpServers?: Array<{
id: string
endpoint: string
transport?: 'http' | 'stdio'
}>
output?: {
destination?: 'configmap' | 'secret' | 'stdout' | 'pvc'
name?: string
namespace?: string
claimName?: string
format?: 'pdf' | 'xlsx' | 'json' | 'text' | 'html' | 'multi'
storageSize?: string
}
// ─── Scheduling Fields (Stage 4 / legacy) ─────────────────────────
scheduling?: {
cron: string
timezone?: string
concurrencyPolicy?: 'Forbid' | 'Replace' | 'Allow'
successfulHistoryLimit?: number
failedHistoryLimit?: number
suspend?: boolean
}
runRetention?: {
maxRunDurationSeconds?: number
ttlSecondsAfterFinished?: number
successfulHistoryLimit?: number
failedHistoryLimit?: number
}
// ─── Triggers (Phase 5 — DB-first) ────────────────────────────────
// Plan §Fase 5: WRC reads spec.triggers.schedule and keeps the row
// in workflow_schedules in sync. Both shapes coexist in the CRD
// schema (R5–R8); reconciler prefers triggers.schedule, falls back
// to legacy spec.scheduling.
triggers?: {
onDemand?: {
allowedActors?: Array<'user' | 'autonomous' | 'scheduled'>
}
schedule?: {
cron: string
timezone?: string
concurrencyPolicy?: 'Forbid' | 'Replace' | 'Allow'
suspend?: boolean
}
}
}
export interface WorkflowSnippetRunSpec {
type: 'snippet'
language: 'typescript'
code: string
capabilities?: WorkflowSnippetCapabilities
}
export interface WorkflowRuntimeEgressSpec {
http?: {
egressClass?: 'exact-host' | 'public-web'
allowedHosts?: string[]
}
}
export interface WorkflowSnippetCapabilities {
http?: {
egressClass?: 'exact-host' | 'public-web'
allowedHosts?: string[]
}
secrets?: Array<{
alias: string
secretRef: {
name: string
key: string
}
}>
mongo?: {
workloads?: string[]
access?: 'read' | 'readWrite'
}
postgres?: {
workloads?: string[]
access?: 'read' | 'readWrite'
}
mcp?: {
servers?: string[]
allowedTools?: {
include?: string[]
}
}
artifacts?: {
maxCount?: number
}
pluginWorkloadSdk?: PluginWorkloadSdkSnippetConfig
}
export interface WorkloadStatus {
id: string
type: WorkloadType
phase: string
message?: string
ready?: boolean
}
export interface WorkflowRecipeStatus {
phase: RecipePhase
message?: string
lastTransitionTime?: string
workloads?: WorkloadStatus[]
conditions?: StatusCondition[]
// ─── Workload Instance Mapping (UUID naming) ────────────────────────
/** Mapping of workload ID → K8s resource name (UUID-based). Persisted on first deploy. */
workloadInstances?: Record<string, string>
// ─── Resource Instance Mapping (UUID naming) ────────────────────────
/**
* Mapping of spec.resources[] ID (PVC/Secret/ConfigMap) → recipe-scoped K8s
* resource name. Persisted on first deploy, mirroring workloadInstances, so
* generic resource ids (e.g. "data", "credentials") cannot collide across
* recipes in a shared namespace. See issue #571.
*/
resourceInstances?: Record<string, string>
// ─── Plugin Workload SDK ────────────────────────────────────────────
/** Validated SDK capability projection — read by control-api at runtime. */
pluginWorkloadSdk?: PluginWorkloadSdkCapabilityStatus
// ─── Workflow Status (Stage 1) ─────────────────────────────────────
workflowExecution?: {
phase: string
startedAt?: string
completedAt?: string
currentStep?: string
attempt?: number
message?: string
}
steps?: Array<{
id: string
phase: string
startedAt?: string
completedAt?: string
output?: string
error?: string
executor?: 'agentic' | 'snippet' | 'custom'
modelUsed?: string
approvalBindingSha256?: string
toolsCalled?: Array<{
serverName: string
toolName: string
args?: Record<string, unknown>
result?: unknown
durationMs?: number
}>
toolsCalledTruncated?: boolean
}>
}
export interface StatusCondition {
type: string
status: 'True' | 'False' | 'Unknown'
lastTransitionTime: string
reason?: string
message?: string
}
export interface WorkflowRecipeCRD {
apiVersion: 'clerum.io/v1alpha1'
kind: 'WorkflowRecipe'
metadata: {
name: string
namespace: string
uid?: string
resourceVersion?: string
generation?: number
creationTimestamp?: string
annotations?: Record<string, string>
labels?: Record<string, string>
finalizers?: string[]
deletionTimestamp?: string
ownerReferences?: Array<{
apiVersion?: string
kind?: string
name?: string
uid?: string
controller?: boolean
blockOwnerDeletion?: boolean
}>
}
spec: WorkflowRecipeSpec
status?: WorkflowRecipeStatus
}
// ─── WorkflowRecipePolicy CRD (Phase 7) ────────────────────────────────
export interface WorkflowRecipePolicySpec {
description?: string
governance?: {
requireApproval?: boolean
maxWorkloadsPerRecipe?: number
maxReplicasPerWorkload?: number
allowedWorkloadTypes?: WorkloadType[]
requiredSecurityLevel?: SecurityIsolationLevel
}
detection?: {
enabled?: boolean
scanOnDeploy?: boolean
imageAllowlist?: string[]
imageDenylist?: string[]
allowedStorageClasses?: string[]
maxResourceLimits?: { cpu?: string; memory?: string }
}
publication?: {
registry?: string
autoPublish?: boolean
registries?: Array<{
name: string
url: string
type: 'public' | 'private'
ociBackend?: 'docr' | 'ghcr' | 'generic-oci'
default?: boolean
credentialsSecret?: string
}>
}
deployment?: {
autoPromoteAfterSeconds?: number
rollbackOnFailure?: boolean
maxConcurrentDeploys?: number
}
notification?: {
channels?: string[]
events?: ('deploy' | 'rollback' | 'failed' | 'deprecated')[]
}
deprecation?: {
gracePeriodDays?: number
autoDeprecateInactiveDays?: number
}
}
export interface WorkflowRecipePolicyCRD {
apiVersion: 'clerum.io/v1alpha1'
kind: 'WorkflowRecipePolicy'
metadata: {
name: string
namespace: string
}
spec: WorkflowRecipePolicySpec
}