feat(swarm): HM-side Swarm Run read-only query — Phase1 (#45)

#45 Phase1 的只读查询(list/status/events?after/artifacts),全部基于 HM 已持久化的
运行时回调数据(Swarm → HM 带签名回调,见 agent_callback.go),**无需实时调 Swarm**,
因此不被 agent_swarm#2 契约冻结阻塞、返工风险低:

- GET /api/heicode/swarms — 列出当前用户的 swarm 运行(AgentDeployment, sub_mode=swarm 或有 runtime_swarm_id)
- GET /api/heicode/swarms/:id — 状态(:id = deployment_id/swarm_id/correlation_id 任一)
- GET /api/heicode/swarms/:id/events?after=&limit= — 事件增量拉取(id 游标 next_after,oldest-first)
  新增 model.ListSwarmCallbackEventsAfter(按 deployment_id/swarm_id + id>after)
- GET /api/heicode/swarms/:id/artifacts — 从已存事件(event_type 含 artifact)派生
- POST /api/heicode/swarms/:id/stop — 唯一写操作;在 agent_swarm#2 冻结 + SWARM_RUNTIME_ENABLED=true
  前默认关闭并明确提示(不臆造未冻结写接口)

字段口径对齐 agent_swarm/docs/integration/runtime-contract.md(deployment_id↔swarm_id↔
manager_deployment_id;状态机 waiting_approval→running→…)。所有查询按 user 作用域,视图脱敏
(不含 plan/payload 大字段与凭据)。

测试:TestListSwarmCallbackEventsAfter(游标/过滤/空标识);TestMain 迁移 AgentDeployment +
AgentCallbackEvent。go build/vet 干净,controller+model 全套回归通过。文档 §5.2。

Affects: Manager only(新增只读查询端点 + 一个 gated 写端点)。无计费/审计 schema 改动;
不依赖未冻结契约。stop 真实接入随 agent_swarm#2 冻结落地。

