@nylorun/runtime 0.42.0-beta → 0.44.0-beta

package/dist/openapi.json

@@ -2,7 +2,7 @@
   "openapi": "3.2.0",
   "info": {
     "title": "Nylorun Runtime API",
-    "version": "0.42.0-beta",
+    "version": "0.44.0-beta",
     "x-nylorun-protocol": {
       "min": 16,
       "max": 18,
@@ -43,7 +43,9 @@
         "agent-model",
         "secret-routing",
         "remove-tools",
-        "agent-triggers"
+        "agent-triggers",
+        "steer",
+        "queue-recovery"
       ]
     },
     "description": "For developers: agent manifests, Agents, sessions (with AG-UI and A2A) and artifacts, through `@nylorun/agents`. Every Tenant serves it on its one URL, beside the Management API (`/openapi/management.json`). Every request sends `Nylorun-Protocol`. Who may call each operation is its `security` and its `x-nylorun-credentials` and `x-nylorun-scopes` (the subject scopes that reach it) fields: an application key (servers) or a trusted issuer's token (browsers and apps)."
@@ -6336,6 +6338,13 @@
             "type": "integer",
             "exclusiveMinimum": 0
           },
+          "delivery": {
+            "type": "string",
+            "enum": [
+              "queue",
+              "steer"
+            ]
+          },
           "cursor": {
             "type": [
               "string",
@@ -6788,7 +6797,8 @@
                 "type": "string",
                 "enum": [
                   "queue",
-                  "reject"
+                  "reject",
+                  "steer"
                 ]
               },
               "content": {
@@ -7116,7 +7126,8 @@
                 "type": "string",
                 "enum": [
                   "queue",
-                  "reject"
+                  "reject",
+                  "steer"
                 ]
               },
               "data": {
@@ -7443,7 +7454,8 @@
                 "type": "string",
                 "enum": [
                   "queue",
-                  "reject"
+                  "reject",
+                  "steer"
                 ]
               },
               "parts": {
@@ -9330,6 +9342,12 @@
           {
             "$ref": "#/components/schemas/QueueTakenEvent"
           },
+          {
+            "$ref": "#/components/schemas/QueueFailedEvent"
+          },
+          {
+            "$ref": "#/components/schemas/QueueCancelledEvent"
+          },
           {
             "$ref": "#/components/schemas/WakeBookedEvent"
           },
@@ -9426,6 +9444,8 @@
             "session.model": "#/components/schemas/SessionModelEvent",
             "message.queued": "#/components/schemas/MessageQueuedEvent",
             "queue.taken": "#/components/schemas/QueueTakenEvent",
+            "queue.failed": "#/components/schemas/QueueFailedEvent",
+            "queue.cancelled": "#/components/schemas/QueueCancelledEvent",
             "wake.booked": "#/components/schemas/WakeBookedEvent",
             "wake.cancelled": "#/components/schemas/WakeCancelledEvent",
             "memory.updated": "#/components/schemas/MemoryUpdatedEvent",
@@ -9700,7 +9720,8 @@
                 "type": "string",
                 "enum": [
                   "queue",
-                  "reject"
+                  "reject",
+                  "steer"
                 ]
               },
               "queuedMessageId": {
@@ -11856,6 +11877,13 @@
               },
               "requestId": {
                 "type": "string"
+              },
+              "delivery": {
+                "type": "string",
+                "enum": [
+                  "queue",
+                  "steer"
+                ]
               }
             },
             "required": [
@@ -12057,6 +12085,10 @@
                   "type": "string"
                 },
                 "minItems": 1
+              },
+              "step": {
+                "type": "integer",
+                "exclusiveMinimum": 0
               }
             },
             "required": [
@@ -12087,6 +12119,421 @@
           "payload"
         ]
       },
+      "QueueFailedEvent": {
+        "type": "object",
+        "properties": {
+          "schema": {
+            "type": "string",
+            "enum": [
+              "nylorun.event/2"
+            ]
+          },
+          "eventId": {
+            "type": "string"
+          },
+          "tenantId": {
+            "type": "string",
+            "minLength": 1
+          },
+          "sessionId": {
+            "type": "string"
+          },
+          "runId": {
+            "type": [
+              "string",
+              "null"
+            ]
+          },
+          "turnId": {
+            "type": [
+              "string",
+              "null"
+            ]
+          },
+          "incarnation": {
+            "type": "integer",
+            "minimum": 0
+          },
+          "epoch": {
