feat(swarm): autonomous agent containers, Swarm Manager with auto-stop, /nodes UI overhaul

## 1. Fix /nodes Swarm Status Display - Add SwarmStatusBanner component: clear green/red/loading state - Shows nodeId, managerAddr, isManager badge - Error state explains what to check (docker.sock mount) - Header now shows 'swarm unreachable — check gateway' vs 'active' - swarmOk now checks nodeId presence, not just data existence ## 2. Autonomous Agent Container - New docker/Dockerfile.agent — builds Go agent binary from gateway/cmd/agent/ - New gateway/cmd/agent/main.go — standalone HTTP microservice: * GET /health — liveness probe with idle time info * POST /task — receives task, forwards to Gateway orchestrator * GET /info — agent metadata (id, hostname, gateway url) * Idle watchdog: calls /api/swarm/agents/{name}/stop after IdleTimeoutMinutes * Connects to Swarm overlay network (goclaw-net) → reaches DB/Gateway by DNS * Env: AGENT_ID, GATEWAY_URL, DATABASE_URL, IDLE_TIMEOUT_MINUTES ## 3. Swarm Manager Agent (auto-stop after 15min idle) - New gateway/internal/api/swarm_manager.go: * SwarmManager goroutine checks every 60s * Scales idle GoClaw agent services to 0 replicas after 15 min * Tracks lastActivity from task UpdatedAt timestamps - New REST endpoints in gateway: * GET /api/swarm/agents — list agents with idleMinutes * POST /api/swarm/agents/{name}/start — scale up agent * POST /api/swarm/agents/{name}/stop — scale to 0 * DELETE /api/swarm/services/{id} — remove service permanently - SwarmManager started as background goroutine in main.go with context cancel ## 4. Docker Client Enhancements - Added NetworkAttachment type and Networks field to ServiceSpec - CreateAgentServiceFull(opts) — supports overlay networks, custom labels - CreateAgentService() delegates to CreateAgentServiceFull for backward compat - RemoveService(id) — DELETE /v1.44/services/{id} - GetServiceLastActivity(id) — finds latest task UpdatedAt for idle detection ## 5. tRPC & Gateway Proxy - New functions: removeSwarmService, listSwarmAgents, startSwarmAgent, stopSwarmAgent - SwarmAgentInfo type with idleMinutes, lastActivity, desiredReplicas - createAgentService now accepts networks[] parameter - New tRPC endpoints: nodes.removeService, nodes.listAgents, nodes.startAgent, nodes.stopAgent ## 6. Nodes.tsx UI Overhaul - SwarmStatusBanner component at top — no more silent 'connecting…' - New 'Agents' tab with AgentManagerRow: idle time, auto-stop warning, start/stop/remove buttons - IdleColor coding: green < 5m, yellow 5-10m, red 10m+ with countdown to auto-stop - ServiceRow: added Remove button with confirmation dialog - RemoveConfirmDialog component - DeployAgentDialog: added overlay networks field, default env includes GATEWAY_URL - All queries refetch after agent start/stop/remove
2026-03-21 20:37:21 +00:00
parent 12b8332b2f
commit a8a8ea1ee2
9 changed files with 1168 additions and 194 deletions
--- a/gateway/cmd/agent/main.go
+++ b/gateway/cmd/agent/main.go
@@ -0,0 +1,270 @@
+// GoClaw Agent Server — autonomous agent microservice
+//
+// Each agent runs as an independent container in the Docker Swarm overlay
+// network. It exposes an HTTP API that the GoClaw Orchestrator can reach
+// via the Swarm DNS name (e.g. http://goclaw-agent-researcher:8080).
