fix(deploy): many docker compose bug

DEPLOYMENT.md

@@ -127,7 +127,7 @@ host = "postgres"
 password = "your_secure_password"  # Must change in config.toml
 
 [containerd]
-socket_path = "unix:///var/run/docker.sock"  # Use host Docker
+socket_path = "/run/containerd/containerd.sock"
 
 [qdrant]
 base_url = "http://qdrant:6334"

agent/src/config.ts

@@ -4,6 +4,7 @@ import { parse } from 'toml'
 type AgentGatewayConfig = {
   'agent_gateway': {
     host?: string
+    server_addr?: string
     port?: number
   },
   'server': {

agent/src/index.ts

@@ -3,7 +3,6 @@ import { chatModule } from './modules/chat'
 import { corsMiddleware } from './middlewares/cors'
 import { errorMiddleware } from './middlewares/error'
 import { loadConfig } from './config'
-import { join } from 'path'
 
 const config = loadConfig('../config.toml')
 
@@ -15,17 +14,26 @@ export const getBraveConfig = () => {
 }
 
 export const getBaseUrl = () => {
-  let baseUrl = ''
-  if (!baseUrl) {
-    baseUrl = 'http://127.0.0.1'
+  const rawAddr =
+    typeof config.agent_gateway.server_addr === 'string'
+      ? config.agent_gateway.server_addr.trim()
+      : typeof config.server.addr === 'string'
+        ? config.server.addr.trim()
+        : ''
+
+  if (!rawAddr) {
+    return 'http://127.0.0.1'
   }
-  if (
-    typeof config.server.addr === 'string' &&
-    config.server.addr.startsWith(':')
-  ) {
-    baseUrl = `http://127.0.0.1${config.server.addr}`
+
+  if (rawAddr.startsWith('http://') || rawAddr.startsWith('https://')) {
+    return rawAddr.replace(/\/+$/, '')
   }
-  return baseUrl
+
+  if (rawAddr.startsWith(':')) {
+    return `http://127.0.0.1${rawAddr}`
+  }
+
+  return `http://${rawAddr}`
 }
 
 export type AuthFetcher = (
@@ -40,7 +48,12 @@ export const createAuthFetcher = (bearer: string | undefined): AuthFetcher => {
       headers.set('Authorization', `Bearer ${bearer}`)
     }
 
-    return await fetch(join(getBaseUrl(), url), {
+    const requestUrl = new URL(
+      url,
+      `${getBaseUrl().replace(/\/+$/, '')}/`,
+    ).toString()
+
+    return await fetch(requestUrl, {
       ...requestOptions,
       headers,
     })
@@ -50,6 +63,9 @@ export const createAuthFetcher = (bearer: string | undefined): AuthFetcher => {
 const app = new Elysia()
   .use(corsMiddleware)
   .use(errorMiddleware)
+  .get('/health', () => ({
+    status: 'ok',
+  }))
   .use(chatModule)
   .listen({
     port: config.agent_gateway.port ?? 8081,

config.toml.example

@@ -50,6 +50,7 @@ timeout_seconds = 10
 [agent_gateway]
 host = "127.0.0.1"
 port = 8081
+server_addr = ":8080"
 
 [brave]
 api_key = ""