+            "type": "integer",
+            "minimum": 0
+          },
+          "seq": {
+            "type": "integer",
+            "minimum": 0
+          },
+          "cursor": {
+            "type": "string"
+          },
+          "time": {
+            "type": "string"
+          },
+          "schemaVersion": {
+            "type": "integer",
+            "exclusiveMinimum": 0
+          },
+          "source": {
+            "type": "object",
+            "properties": {
+              "kind": {
+                "type": "string",
+                "enum": [
+                  "client",
+                  "core",
+                  "model-gate",
+                  "tool-gate",
+                  "egress-gate",
+                  "sandboxes",
+                  "edge",
+                  "init",
+                  "controller",
+                  "loop",
+                  "supervisor",
+                  "harness",
+                  "evidence",
+                  "operator",
+                  "human",
+                  "api"
+                ]
+              },
+              "id": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "kind",
+              "id"
+            ]
+          },
+          "evidence": {
+            "type": "string",
+            "enum": [
+              "observed",
+              "measured",
+              "attested",
+              "claimed",
+              "asserted"
+            ]
+          },
+          "principal": {
+            "type": "object",
+            "properties": {
+              "kind": {
+                "type": "string",
+                "enum": [
+                  "application",
+                  "subject",
+                  "flow-host",
+                  "core",
+                  "trigger"
+                ]
+              },
+              "id": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "kind",
+              "id"
+            ],
+            "additionalProperties": false
+          },
+          "trace": {
+            "type": "object",
+            "properties": {
+              "traceId": {
+                "type": "string"
+              },
+              "spanId": {
+                "type": "string"
+              },
+              "parentSpanId": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "traceId",
+              "spanId"
+            ]
+          },
+          "visibility": {
+            "type": "string",
+            "enum": [
+              "public",
+              "internal",
+              "restricted"
+            ]
+          },
+          "retention": {
+            "type": "string",
+            "enum": [
+              "full",
+              "metadata"
+            ]
+          },
+          "type": {
+            "type": "string",
+            "enum": [
+              "queue.failed"
+            ]
+          },
+          "payload": {
+            "type": "object",
+            "properties": {
+              "queuedMessageIds": {
+                "type": "array",
+                "items": {
+                  "type": "string"
+                },
+                "minItems": 1
+              },
+              "error": {
+                "type": "object",
+                "properties": {
+                  "code": {
+                    "type": "string"
+                  },
+                  "message": {
+                    "type": "string"
+                  },
+                  "reason": {
+                    "type": "string"
+                  }
+                },
+                "required": [
+                  "code",
+                  "message"
+                ],
+                "additionalProperties": {}
+              }
+            },
+            "required": [
+              "queuedMessageIds",
+              "error"
+            ],
+            "additionalProperties": {}
+          }
+        },
+        "required": [
+          "schema",
+          "eventId",
+          "tenantId",
+          "sessionId",
+          "runId",
+          "turnId",
+          "incarnation",
+          "epoch",
+          "seq",
+          "cursor",
+          "time",
+          "schemaVersion",
+          "source",
+          "evidence",
+          "visibility",
+          "retention",
+          "type",
+          "payload"
+        ]
+      },
+      "QueueCancelledEvent": {
+        "type": "object",
+        "properties": {
+          "schema": {
+            "type": "string",
+            "enum": [
+              "nylorun.event/2"
+            ]
+          },
+          "eventId": {
+            "type": "string"
+          },
+          "tenantId": {
+            "type": "string",
+            "minLength": 1
+          },
+          "sessionId": {
+            "type": "string"
+          },
+          "runId": {
+            "type": [
+              "string",
