fix(swarm): production deploy bugs (volumes, sessions, providers, theme, missing assets)

.dockerignore

@@ -3,6 +3,8 @@
 .env
 __pycache__/
 *.pyc
+dashboard/data/
+workspace/
 ADWs/logs/
 ADWs/__pycache__/
 .claude/agent-memory/

Dockerfile.dashboard

@@ -40,6 +40,14 @@ COPY dashboard/backend/ dashboard/backend/
 COPY social-auth/ social-auth/
 COPY scheduler.py ./
 
+# Copy workspace assets the backend reads at runtime.
+# Without these, /api/agents, /api/skills, /api/commands etc. all return empty
+# and the UI shows "No agents found" / "No skills found" on a fresh deploy.
+# .claude/agent-memory and .claude/.env are excluded by .dockerignore so user
+# data and secrets stay out of the image.
+COPY .claude/ .claude/
+COPY docs/ docs/
+
 # Copy built frontend from stage 1
 COPY --from=frontend-build /frontend/dist dashboard/frontend/dist
 

dashboard/terminal-server/src/claude-bridge.js

@@ -15,6 +15,7 @@ class ClaudeBridge {
   _loadProviderConfig() {
     const ALLOWED_CLI = new Set(['claude', 'openclaude']);
     const ALLOWED_VARS = new Set([
+      'ANTHROPIC_API_KEY',
       'CLAUDE_CODE_USE_OPENAI', 'CLAUDE_CODE_USE_GEMINI',
       'CLAUDE_CODE_USE_BEDROCK', 'CLAUDE_CODE_USE_VERTEX',
       'OPENAI_BASE_URL', 'OPENAI_API_KEY', 'OPENAI_MODEL',
@@ -138,7 +139,22 @@ class ClaudeBridge {
 
   async startSession(sessionId, options = {}) {
     if (this.sessions.has(sessionId)) {
-      throw new Error(`Session ${sessionId} already exists`);
+      const existing = this.sessions.get(sessionId);
+      if (existing.active) {
+        // Idempotent: a duplicate startSession can arrive when the WebSocket
+        // reconnects through a reverse proxy (Traefik) and the frontend
+        // re-sends start_claude before learning the session is still alive.
+        // Returning the existing session instead of throwing prevents a
+        // confusing "Session already exists" toast on the user's terminal
+        // while keeping the original PTY intact.
+        console.log(`[bridge] startSession(${sessionId}) — already active, returning existing session`);
+        return existing;
+      }
+      // Orphaned dead session — clean up and restart
+      if (existing.process) {
+        try { existing.process.kill('SIGKILL'); } catch (_) {}
+      }
+      this.sessions.delete(sessionId);
     }
 
     const {

dashboard/terminal-server/src/server.js

@@ -778,7 +778,11 @@ class TerminalServer {
     if (!session) return;
 
     if (session.active) {
-      this.sendToWebSocket(wsInfo.ws, { type: 'error', message: 'An agent is already running in this session' });
+      // Frontend may re-send start_claude on WebSocket reconnect (common
+      // through reverse proxies like Traefik). The session is already
+      // running — replay the buffer and tell the client it's attached
+      // instead of surfacing a misleading error toast.
+      this.sendToWebSocket(wsInfo.ws, { type: 'claude_started', sessionId: wsInfo.claudeSessionId });
       return;
     }
 

docker-compose.yml

@@ -3,25 +3,27 @@ services:
   dashboard:
     build:
       context: .
-      dockerfile: Dockerfile.dashboard
+      dockerfile: Dockerfile.swarm.dashboard
     container_name: evonexus-dashboard
     ports:
-      - "${EVONEXUS_PORT:-8080}:8080"
+      - "8081:8080"
+      - "32352:32352"
     env_file: .env
     environment:
       - TZ=America/Sao_Paulo
       - EVONEXUS_PORT=8080
+      - TERMINAL_SERVER_PORT=32352
     volumes:
-      - ./.env:/workspace/.env:ro
-      - ./config:/workspace/config:ro
+      - claude-auth:/root/.claude
+      - ./config:/workspace/config
       - ./workspace:/workspace/workspace
       - ./dashboard/data:/workspace/dashboard/data
       - ./.claude/agents:/workspace/.claude/agents:ro
       - ./.claude/skills:/workspace/.claude/skills:ro
       - ./.claude/commands:/workspace/.claude/commands:ro
       - ./.claude/templates:/workspace/.claude/templates:ro
       - ./memory:/workspace/memory:ro
-      - ./ADWs/logs:/workspace/ADWs/logs:ro
+      - ./ADWs/logs:/workspace/ADWs/logs
     restart: unless-stopped
     healthcheck:
       test: ["CMD", "curl", "-f", "http://localhost:8080/api/version"]
@@ -79,6 +81,7 @@ services:
       - manual
 
 volumes:
+  claude-auth:
   daily-logs:
   projects:
   community:

evonexus.portainer.stack.yml

@@ -0,0 +1,166 @@
+version: "3.8"
+
+# ============================================================
+# EvoNexus - Stack de Produção (Docker Swarm / Portainer)
+#
+# Baseado na documentação oficial do projeto:
+# - README.swarm.md
+# - evonexus.stack.yml
+#
+# Ajustado para este servidor:
+# - Domínio: evonexus.advancedbot.com.br
+# - Rede Traefik: network_public
+# - Entrypoint Traefik: websecure
+# - Certresolver: letsencryptresolver
+#
+# Antes do deploy:
+# 1. Garanta que a rede `network_public` já exista no Swarm
+# 2. As imagens publicadas no Docker Hub usam a tag latest:
+#    - marcelolealhub/evo-nexus-dashboard:latest
+#    - marcelolealhub/evo-nexus-runtime:latest
+# ============================================================
+
+services:
+
+  evonexus_dashboard:
+    image: marcelolealhub/evo-nexus-dashboard:latest
+
+    volumes:
+      - evonexus_config:/workspace/config
+      - evonexus_workspace:/workspace/workspace
+      - evonexus_dashboard_data:/workspace/dashboard/data
+      - evonexus_memory:/workspace/memory
+      - evonexus_adw_logs:/workspace/ADWs/logs
+      - evonexus_agent_memory:/workspace/.claude/agent-memory
+      - evonexus_claude_auth:/root/.claude
+      - evonexus_codex_auth:/root/.codex
+
+    networks:
+      - network_public
+
+    environment:
+      - TZ=America/Sao_Paulo
+      - EVONEXUS_PORT=8080
+      - TERMINAL_SERVER_PORT=32352
+      - FORWARDED_ALLOW_IPS=*
+
+    deploy:
+      mode: replicated
+      replicas: 1
+      restart_policy:
+        condition: on-failure
+      placement:
+        constraints:
+          - node.role == manager
+      resources:
+        limits:
+          cpus: "1"
+          memory: 1024M
+      labels:
+        - traefik.enable=true
+        - traefik.docker.network=network_public
+
+        - traefik.http.routers.evonexus.rule=Host(`evonexus.advancedbot.com.br`)
+        - traefik.http.routers.evonexus.entrypoints=websecure
+        - traefik.http.routers.evonexus.priority=1
+        - traefik.http.routers.evonexus.tls.certresolver=letsencryptresolver
+        - traefik.http.routers.evonexus.service=evonexus
+        - traefik.http.services.evonexus.loadbalancer.server.port=8080
+        - traefik.http.services.evonexus.loadbalancer.passHostHeader=true
+
+        - traefik.http.routers.evonexus-terminal.rule=Host(`evonexus.advancedbot.com.br`) && PathPrefix(`/terminal`)
+        - traefik.http.routers.evonexus-terminal.entrypoints=websecure
+        - traefik.http.routers.evonexus-terminal.priority=10
+        - traefik.http.routers.evonexus-terminal.tls.certresolver=letsencryptresolver
+        - traefik.http.routers.evonexus-terminal.service=evonexus-terminal
+        - traefik.http.routers.evonexus-terminal.middlewares=evonexus-terminal-strip
+        - traefik.http.middlewares.evonexus-terminal-strip.stripprefix.prefixes=/terminal
+        - traefik.http.services.evonexus-terminal.loadbalancer.server.port=32352
+        - traefik.http.services.evonexus-terminal.loadbalancer.passHostHeader=true
+
+  evonexus_telegram:
+    image: marcelolealhub/evo-nexus-runtime:latest
+    command:
+      - "claude"
+      - "--channels"
+      - "plugin:telegram@claude-plugins-official"
+      - "--dangerously-skip-permissions"
+
+    volumes:
+      - evonexus_config:/workspace/config
+      - evonexus_workspace:/workspace/workspace
+      - evonexus_memory:/workspace/memory
+      - evonexus_adw_logs:/workspace/ADWs/logs
+      - evonexus_agent_memory:/workspace/.claude/agent-memory
+      - evonexus_claude_auth:/root/.claude
+      - evonexus_codex_auth:/root/.codex
+
+    networks:
+      - network_public
+
+    environment:
+      - TZ=America/Sao_Paulo
+      - REQUIRE_ANTHROPIC_KEY=1
+
+    stdin_open: true
+    tty: true
+
+    deploy:
+      mode: replicated
+      replicas: 1
+      restart_policy:
+        condition: on-failure
+      placement:
+        constraints:
+          - node.role == manager
+      resources:
+        limits:
+          cpus: "1"
+          memory: 1024M
+
+  evonexus_scheduler:
+    image: marcelolealhub/evo-nexus-runtime:latest
+    command: ["uv", "run", "python", "scheduler.py"]
+
+    volumes:
+      - evonexus_config:/workspace/config
+      - evonexus_workspace:/workspace/workspace
+      - evonexus_memory:/workspace/memory
+      - evonexus_adw_logs:/workspace/ADWs/logs
+      - evonexus_agent_memory:/workspace/.claude/agent-memory
+      - evonexus_claude_auth:/root/.claude
+      - evonexus_codex_auth:/root/.codex
+
+    networks:
+      - network_public
+
+    environment:
+      - TZ=America/Sao_Paulo
+      - REQUIRE_ANTHROPIC_KEY=1
+
+    deploy:
+      mode: replicated
+      replicas: 1
+      restart_policy:
+        condition: on-failure
+      placement:
+        constraints:
+          - node.role == manager
+      resources:
+        limits:
+          cpus: "1"
+          memory: 1024M
+
+volumes:
+  evonexus_config:
+  evonexus_workspace:
+  evonexus_dashboard_data:
+  evonexus_memory:
+  evonexus_adw_logs:
+  evonexus_agent_memory:
+  evonexus_claude_auth:
+  evonexus_codex_auth:
+
+networks:
+  network_public:
+    external: true