+//
+// The agent:
+//   - Receives task requests from the orchestrator
+//   - Calls the LLM via the centrally-managed GoClaw Gateway
+//   - Reads/writes shared state in the MySQL database
+//   - Reports its last-activity time so the SwarmManager can auto-stop it
+//   - Gracefully shuts down after IdleTimeout with no requests
+package main
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"log"
+	"net/http"
+	"os"
+	"os/signal"
+	"strconv"
+	"strings"
+	"syscall"
+	"time"
+)
+
+// ─── Config ──────────────────────────────────────────────────────────────────
+
+type AgentConfig struct {
+	AgentID            string
+	Port               string
+	GatewayURL         string
+	LLMURL             string
+	LLMAPIKey          string
+	DatabaseURL        string
+	IdleTimeoutMinutes int
+}
+
+func loadConfig() AgentConfig {
+	idleMin := 15
+	if v := os.Getenv("IDLE_TIMEOUT_MINUTES"); v != "" {
+		if n, err := strconv.Atoi(v); err == nil {
+			idleMin = n
+		}
+	}
+	port := os.Getenv("AGENT_PORT")
+	if port == "" {
+		port = "8080"
+	}
+	return AgentConfig{
+		AgentID:            getEnv("AGENT_ID", "unnamed-agent"),
+		Port:               port,
+		GatewayURL:         getEnv("GATEWAY_URL", "http://goclaw-gateway:18789"),
+		LLMURL:             getEnv("LLM_BASE_URL", "https://ollama.com/v1"),
+		LLMAPIKey:          os.Getenv("LLM_API_KEY"),
+		DatabaseURL:        os.Getenv("DATABASE_URL"),
+		IdleTimeoutMinutes: idleMin,
+	}
+}
+
+func getEnv(key, fallback string) string {
+	if v := os.Getenv(key); v != "" {
+		return v
+	}
+	return fallback
+}
+
+// ─── State ───────────────────────────────────────────────────────────────────
+
+type Agent struct {
+	cfg          AgentConfig
+	lastActivity time.Time
+	httpClient   *http.Client
+}
+
+func NewAgent(cfg AgentConfig) *Agent {
+	return &Agent{
+		cfg:          cfg,
+		lastActivity: time.Now(),
+		httpClient:   &http.Client{Timeout: 120 * time.Second},
+	}
+}
+
+func (a *Agent) touch() {
+	a.lastActivity = time.Now()
+}
+
+// ─── HTTP handlers ────────────────────────────────────────────────────────────
+
+// GET /health — liveness probe
+func (a *Agent) handleHealth(w http.ResponseWriter, r *http.Request) {
+	respond(w, 200, map[string]any{
+		"ok":           true,
+		"agentId":      a.cfg.AgentID,
+		"lastActivity": a.lastActivity.Format(time.RFC3339),
+		"idleMinutes":  time.Since(a.lastActivity).Minutes(),
+	})
+}
+
+// POST /task — receive a task from the orchestrator
+// Body: { "sessionId": "abc", "messages": [...], "model": "qwen2.5:7b", "maxIter": 5 }
+func (a *Agent) handleTask(w http.ResponseWriter, r *http.Request) {
+	a.touch()
+	var body struct {
+		SessionID string          `json:"sessionId"`
+		Messages  json.RawMessage `json:"messages"`
+		Model     string          `json:"model"`
+		MaxIter   int             `json:"maxIter"`
+	}
+	if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+		respondError(w, 400, "invalid request: "+err.Error())
+		return
+	}
+	// Forward the task to the GoClaw Gateway orchestrator
+	gatewayURL := a.cfg.GatewayURL + "/api/orchestrator/chat"
+	reqBody, _ := json.Marshal(map[string]any{
+		"messages": body.Messages,
+		"model":    body.Model,
+		"maxIter":  body.MaxIter,
+	})
+
+	req, err := http.NewRequestWithContext(r.Context(), "POST", gatewayURL, strings.NewReader(string(reqBody)))
+	if err != nil {
+		respondError(w, 500, "request build error: "+err.Error())
+		return
+	}
+	req.Header.Set("Content-Type", "application/json")
+
+	resp, err := a.httpClient.Do(req)
+	if err != nil {
+		respondError(w, 502, "gateway error: "+err.Error())
+		return
+	}
+	defer resp.Body.Close()
+
+	var result map[string]any
+	if err := json.NewDecoder(resp.Body).Decode(&result); err != nil {
+		respondError(w, 502, "gateway response error: "+err.Error())
+		return
+	}
+
+	a.touch()
+	respond(w, 200, map[string]any{
+		"ok":        true,
+		"agentId":   a.cfg.AgentID,
+		"sessionId": body.SessionID,
+		"result":    result,
+	})
+}
+
+// GET /info — agent metadata
+func (a *Agent) handleInfo(w http.ResponseWriter, r *http.Request) {
+	hostname, _ := os.Hostname()
+	respond(w, 200, map[string]any{
+		"agentId":       a.cfg.AgentID,
+		"hostname":      hostname,
+		"gatewayUrl":    a.cfg.GatewayURL,
+		"idleTimeout":   a.cfg.IdleTimeoutMinutes,
+		"lastActivity":  a.lastActivity.Format(time.RFC3339),
+		"idleMinutes":   time.Since(a.lastActivity).Minutes(),
+	})
+}
+
+// ─── Idle watchdog ────────────────────────────────────────────────────────────
+
+func (a *Agent) runIdleWatchdog(cancel context.CancelFunc) {
+	threshold := time.Duration(a.cfg.IdleTimeoutMinutes) * time.Minute
+	ticker := time.NewTicker(30 * time.Second)
+	defer ticker.Stop()
+	for range ticker.C {
+		idle := time.Since(a.lastActivity)
+		if idle >= threshold {
+			log.Printf("[Agent %s] Idle for %.1f min — requesting self-stop via gateway",
+				a.cfg.AgentID, idle.Minutes())
+			a.selfStop()
+			cancel()
+			return
+		}
+	}
+}
+
+// selfStop asks the GoClaw Gateway to scale this service to 0.