+              "null"
+            ]
+          },
+          "turnId": {
+            "type": [
+              "string",
+              "null"
+            ]
+          },
+          "incarnation": {
+            "type": "integer",
+            "minimum": 0
+          },
+          "epoch": {
+            "type": "integer",
+            "minimum": 0
+          },
+          "seq": {
+            "type": "integer",
+            "minimum": 0
+          },
+          "cursor": {
+            "type": "string"
+          },
+          "time": {
+            "type": "string"
+          },
+          "schemaVersion": {
+            "type": "integer",
+            "exclusiveMinimum": 0
+          },
+          "source": {
+            "type": "object",
+            "properties": {
+              "kind": {
+                "type": "string",
+                "enum": [
+                  "client",
+                  "core",
+                  "model-gate",
+                  "tool-gate",
+                  "egress-gate",
+                  "sandboxes",
+                  "edge",
+                  "init",
+                  "controller",
+                  "loop",
+                  "supervisor",
+                  "harness",
+                  "evidence",
+                  "operator",
+                  "human",
+                  "api"
+                ]
+              },
+              "id": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "kind",
+              "id"
+            ]
+          },
+          "evidence": {
+            "type": "string",
+            "enum": [
+              "observed",
+              "measured",
+              "attested",
+              "claimed",
+              "asserted"
+            ]
+          },
+          "principal": {
+            "type": "object",
+            "properties": {
+              "kind": {
+                "type": "string",
+                "enum": [
+                  "application",
+                  "subject",
+                  "flow-host",
+                  "core",
+                  "trigger"
+                ]
+              },
+              "id": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "kind",
+              "id"
+            ],
+            "additionalProperties": false
+          },
+          "trace": {
+            "type": "object",
+            "properties": {
+              "traceId": {
+                "type": "string"
+              },
+              "spanId": {
+                "type": "string"
+              },
+              "parentSpanId": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "traceId",
+              "spanId"
+            ]
+          },
+          "visibility": {
+            "type": "string",
+            "enum": [
+              "public",
+              "internal",
+              "restricted"
+            ]
+          },
+          "retention": {
+            "type": "string",
+            "enum": [
+              "full",
+              "metadata"
+            ]
+          },
+          "type": {
+            "type": "string",
+            "enum": [
+              "queue.cancelled"
+            ]
+          },
+          "payload": {
+            "type": "object",
+            "properties": {
+              "queuedMessageIds": {
+                "type": "array",
+                "items": {
+                  "type": "string"
+                },
+                "minItems": 1
+              },
+              "reason": {
+                "type": "string"
+              }
+            },
+            "required": [
+              "queuedMessageIds"
+            ],
+            "additionalProperties": {}
+          }
+        },
+        "required": [
+          "schema",
+          "eventId",
+          "tenantId",
+          "sessionId",
+          "runId",
+          "turnId",
+          "incarnation",
+          "epoch",
+          "seq",
+          "cursor",
+          "time",
+          "schemaVersion",
+          "source",
+          "evidence",
+          "visibility",
+          "retention",
+          "type",
+          "payload"
+        ]
+      },
       "WakeBookedEvent": {
         "type": "object",
         "properties": {
@@ -34432,6 +34879,12 @@
                     {
                       "$ref": "#/components/schemas/QueueTakenEvent"
                     },
+                    {
+                      "$ref": "#/components/schemas/QueueFailedEvent"
+                    },
+                    {
+                      "$ref": "#/components/schemas/QueueCancelledEvent"
+                    },
                     {
                       "$ref": "#/components/schemas/WakeBookedEvent"
                     },
@@ -36864,7 +37317,7 @@
           "A2A"
         ],
         "summary": "Call an agent over A2A",