evonexus.stack.yml

@@ -32,6 +32,7 @@ services:
       - evonexus_memory:/workspace/memory
       - evonexus_adw_logs:/workspace/ADWs/logs
       - evonexus_agent_memory:/workspace/.claude/agent-memory
+      - evonexus_claude_auth:/root/.claude
       - evonexus_codex_auth:/root/.codex
 
     networks:
@@ -96,6 +97,7 @@ services:
       - evonexus_memory:/workspace/memory
       - evonexus_adw_logs:/workspace/ADWs/logs
       - evonexus_agent_memory:/workspace/.claude/agent-memory
+      - evonexus_claude_auth:/root/.claude
       - evonexus_codex_auth:/root/.codex
 
     networks:
@@ -155,6 +157,7 @@ volumes:
   evonexus_memory:
   evonexus_adw_logs:
   evonexus_agent_memory:
+  evonexus_claude_auth:
   evonexus_codex_auth:
 
 networks:

start-dashboard.sh

@@ -22,6 +22,59 @@ FLASK_PORT="${EVONEXUS_PORT:-8080}"
 
 echo "[start-dashboard] terminal-server on :${TERMINAL_PORT}, Flask on :${FLASK_PORT}"
 
+# ----------------------------------------------------------------------------
+# Pre-seed Claude Code global settings so the first-run theme/onboarding
+# prompts are skipped on every new agent terminal. Each agent runs in its
+# own working directory, which Claude Code treats as a separate project —
+# without this, the user has to pick a theme on every single agent.
+# Only writes the file if it doesn't already exist (preserves user choices).
+# ----------------------------------------------------------------------------
+mkdir -p /root/.claude
+if [ ! -f /root/.claude/settings.json ]; then
+    echo "[start-dashboard] seeding /root/.claude/settings.json with default theme"
+    cat > /root/.claude/settings.json <<'EOF'
+{
+  "theme": "dark",
+  "hasCompletedOnboarding": true,
+  "hasSeenWelcome": true,
+  "telemetry": false
+}
+EOF
+fi
+
+# ----------------------------------------------------------------------------
+# Restore /root/.claude.json from the most recent backup when missing.
+#
+# Claude Code's main config (theme, OAuth tokens, per-project state) lives
+# at /root/.claude.json — a SIBLING of the /root/.claude/ directory, NOT
+# inside it. The Swarm volume mounts /root/.claude/, so /root/.claude.json
+# sits in the container's writable layer and is wiped on every redeploy.
+# Result: theme picker and onboarding reappear on every release.
+#
+# Claude Code itself writes timestamped backups into /root/.claude/backups/
+# (which IS in the volume). We just need to restore the latest on startup
+# if the main file is missing. If no backup exists either, seed a minimal
+# config so the first-run prompts are skipped.
+# ----------------------------------------------------------------------------
+if [ ! -f /root/.claude.json ]; then
+    latest_backup=$(ls -t /root/.claude/backups/.claude.json.backup.* 2>/dev/null | head -n1 || true)
+    if [ -n "${latest_backup:-}" ] && [ -f "${latest_backup}" ]; then
+        echo "[start-dashboard] restoring /root/.claude.json from ${latest_backup}"
+        cp "${latest_backup}" /root/.claude.json
+    else
+        echo "[start-dashboard] seeding minimal /root/.claude.json (no backup found)"
+        cat > /root/.claude.json <<'EOF'
+{
+  "theme": "dark",
+  "hasCompletedOnboarding": true,
+  "hasSeenWelcome": true,
+  "bypassPermissionsModeAccepted": true,
+  "telemetry": false
+}
+EOF
+    fi
+fi
+
 # Start terminal-server in the background
 node /workspace/dashboard/terminal-server/bin/server.js --port "${TERMINAL_PORT}" &
 TERMINAL_PID=$!