+func (a *Agent) selfStop() {
+	url := fmt.Sprintf("%s/api/swarm/agents/%s/stop", a.cfg.GatewayURL, a.cfg.AgentID)
+	req, err := http.NewRequest("POST", url, nil)
+	if err != nil {
+		log.Printf("[Agent %s] selfStop error building request: %v", a.cfg.AgentID, err)
+		return
+	}
+	resp, err := a.httpClient.Do(req)
+	if err != nil {
+		log.Printf("[Agent %s] selfStop error: %v", a.cfg.AgentID, err)
+		return
+	}
+	body, _ := io.ReadAll(resp.Body)
+	resp.Body.Close()
+	log.Printf("[Agent %s] selfStop response %d: %s", a.cfg.AgentID, resp.StatusCode, string(body))
+}
+
+// ─── Helpers ─────────────────────────────────────────────────────────────────
+
+func respond(w http.ResponseWriter, status int, data any) {
+	w.Header().Set("Content-Type", "application/json")
+	w.WriteHeader(status)
+	json.NewEncoder(w).Encode(data)
+}
+
+func respondError(w http.ResponseWriter, status int, msg string) {
+	respond(w, status, map[string]any{"error": msg})
+}
+
+// ─── Main ─────────────────────────────────────────────────────────────────────
+
+func main() {
+	log.SetFlags(log.LstdFlags | log.Lshortfile)
+
+	cfg := loadConfig()
+	agent := NewAgent(cfg)
+
+	log.Printf("[Agent] %s starting on port %s (idle timeout: %d min)",
+		cfg.AgentID, cfg.Port, cfg.IdleTimeoutMinutes)
+	log.Printf("[Agent] Gateway: %s", cfg.GatewayURL)
+
+	// ── HTTP server ──────────────────────────────────────────────────────────
+	mux := http.NewServeMux()
+	mux.HandleFunc("GET /health", agent.handleHealth)
+	mux.HandleFunc("POST /task", agent.handleTask)
+	mux.HandleFunc("GET /info", agent.handleInfo)
+
+	srv := &http.Server{
+		Addr:         ":" + cfg.Port,
+		Handler:      mux,
+		ReadTimeout:  30 * time.Second,
+		WriteTimeout: 150 * time.Second,
+		IdleTimeout:  120 * time.Second,
+	}
+
+	ctx, cancel := context.WithCancel(context.Background())
+
+	// ── Idle watchdog ────────────────────────────────────────────────────────
+	go agent.runIdleWatchdog(cancel)
+
+	// ── Graceful shutdown ────────────────────────────────────────────────────
+	quit := make(chan os.Signal, 1)
+	signal.Notify(quit, syscall.SIGINT, syscall.SIGTERM)
+
+	go func() {
+		log.Printf("[Agent %s] Listening on :%s", cfg.AgentID, cfg.Port)
+		if err := srv.ListenAndServe(); err != nil && err != http.ErrServerClosed {
+			log.Fatalf("[Agent %s] Server error: %v", cfg.AgentID, err)
+		}
+	}()
+
+	select {
+	case <-quit:
+		log.Printf("[Agent %s] Signal received — shutting down", cfg.AgentID)
+	case <-ctx.Done():
+		log.Printf("[Agent %s] Context cancelled — shutting down", cfg.AgentID)
+	}
+
+	shutCtx, shutCancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer shutCancel()
+	if err := srv.Shutdown(shutCtx); err != nil {
+		log.Printf("[Agent %s] Shutdown error: %v", cfg.AgentID, err)
+	}
+	log.Printf("[Agent %s] Stopped.", cfg.AgentID)
+}
--- a/gateway/cmd/gateway/main.go
+++ b/gateway/cmd/gateway/main.go
@@ -135,12 +135,22 @@ func main() {
 		r.Post("/swarm/nodes/{id}/availability", h.SwarmSetNodeAvailability)
 		r.Get("/swarm/services", h.SwarmServices)
 		r.Post("/swarm/services/create", h.SwarmCreateService)
+		r.Delete("/swarm/services/{id}", h.SwarmRemoveService)
 		r.Get("/swarm/services/{id}/tasks", h.SwarmServiceTasks)
 		r.Post("/swarm/services/{id}/scale", h.SwarmScaleService)
 		r.Get("/swarm/join-token", h.SwarmJoinToken)
 		r.Post("/swarm/shell", h.SwarmShell)
+		r.Get("/swarm/agents", h.SwarmListAgents)
+		r.Post("/swarm/agents/{name}/start", h.SwarmStartAgent)
+		r.Post("/swarm/agents/{name}/stop", h.SwarmStopAgent)
 	})

+	// ── Swarm Manager: auto-stop idle agents after 15 min ────────────────────
+	swarmMgr := api.NewSwarmManager(h, 60*time.Second)
+	managerCtx, managerCancel := context.WithCancel(context.Background())
+	go swarmMgr.Start(managerCtx)
+	defer managerCancel()
+
 	// ── Start Server ─────────────────────────────────────────────────────────
 	srv := &http.Server{
 		Addr:         ":" + cfg.Port,