-        "description": "A2A 1.0 JSON-RPC, for a subject, under the Nylorun profile v2 (Host feature `a2a-profile-v2`). `SendMessage` blocks up to the Runtime limit `a2aWaitMs` (5 minutes by default; then the task is answered still `WORKING`) unless `returnImmediately`; `GetTask` (with `historyLength`), `CancelTask` and `ListTasks`; and the four push notification config methods, plus `configuration.taskPushNotificationConfig` in `SendMessage`: each update is POSTed as one `StreamResponse`, signed with Standard Webhooks (`webhook-id`, `webhook-timestamp`, `webhook-signature` with the Runtime's `a2a.push.secret`) and `X-A2A-Notification-Token`, at least once, retried for up to 24 hours, to hosts the Runtime allows (`a2a.push.allowHosts` for private ones). A message to a busy context is queued and answered with the task that will take it (`TASK_STATE_SUBMITTED`); `message.metadata.delivery: \"reject\"` asks for `UnsupportedOperationError` (reason `CONTEXT_BUSY`) instead. `message.metadata.removeTools` (Host feature `remove-tools`) names the tools a new message's turn may not use, as a session message's `removeTools`. A message is text, one data part, or text with `raw` files (stored as artifacts of the context's session) and `url` files (fetched by the Runtime), each at most the Runtime limit `inlineFileBytes` (20 MiB by default). An approval, question or connection the task waits for (`AUTH_REQUIRED` or `INPUT_REQUIRED`) is shown under `interactions/v1` and answered on the same task with `{ interactionId, approved }` or `{ interactionId, value }` under that URI in the message's `metadata`; text never approves. Artifacts carry the turn's output and the files it wrote, as `url` parts at `/v1/a2a/files/…` with `files/v1`; the task's metadata carries `progress/v1`, `telemetry/v1` and `effects/v1`. Extension data is answered only for the extensions activated with `A2A-Extensions`. A context is one session per subject, agent and `contextId`; a task is one turn. JSON-RPC errors are answered with 200; limits, scopes and unavailable streams stay HTTP errors. Not served: streaming, `GetExtendedAgentCard`.",
+        "description": "A2A 1.0 JSON-RPC, for a subject, under the Nylorun profile v2 (Host feature `a2a-profile-v2`). `SendMessage` blocks up to the Runtime limit `a2aWaitMs` (5 minutes by default; then the task is answered still `WORKING`) unless `returnImmediately`; `GetTask` (with `historyLength`), `CancelTask` and `ListTasks`; and the four push notification config methods, plus `configuration.taskPushNotificationConfig` in `SendMessage`: each update is POSTed as one `StreamResponse`, signed with Standard Webhooks (`webhook-id`, `webhook-timestamp`, `webhook-signature` with the Runtime's `a2a.push.secret`) and `X-A2A-Notification-Token`, at least once, retried for up to 24 hours, to hosts the Runtime allows (`a2a.push.allowHosts` for private ones). A message to a busy context is queued and answered with the task that will take it (`TASK_STATE_SUBMITTED`); `message.metadata.delivery: \"reject\"` asks for `UnsupportedOperationError` (reason `CONTEXT_BUSY`) instead. A queued task never waits for good (Host feature `queue-recovery`): when its Agent refuses its turn's start (a disk copy, a lost pod), the start is tried again until the Runtime limit `queuedStartMs` (15 minutes by default), and then, or at once for a refusal no retry changes, the task is `TASK_STATE_FAILED` with a status message saying why; `CancelTask` on a queued task takes it out of the queue, a task running in the context or not, and answers `TASK_STATE_CANCELED`. Both are pushed. `message.metadata.removeTools` (Host feature `remove-tools`) names the tools a new message's turn may not use, as a session message's `removeTools`. A message is text, one data part, or text with `raw` files (stored as artifacts of the context's session) and `url` files (fetched by the Runtime), each at most the Runtime limit `inlineFileBytes` (20 MiB by default). An approval, question or connection the task waits for (`AUTH_REQUIRED` or `INPUT_REQUIRED`) is shown under `interactions/v1` and answered on the same task with `{ interactionId, approved }` or `{ interactionId, value }` under that URI in the message's `metadata`; text never approves. Artifacts carry the turn's output and the files it wrote, as `url` parts at `/v1/a2a/files/…` with `files/v1`; the task's metadata carries `progress/v1`, `telemetry/v1` and `effects/v1`. Extension data is answered only for the extensions activated with `A2A-Extensions`. A context is one session per subject, agent and `contextId`; a task is one turn. JSON-RPC errors are answered with 200; limits, scopes and unavailable streams stay HTTP errors. Not served: streaming, `GetExtendedAgentCard`.",
         "externalDocs": {
           "description": "The A2A specification",
           "url": "https://a2a-protocol.org/latest/specification/"

