Resources / Agent

Agent

An always-on agent: a language model that pursues the goal you write, observes this project through the API and acts through it with its own identity, which can do exactly what its role allows. Every step is recorded durably and every action is in the audit log under the agent's name. Text it reads is treated as data, never as instructions.

Overview

APIintelligence.etorm.com/v1alpha1
Resourceagents (cluster-scoped inside its cluster)
Short namesagent
Available inproject clusters only
Statuswritten by a controller through /status

Minimal example

apiVersion: intelligence.etorm.com/v1alpha1
kind: Agent
metadata:
  name: example
spec:
  goal: "<goal>"
  endpointRef:
    name: "<name>"
  roleRef:
    name: "<name>"

Spec

FieldTypeRequiredDefaultAllowedDescription
goalstringyeslength 1 to 4096What the agent should achieve, in your words.
instructionsstringup to 16384 charactersHow to go about it: constraints, preferences, what to report.
endpointRefreference to InferenceEndpointyesThe chat InferenceEndpoint (the model) the agent thinks with.
roleRefobjectyesThe Role the agent holds in this project: its tools. It may not grant access control, billing, secret values, status or wildcard writes. Setting it needs the right to bind this role.
roleRef.namestringyeslength 1 to 63Name of a Role in this project or a platform role.
statestring (one of)"Running"Running, Paused, StoppedRunning works; Paused keeps the agent's place and calls nothing; Stopped ends it at once (the kill switch).
extraClusterslist of objectup to 6 itemsOther clusters the agent may read (never write). Only for agents of the platform's own agents project, and only the cluster platform and the projects of its swarm's change repositories.
extraClusters[].namestringyeslength 1 to 63The cluster (only platform).
extraClusters[].roleRefobjectyesA Role that grants read verbs only; the agent holds it in that cluster.
extraClusters[].roleRef.namestringyeslength 1 to 63Name of the Role.
triggersobject{}When the agent works.
triggers.modestring (one of)"AlwaysOn"AlwaysOn, Schedule, EventsAlwaysOn loops and waits intervalSeconds when it has nothing to do; Schedule works at the cron times; Events works when a watched object changes.
triggers.intervalSecondsinteger30030 to 86400How long the agent waits after it has nothing more to do.
triggers.cronlist of stringup to 8 itemsCron expressions in UTC, for Schedule.
triggers.watcheslist of objectup to 16 itemsObjects of this project whose changes wake the agent, for Events.
triggers.watches[].groupstringyesup to 253 charactersAPI group of the kind.
triggers.watches[].resourcestringyesup to 63 charactersPlural resource name, for example vpcs.
triggers.watches[].namestringup to 63 charactersOne object; all of the kind when empty.
triggers.watches[].clusterstringup to 63 charactersThe cluster of the objects; the agent's own project when empty (otherwise one of its extraClusters).
budgetobject{}Limits the agent stays within.
budget.tokensPerDayinteger2000001 to 10000000Model tokens (input and output) per UTC day.
budget.cpuSecondsPerDayinteger6001 to 86400CPU seconds per UTC day spent on the agent's behalf.
budget.actionsPerHourinteger201 to 600API calls per hour.
budget.stepsPerRuninteger501 to 1000Steps before the agent rests, as if it had nothing more to do.
budget.spendCapMicrosinteger0at least 0Money the agent may spend, in millionths of the account currency. 0 unless a billing owner sets it.
memoryRefslist of Memoryup to 8 itemsMemories the agent may recall from and write notes to (its role must also allow memories/records).
memoryRefs[].namestringyeslength 1 to 63Name of the Memory in this project.
approvalobject{}Which actions need a person's approval before they run.
approval.requiredFromstring (one of)"high"low, medium, high, neverActions of this risk and above need approval: low is reading, medium is creating and changing, high is deleting. never asks for nothing.
approval.alwaysAsklist of objectup to 32 itemsKinds whose writes always need approval.
approval.alwaysAsk[].groupstringyesup to 253 charactersAPI group of the kind.
approval.alwaysAsk[].resourcestringyesup to 63 charactersPlural resource name.
approval.expiresAfterSecondsinteger8640060 to 2592000A request nobody decided on expires after this long.
monitoringRefreference to MonitoringWorkspaceOptional MonitoringWorkspace for the agent's step metrics.

Status

FieldTypeRequiredDefaultAllowedDescription
phasestringPending, Running, Paused, Stopped, Degraded or Failed.
serviceAccountstringThe agent's own identity in this project.
roleBindingstringThe binding that gives the agent its role.
workflowobjectThe durable run of the agent's loop.
workflow.idstringWorkflow id.
workflow.runIdstringThe current run.
workflow.runsintegerRuns started (a stopped agent that is set Running again starts a new run).
lastStepAtstringWhen the agent last did a step (RFC 3339).
lastDecisionobjectWhat the agent decided last.
lastDecision.kindstringact, remember, recall, wait, done, paused, invalid or overBudget.
lastDecision.summarystringA short description.
pendingApprovalsintegerActions waiting for a person's decision.
usageobjectWhat the agent used.
usage.daystringThe UTC day of the daily counters (YYYY-MM-DD).
usage.tokensTodayintegerTokens on that day.
usage.cpuSecondsTodayintegerCPU seconds on that day.
usage.inputTokensintegerInput tokens in all.
usage.outputTokensintegerOutput tokens in all.
usage.hourstringThe UTC hour of actionsThisHour.
usage.actionsThisHourintegerAPI calls in that hour.
usage.actionsintegerAPI calls in all.
usage.refusedintegerCalls refused (by its role, a budget or a person) in all.
usage.stepsintegerSteps in all.
conditionslist of objectDetailed state; Ready says whether the agent is working.
conditions[].typestringAspect.
conditions[].statusstringTrue or False.
conditions[].reasonstringMachine-readable reason.
conditions[].messagestringHuman-readable detail.

Columns in listings

ColumnFrom
State.spec.state
Role.spec.roleRef.name
Steps.status.usage.steps
Actions.status.usage.actions
Phase.status.phase
Detail.status.conditions[0].message

All resources