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

heicode/controller/agent_swarm_query.go

@@ -0,0 +1,191 @@
+package controller
+
+import (
+	"errors"
+	"strconv"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+	"github.com/heicode/manager/common"
+	"github.com/heicode/manager/model"
+	"gorm.io/gorm"
+)
+
+// HM-side Swarm Run query (#45 Phase1: list / status / events?after / artifacts / stop).
+//
+// 设计要点(关键):**只读查询全部基于 HM 本地已持久化的数据**——Swarm 运行时通过带签名回调
+// (`/api/agent/callbacks/runtime-events`,见 agent_callback.go)把生命周期事件推给 HM,HM 落
+// 到 AgentDeployment + AgentCallbackEvent。因此 list/status/events/artifacts **无需**调用 Swarm
+// 运行时,也就**不依赖 `agent_swarm#2` 尚未冻结的拉取契约**,返工风险低。
+//
+// 仅 `stop`(写操作)需要真正调用 Swarm 运行时;在契约冻结 + `SWARM_RUNTIME_ENABLED=true` 前
+// 默认关闭并明确提示(不 mock,不臆造未冻结的写接口)。
+//
+// 字段口径对齐 agent_swarm/docs/integration/runtime-contract.md(草案):
+// deployment_id ↔ swarm_id ↔ manager_deployment_id 三者映射;事件按 swarm_id/deployment_id 持久化。
+
+// findUserSwarmDeployment 按 :id(匹配 deployment_id / runtime_swarm_id / correlation_id)加载
+// 当前用户的 swarm 部署。非本人或非 swarm 模式 → 失败。
+func findUserSwarmDeployment(c *gin.Context) (model.AgentDeployment, bool) {
+	userID := c.GetInt("id")
+	id := strings.TrimSpace(c.Param("id"))
+	var dep model.AgentDeployment
+	err := model.DB.Where(
+		"user_id = ? AND (deployment_id = ? OR runtime_swarm_id = ? OR correlation_id = ?)",
+		strconv.Itoa(userID), id, id, id,
+	).First(&dep).Error
+	if err != nil {
+		if errors.Is(err, gorm.ErrRecordNotFound) {
+			agentError(c, "POLICY_REJECTED", "swarm run not found")
+		} else {
+			agentError(c, "DEPLOYMENT_CONFLICT", "failed to load swarm run")
+		}
+		return model.AgentDeployment{}, false
+	}
+	if !isSwarmDeployment(dep) {
+		agentError(c, "POLICY_REJECTED", "deployment is not a swarm run")
+		return model.AgentDeployment{}, false
+	}
+	return dep, true
+}
+
+func isSwarmDeployment(dep model.AgentDeployment) bool {
+	return strings.EqualFold(dep.SubMode, "swarm") || strings.TrimSpace(dep.RuntimeSwarmID) != ""
+}
+
+// swarmDeploymentView 是 swarm 运行的状态摘要视图(脱敏:不含 plan/payload 等大字段与凭据)。
+func swarmDeploymentView(dep model.AgentDeployment) gin.H {
+	return gin.H{
+		"deployment_id":         dep.DeploymentID,
+		"swarm_id":              dep.RuntimeSwarmID,
+		"runtime_deployment_id": dep.RuntimeDeploymentID,
+		"correlation_id":        dep.CorrelationID,
+		"status":                dep.Status,
+		"phase":                 dep.Phase,
+		"runtime_state":         dep.RuntimeState,
+		"failure_reason":        dep.FailureReason,
+		"created_at":            dep.CreatedAtText,
+		"updated_at":            dep.UpdatedAtText,
+		"runtime_last_sync_at":  dep.RuntimeLastSyncAtText,
+	}
+}
+
+// swarmEventView maps a persisted callback event to the client view.
+func swarmEventView(e model.AgentCallbackEvent) gin.H {
+	return gin.H{
+		"id":                e.Id, // 作为下一次 ?after= 的游标
+		"event_id":          e.EventID,
+		"event_type":        e.EventType,
+		"task_id":           e.TaskID,
+		"agent_instance_id": e.AgentInstanceID,
+		"result":            e.Result,
+		"occurred_at":       e.OccurredAt,
+		"created_at_ms":     e.CreatedAtMs,
+		"payload":           unmarshalResourceJSON(e.PayloadJSON),
+	}
+}
+
+// HeicodeListSwarms: GET /api/heicode/swarms — 列出当前用户的 swarm 运行(读 HM 本地部署表)。
+func HeicodeListSwarms(c *gin.Context) {
+	userID := c.GetInt("id")
+	if userID <= 0 {
+		agentError(c, "POLICY_REJECTED", "authentication required")
+		return
+	}
+	if model.DB == nil {
+		agentError(c, "DEPLOYMENT_PERSIST_FAILED", "database not initialised")
+		return
+	}
+	var rows []model.AgentDeployment
+	err := model.DB.Where(
+		"user_id = ? AND (LOWER(sub_mode) = ? OR runtime_swarm_id <> '')",
+		strconv.Itoa(userID), "swarm",
+	).Order("created_at_ms desc, id desc").Limit(200).Find(&rows).Error
+	if err != nil {
+		agentError(c, "DEPLOYMENT_CONFLICT", "failed to list swarm runs")
+		return
+	}
+	items := make([]gin.H, 0, len(rows))
+	for _, r := range rows {
+		items = append(items, swarmDeploymentView(r))
+	}
+	common.ApiSuccess(c, gin.H{"items": items, "total": len(items)})
+}
+
+// HeicodeGetSwarmStatus: GET /api/heicode/swarms/:id — 单个 swarm 运行状态。
+func HeicodeGetSwarmStatus(c *gin.Context) {
+	dep, ok := findUserSwarmDeployment(c)
+	if !ok {
+		return
+	}
+	common.ApiSuccess(c, swarmDeploymentView(dep))
+}
+
+// HeicodeListSwarmEvents: GET /api/heicode/swarms/:id/events?after=&limit= — 事件增量拉取。
+func HeicodeListSwarmEvents(c *gin.Context) {
+	dep, ok := findUserSwarmDeployment(c)
+	if !ok {
+		return
+	}
+	after, _ := strconv.Atoi(strings.TrimSpace(c.Query("after")))
+	limit, _ := strconv.Atoi(strings.TrimSpace(c.Query("limit")))
+	events, err := model.ListSwarmCallbackEventsAfter(dep.DeploymentID, dep.RuntimeSwarmID, after, limit)
+	if err != nil {
+		agentError(c, "DEPLOYMENT_CONFLICT", "failed to list swarm events")
+		return
+	}
+	items := make([]gin.H, 0, len(events))
+	nextAfter := after
+	for _, e := range events {
+		items = append(items, swarmEventView(e))
+		if e.Id > nextAfter {
+			nextAfter = e.Id
+		}
+	}
+	common.ApiSuccess(c, gin.H{"items": items, "next_after": nextAfter, "count": len(items)})
+}
+
+// HeicodeListSwarmArtifacts: GET /api/heicode/swarms/:id/artifacts — 从已持久化事件中筛产物。
+// 当前从 event_type 含 "artifact" 的回调事件派生(真实数据);专用 artifact 端点待 agent_swarm#2 冻结后补。
+func HeicodeListSwarmArtifacts(c *gin.Context) {
+	dep, ok := findUserSwarmDeployment(c)
+	if !ok {
+		return
+	}
+	events, err := model.ListSwarmCallbackEventsAfter(dep.DeploymentID, dep.RuntimeSwarmID, 0, 1000)
+	if err != nil {
+		agentError(c, "DEPLOYMENT_CONFLICT", "failed to list swarm artifacts")
+		return
+	}
+	items := make([]gin.H, 0)
+	for _, e := range events {
+		if strings.Contains(strings.ToLower(e.EventType), "artifact") {
+			items = append(items, swarmEventView(e))
+		}
+	}
+	common.ApiSuccess(c, gin.H{"items": items, "total": len(items),
+		"note": "derived from persisted runtime events; dedicated artifact contract pending agent_swarm#2 freeze"})
+}
+
+// HeicodeStopSwarm: POST /api/heicode/swarms/:id/stop — 停止 swarm 运行(写操作)。
+// 这是唯一需要真正调用 Swarm 运行时的操作。在 agent_swarm#2 契约冻结 + SWARM_RUNTIME_ENABLED=true
+// 前默认关闭并明确提示(不臆造未冻结的写接口)。冻结后在此接入运行时 stop(草案路径
+// /api/agent/swarm/deployments/{id}/stop)。
+func HeicodeStopSwarm(c *gin.Context) {
+	dep, ok := findUserSwarmDeployment(c)
+	if !ok {
+		return
+	}
+	if !common.GetEnvOrDefaultBool("SWARM_RUNTIME_ENABLED", false) {
+		agentError(c, "POLICY_REJECTED",
+			"swarm stop is not enabled yet: pending agent_swarm#2 runtime-contract freeze and SWARM_RUNTIME_ENABLED=true")
+		return
+	}
+	// 契约冻结后在此调用 Swarm 运行时 stop;当前仅在本地记录意图,避免对未冻结写接口下注。
+	common.ApiSuccess(c, gin.H{
+		"deployment_id": dep.DeploymentID,
+		"swarm_id":      dep.RuntimeSwarmID,
+		"accepted":      true,
+		"note":          "runtime stop wiring lands with agent_swarm#2 contract freeze",
+	})
+}