deploy.sh

@@ -44,12 +44,39 @@ if [ ! -f config.toml ]; then
     echo ""
 fi
 
-# Build MCP image
-echo -e "${GREEN}Building MCP image...${NC}"
-if docker build -f docker/Dockerfile.mcp -t memoh-mcp:latest . > /dev/null 2>&1; then
-    echo -e "${GREEN}✓ MCP image built successfully${NC}"
+# Prepare data root path for host/containerd compatibility
+MEMOH_DATA_ROOT="$(pwd)/.data/memoh"
+mkdir -p "${MEMOH_DATA_ROOT}"
+export MEMOH_DATA_ROOT
+if grep -q '^data_root[[:space:]]*=' config.toml; then
+    awk -v path="${MEMOH_DATA_ROOT}" '
+        $0 ~ /^data_root[[:space:]]*=/ { print "data_root = \"" path "\""; next }
+        { print }
+    ' config.toml > config.toml.tmp && mv config.toml.tmp config.toml
+fi
+echo -e "${GREEN}✓ Data root: ${MEMOH_DATA_ROOT}${NC}"
+echo ""
+
+# Prepare container runtime environment
+echo -e "${GREEN}Preparing container runtime environment...${NC}"
+if sh scripts/containerd-install.sh > /dev/null 2>&1; then
+    echo -e "${GREEN}✓ Container runtime environment is ready${NC}"
 else
-    echo -e "${YELLOW}⚠ MCP image build failed, will try to pull at runtime${NC}"
+    echo -e "${YELLOW}⚠ Failed to prepare container runtime environment, MCP build may be skipped${NC}"
+fi
+echo ""
+
+# Build MCP image on host with nerdctl
+MCP_IMAGE="docker.io/library/memoh-mcp:latest"
+echo -e "${GREEN}Building MCP image on host with nerdctl...${NC}"
+if command -v nerdctl &> /dev/null && command -v buildctl &> /dev/null && command -v buildkitd &> /dev/null; then
+    if nerdctl build -f docker/Dockerfile.mcp -t "${MCP_IMAGE}" . > /dev/null 2>&1; then
+        echo -e "${GREEN}✓ MCP image built successfully (on host)${NC}"
+    else
+        echo -e "${YELLOW}⚠ MCP image build failed on host, will try to pull at runtime${NC}"
+    fi
+else
+    echo -e "${YELLOW}⚠ nerdctl/buildkit environment not found on host, skipping MCP build${NC}"
 fi
 echo ""
 

docker-compose.yml

@@ -1,5 +1,5 @@
+name: "memoh"
 services:
-  
   postgres:
     image: postgres:18.1-alpine
     container_name: memoh-postgres
@@ -8,7 +8,7 @@ services:
       POSTGRES_USER: memoh
       POSTGRES_PASSWORD: memoh123
     volumes:
-      - postgres_data:/var/lib/postgresql/data
+      - postgres_data:/var/lib/postgresql
       - ./db/migrations:/docker-entrypoint-initdb.d:ro
     expose:
       - "5432"
@@ -38,26 +38,24 @@ services:
     networks:
       - memoh-network
 
-  docker-cli:
-    image: docker:27-cli
-    container_name: memoh-docker-cli
-    volumes:
-      - /var/run/docker.sock:/var/run/docker.sock
-      - memoh_bot_data:/var/lib/memoh/data
-    command: ["tail", "-f", "/dev/null"]
-    restart: unless-stopped
-    networks:
-      - memoh-network
-
   server:
     build:
-      context: ./docker
-      dockerfile: Dockerfile.server
+      context: .
+      dockerfile: docker/Dockerfile.server
     container_name: memoh-server
+    pid: host
     volumes:
       - ./config.toml:/app/config.toml:ro
-      - /var/run/docker.sock:/var/run/docker.sock
-      - memoh_bot_data:/var/lib/memoh/data
+      - /run/containerd/containerd.sock:/run/containerd/containerd.sock
+      - /var/lib/containerd:/var/lib/containerd
+      - server_cni_state:/var/lib/cni
+      - ${MEMOH_DATA_ROOT:-/opt/memoh/data}:${MEMOH_DATA_ROOT:-/opt/memoh/data}
+    cap_add:
+      - SYS_ADMIN
+      - NET_ADMIN
+    security_opt:
+      - seccomp:unconfined
+      - apparmor:unconfined
     ports:
       - "8080:8080"
     depends_on:
@@ -106,7 +104,7 @@ volumes:
     driver: local
   qdrant_data:
     driver: local
-  memoh_bot_data:
+  server_cni_state:
     driver: local
 
 networks:

docker/Dockerfile.server

@@ -17,11 +17,43 @@ FROM alpine:latest
 
 WORKDIR /app
 
-RUN apk add --no-cache ca-certificates tzdata wget
+RUN apk add --no-cache ca-certificates tzdata wget nerdctl cni-plugins iptables \
+    && mkdir -p /opt/cni/bin \
+    && (cp -a /usr/lib/cni/. /opt/cni/bin/ 2>/dev/null || true) \
+    && (cp -a /usr/libexec/cni/. /opt/cni/bin/ 2>/dev/null || true) \
+    && mkdir -p /etc/cni/net.d /var/lib/cni \
+    && printf '%s\n' \
+      '{' \
+      '  "cniVersion": "1.0.0",' \
+      '  "name": "memoh-cni",' \
+      '  "plugins": [' \
+      '    {' \
+      '      "type": "bridge",' \
+      '      "bridge": "cni0",' \
+      '      "isGateway": true,' \
+      '      "ipMasq": true,' \
+      '      "promiscMode": true,' \
+      '      "ipam": {' \
+      '        "type": "host-local",' \
+      '        "ranges": [[' \
+      '          { "subnet": "10.88.0.0/16" }' \
+      '        ]],' \
+      '        "routes": [' \
+      '          { "dst": "0.0.0.0/0" }' \
+      '        ]' \
+      '      }' \
+      '    },' \
+      '    {' \
+      '      "type": "portmap",' \
+      '      "capabilities": { "portMappings": true }' \
+      '    }' \
+      '  ]' \
+      '}' > /etc/cni/net.d/10-memoh.conflist
 
 COPY --from=builder /build/memoh-server /app/memoh-server
+COPY --from=builder /build/spec /app/spec
 
-RUN mkdir -p /var/lib/memoh/data
+RUN mkdir -p /opt/memoh/data
 
 EXPOSE 8080
 

docker/Dockerfile.web

@@ -28,6 +28,6 @@ COPY docker/config/nginx.conf /etc/nginx/conf.d/default.conf
 EXPOSE 80
 
 HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
-    CMD wget --no-verbose --tries=1 --spider http://localhost/ || exit 1
+    CMD wget --no-verbose --tries=1 --spider http://localhost/health || exit 1
 
 CMD ["nginx", "-g", "daemon off;"]

docker/config/config.docker.toml

@@ -4,7 +4,7 @@ level = "info"
 format = "text"
 
 [server]
-addr = ":8080"
+addr = "server:8080"
 
 ## Admin
 [admin]
@@ -19,13 +19,13 @@ jwt_expires_in = "168h"
 
 ## Docker configuration
 [containerd]
-socket_path = "unix:///var/run/docker.sock"
+socket_path = "/run/containerd/containerd.sock"
 namespace = "default"
 
 [mcp]
-busybox_image = "memoh-mcp:latest"
+busybox_image = "docker.io/library/memoh-mcp:latest"
 snapshotter = "overlayfs"
-data_root = "/var/lib/memoh/data"
+data_root = "/opt/memoh/data"
 data_mount = "/data"
 
 ## Postgres configuration
@@ -48,6 +48,7 @@ timeout_seconds = 10
 [agent_gateway]
 host = "agent"
 port = 8081
+server_addr = "server:8080"
 
 [brave]
 api_key = ""

docker/config/nginx.conf

@@ -1,5 +1,6 @@
 server {
     listen 80;
+    listen [::]:80;
     server_name _;
     root /usr/share/nginx/html;
     index index.html;
@@ -15,9 +16,14 @@ server {
         try_files $uri $uri/ /index.html;
     }
 
-    # API 代理
-    location /api/ {
-        proxy_pass http://memoh-server:8080/;
+    # Nginx 健康检查
+    location = /health {
+        access_log off;
+        add_header Content-Type text/plain;
+        return 200 "ok\n";
+    }
+
+    # 统一代理参数
     proxy_http_version 1.1;
     proxy_set_header Upgrade $http_upgrade;
     proxy_set_header Connection 'upgrade';
@@ -26,29 +32,23 @@ server {
     proxy_set_header X-Real-IP $remote_addr;
     proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
     proxy_set_header X-Forwarded-Proto $scheme;
-        
-        # 超时设置
     proxy_connect_timeout 60s;
     proxy_send_timeout 60s;
     proxy_read_timeout 60s;
+
+    # Swagger 文档（保留 /api 前缀）
+    location ~ ^/api/(docs(?:/.*)?|swagger\.json)$ {
+        proxy_pass http://memoh-server:8080;
+    }
+
+    # API 代理（其余 /api/* 去掉 /api 前缀转发）
+    location /api/ {
+        proxy_pass http://memoh-server:8080/;
     }
 