package/dist/sandbox/pods/lifecycle.d.ts

@@ -14,7 +14,11 @@
  *   is not the row's is applied again.
  * - **Idle** (`trigger: idle`): no turn on a session that names it, and none ended within its
  *   `lifecycle.idle` (at most the Tenant's `limits.idle`) → `Suspended` (`agent.suspended`,
- *   reason `idle`). The volume is kept; the next turn start asks for `Running` again.
+ *   reason `idle`). The volume is kept; the next turn start asks for `Running` again. Every
+ *   running, ready pod has its idle timer armed (at `lastActiveAt` + idle) whenever it is
+ *   reconciled, whichever path woke it: a turn's end arms it too, but a wake with no turn end
+ *   (a disk copy's `wakePod`, commands through the session's sandbox route) armed nothing and
+ *   ran on for days. A timer that fires during a turn looks again one idle period later.
  * - **Lost**: an incarnation that has run (a pod joined) whose Sandbox is gone or whose volume
  *   is missing is `lost` (`agent.lost`): its host epoch moves, its Sandbox is deleted, and
  *   only a reset brings the Environment back. agent-sandbox recreates a deleted claim empty, so

package/dist/sandbox/pods/lifecycle.js

@@ -118,6 +118,11 @@ export function decide(facts) {
         else
             arm.push({ timer: "idle", at: due });
     }
+    else if (facts.trigger === "idle" && desired === "running") {
+        // Busy: a turn's end arms the timer again, but look again anyway, so a turn that never ends
+        // cleanly doesn't keep the pod running for good.
+        arm.push({ timer: "idle", at: now + facts.config.idleMs });
+    }
     // An incarnation a pod has joined, whose Sandbox or volume is gone: lost.
     if (pod.podUid !== undefined && (!status.exists || status.deleting || status.volume === "missing")) {
         const reason = !status.exists
@@ -174,6 +179,9 @@ export function decide(facts) {
             const digest = pod.imageDigest === undefined ? imageDigestOf(status.imageID, facts.image) : undefined;
             if (digest !== undefined)
                 patch.imageDigest = digest;
+            // Whatever woke it, a running pod has an idle timer (the idle trigger armed its own above).
+            if (facts.trigger !== "idle")
+                arm.push({ timer: "idle", at: Date.parse(pod.lastActiveAt ?? row.createdAt) + facts.config.idleMs });
             return done();
         }
         const startedAt = Date.parse(pod.startedAt ?? row.updatedAt);

package/dist/sandbox/pods/network-gate.d.ts

@@ -16,9 +16,15 @@ export interface NetworkGateOptions {
     readonly pauseMs?: number;
     /** Consecutive failed connections that prove an address blocked. Default `harness.networkConfirmations` (3). */
     readonly confirmations?: number;
+    /**
+     * The pod starts on a disk an earlier pod of this sandbox used (a wake): check quickly, with
+     * `harness.wakeNetworkProbeMs` and `harness.wakeNetworkConfirmations`, until an address answers.
+     */
+    readonly wake?: boolean;
     readonly log?: (message: string, fields: Record<string, unknown>) => void;
 }
 /** Resolves once every address is blocked; rejects past the timeout. */
 export declare function awaitNetworkPolicy(options: NetworkGateOptions): Promise<{
     waitedMs: number;
+    check: "full" | "wake";
 }>;