heicode/model/agent_callback.go

@@ -55,6 +55,37 @@ func InsertAgentCallbackEvent(row *AgentCallbackEvent) (bool, error) {
 	return true, nil
 }
 
+// ListSwarmCallbackEventsAfter returns persisted swarm runtime events for one run
+// (matched by deployment_id OR swarm_id) with an id-based `after` cursor for
+// incremental polling (#45 events?after). Ordered oldest-first so the client can
+// append; the caller uses the last returned Id as the next `after`. Reading from
+// HM-persisted callback rows means this needs no live Swarm call.
+func ListSwarmCallbackEventsAfter(deploymentID, swarmID string, afterID, limit int) ([]AgentCallbackEvent, error) {
+	if DB == nil {
+		return nil, nil
+	}
+	q := DB.Model(&AgentCallbackEvent{})
+	switch {
+	case deploymentID != "" && swarmID != "":
+		q = q.Where("deployment_id = ? OR swarm_id = ?", deploymentID, swarmID)
+	case deploymentID != "":
+		q = q.Where("deployment_id = ?", deploymentID)
+	case swarmID != "":
+		q = q.Where("swarm_id = ?", swarmID)
+	default:
+		return nil, nil
+	}
+	if afterID > 0 {
+		q = q.Where("id > ?", afterID)
+	}
+	if limit <= 0 || limit > 1000 {
+		limit = 200
+	}
+	var items []AgentCallbackEvent
+	err := q.Order("id asc").Limit(limit).Find(&items).Error
+	return items, err
+}
+
 func ListAgentCallbackEvents(f ListAgentCallbackEventsFilter) ([]AgentCallbackEvent, error) {
 	if DB == nil {
 		return nil, nil

heicode/model/agent_swarm_query_test.go

@@ -0,0 +1,45 @@
+package model
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+// #45: events?after 游标 —— 按 deployment_id/swarm_id 过滤,id>after 增量返回,oldest-first。
+func TestListSwarmCallbackEventsAfter(t *testing.T) {
+	require.NoError(t, LOG_DB.Where("1 = 1").Delete(&AgentCallbackEvent{}).Error)
+	mk := func(eventID, dep, swarm, etype string) {
+		_, err := InsertAgentCallbackEvent(&AgentCallbackEvent{
+			EventID: eventID, DeploymentID: dep, SwarmID: swarm, EventType: etype,
+		})
+		require.NoError(t, err)
+	}
+	mk("e1", "dep_A", "swarm_A", "deployment.status_changed")
+	mk("e2", "dep_A", "swarm_A", "task.created")
+	mk("e3", "dep_A", "swarm_A", "artifact.produced")
+	mk("e4", "dep_B", "swarm_B", "task.created") // 另一个 run,应被排除
+
+	// 全量(after=0):dep_A 的 3 条,oldest-first
+	all, err := ListSwarmCallbackEventsAfter("dep_A", "swarm_A", 0, 100)
+	require.NoError(t, err)
+	require.Len(t, all, 3)
+	require.Equal(t, "e1", all[0].EventID)
+
+	// 游标:after = 第一条 id → 只返回其后的 2 条
+	after := all[0].Id
+	rest, err := ListSwarmCallbackEventsAfter("dep_A", "swarm_A", after, 100)
+	require.NoError(t, err)
+	require.Len(t, rest, 2)
+	require.Equal(t, "e2", rest[0].EventID)
+
+	// 仅按 swarm_id 也能查到
+	bySwarm, err := ListSwarmCallbackEventsAfter("", "swarm_A", 0, 100)
+	require.NoError(t, err)
+	require.Len(t, bySwarm, 3)
+
+	// 空标识 → 空
+	none, err := ListSwarmCallbackEventsAfter("", "", 0, 100)
+	require.NoError(t, err)
+	require.Empty(t, none)
+}

heicode/model/task_cas_test.go

@@ -44,6 +44,8 @@ func TestMain(m *testing.M) {
 		&SubscriptionOrder{},
 		&UserSubscription{},
 		&TelemetryEvent{},
+		&AgentDeployment{},
+		&AgentCallbackEvent{},
 	); err != nil {
 		panic("failed to migrate: " + err.Error())
 	}

heicode/router/api-router.go

@@ -547,6 +547,13 @@ func SetApiRouter(router *gin.Engine) {
 			// Client error-telemetry ingest (#24). Device-paired; never bills.
 			// Gated by HEICODE_TELEMETRY_ENABLED (default off -> 410 kill switch).
 			heicodeAgentRoute.POST("/telemetry/events", controller.HeicodeTelemetryEvents)
+			// Swarm Run query (#45 Phase1). Read-only views served from HM-persisted
+			// callback data (no live Swarm call); stop is gated pending agent_swarm#2 freeze.
+			heicodeAgentRoute.GET("/swarms", controller.HeicodeListSwarms)
+			heicodeAgentRoute.GET("/swarms/:id", controller.HeicodeGetSwarmStatus)
+			heicodeAgentRoute.GET("/swarms/:id/events", controller.HeicodeListSwarmEvents)
+			heicodeAgentRoute.GET("/swarms/:id/artifacts", controller.HeicodeListSwarmArtifacts)
+			heicodeAgentRoute.POST("/swarms/:id/stop", controller.HeicodeStopSwarm)
 		}
 
 		// Client↔agent access control (HM-provided, AM-optional). Called by the