package/dist/sandbox/pods/network-gate.js

@@ -6,6 +6,14 @@
  * `KUBERNETES_SERVICE_HOST:KUBERNETES_SERVICE_PORT`, which the kubelet always sets) until each
  * fails `harness.networkConfirmations` times in a row (Host tuning) (refused, unreachable or timed out). A connection that
  * succeeds starts the count again. Past `timeoutMs` the engine exits, and the pod restarts it.
+ *
+ * A wake (`wake`: the disk already holds the engine's `volume-id`, so a pod of this sandbox ran
+ * in this namespace before) starts the pod into a namespace whose policy has long been in force.
+ * The gate then takes `harness.wakeNetworkConfirmations` failed probes of
+ * `harness.wakeNetworkProbeMs` (1 of 150 ms) instead of three of a second: 152 ms instead of
+ * 3.0 s per wake (the A2 wake lab). It trusts a wake's quick check only while nothing answers:
+ * once a blocked address connects, the policy is not in force, and the gate goes back to the full
+ * check for every address.
  */
 import { connect } from "node:net";
 import { hostTuning } from "../../tuning.js";
@@ -47,14 +55,21 @@ export async function awaitNetworkPolicy(options) {
     const tuning = hostTuning().harness;
     const started = Date.now();
     const deadline = started + (options.timeoutMs ?? tuning.networkPolicyTimeoutMs);
-    const needed = options.confirmations ?? tuning.networkConfirmations;
+    const full = {
+        needed: options.confirmations ?? tuning.networkConfirmations,
+        attemptMs: options.attemptMs ?? tuning.networkProbeMs,
+    };
+    const quick = { needed: tuning.wakeNetworkConfirmations, attemptMs: tuning.wakeNetworkProbeMs };
+    let check = options.wake ? "wake" : "full";
     for (const address of options.addresses) {
         let failures = 0;
         let opened = 0;
-        while (failures < needed) {
+        while (failures < (check === "wake" ? quick : full).needed) {
             if (Date.now() > deadline)
                 throw new Error(`The pod's NetworkPolicy is not in force: ${address.host}:${address.port} still answered ${opened} time(s)`);
-            if (await opens(address, options.attemptMs ?? tuning.networkProbeMs)) {
+            if (await opens(address, (check === "wake" ? quick : full).attemptMs)) {
+                // An address the policy blocks answered: whatever the disk says, check in full.
+                check = "full";
                 failures = 0;
                 opened += 1;
                 await new Promise((resolve) => setTimeout(resolve, options.pauseMs ?? tuning.networkPauseMs));
@@ -65,7 +80,8 @@ export async function awaitNetworkPolicy(options) {
         options.log?.("sandbox_network_policy_checked", {
             address: `${address.host}:${address.port}`,
             openedBefore: opened,
+            check,
         });
     }
-    return { waitedMs: Date.now() - started };
+    return { waitedMs: Date.now() - started, check };
 }

package/dist/store/postgres/drizzle/0045_session_queue_delivery.sql

@@ -0,0 +1,2 @@
+ALTER TABLE "nylorun"."session_queue" ADD COLUMN "delivery" text DEFAULT 'queue' NOT NULL;--> statement-breakpoint
+ALTER TABLE "nylorun"."session_queue" ADD CONSTRAINT "session_queue_delivery_check" CHECK ("nylorun"."session_queue"."delivery" IN ('queue', 'steer'));

package/dist/store/postgres/drizzle/meta/_journal.json

@@ -316,6 +316,13 @@
       "when": 1793548858191,
       "tag": "0044_agent_triggers_origin_backfill",
       "breakpoints": true
+    },
+    {
+      "idx": 45,
+      "version": "7",
+      "when": 1793954783377,
+      "tag": "0045_session_queue_delivery",
+      "breakpoints": true
     }
   ]
 }