     # Agent Gateway 代理
     location /agent/ {
         proxy_pass http://memoh-agent:8081/;
-        proxy_http_version 1.1;
-        proxy_set_header Upgrade $http_upgrade;
-        proxy_set_header Connection 'upgrade';
-        proxy_set_header Host $host;
-        proxy_cache_bypass $http_upgrade;
-        proxy_set_header X-Real-IP $remote_addr;
-        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-        
-        # 超时设置
-        proxy_connect_timeout 60s;
-        proxy_send_timeout 60s;
-        proxy_read_timeout 60s;
     }
 
     # 静态资源缓存

internal/containerd/service.go

@@ -9,7 +9,6 @@ import (
 	"io"
 	"log/slog"
 	"os"
-	"path/filepath"
 	"runtime"
 	"strings"
 	"syscall"
@@ -673,15 +672,8 @@ func (s *DefaultService) ExecTaskStreaming(ctx context.Context, containerID stri
 	if req.Terminal {
 		ioOpts = append(ioOpts, cio.WithTerminal)
 	}
-	fifoDir := strings.TrimSpace(req.FIFODir)
-	if fifoDir == "" {
-		if homeDir, err := os.UserHomeDir(); err == nil && homeDir != "" {
-			fifoDir = filepath.Join(homeDir, ".memoh", "containerd-fifo")
-		} else {
-			fifoDir = "/tmp/memoh-containerd-fifo"
-		}
-	}
-	if err := os.MkdirAll(fifoDir, 0o755); err != nil {
+	fifoDir, err := resolveExecFIFODir(req.FIFODir)
+	if err != nil {
 		_ = stdinR.Close()
 		_ = stdinW.Close()
 		_ = stdoutR.Close()
@@ -752,6 +744,27 @@ func (s *DefaultService) ExecTaskStreaming(ctx context.Context, containerID stri
 	}, nil
 }
 
+func resolveExecFIFODir(preferred string) (string, error) {
+	candidates := make([]string, 0, 3)
+	if p := strings.TrimSpace(preferred); p != "" {
+		candidates = append(candidates, p)
+	}
+	candidates = append(candidates, "/var/lib/containerd/memoh-fifo", "/tmp/memoh-containerd-fifo")
+
+	var lastErr error
+	for _, dir := range candidates {
+		if err := os.MkdirAll(dir, 0o755); err == nil {
+			return dir, nil
+		} else {
+			lastErr = err
+		}
+	}
+	if lastErr == nil {
+		lastErr = fmt.Errorf("no fifo directory candidate available")
+	}
+	return "", lastErr
+}
+
 func (s *DefaultService) ListContainersByLabel(ctx context.Context, key, value string) ([]containerd.Container, error) {
 	if key == "" {
 		return nil, ErrInvalidArgument

internal/handlers/fs.go

@@ -4,11 +4,13 @@ import (
 	"bufio"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"log/slog"
 	"net/http"
 	"os/exec"
+	"path/filepath"
 	"runtime"
 	"strings"
 	"sync"
@@ -189,6 +191,7 @@ func (h *ContainerdHandler) getMCPSession(ctx context.Context, containerID strin
 func (h *ContainerdHandler) startContainerdMCPSession(ctx context.Context, containerID string) (*mcpSession, error) {
 	execSession, err := h.service.ExecTaskStreaming(ctx, containerID, ctr.ExecTaskRequest{
 		Args:    []string{"/app/mcp"},
+		FIFODir: h.mcpFIFODir(),
 	})
 	if err != nil {
 		return nil, err
@@ -207,11 +210,15 @@ func (h *ContainerdHandler) startContainerdMCPSession(ctx context.Context, conta
 	go func() {
 		_, err := execSession.Wait()
 		if err != nil {
+			if isBenignMCPSessionExit(err) {
+				sess.closeWithError(io.EOF)
+				return
+			}
 			h.logger.Error("mcp session exited", slog.Any("error", err), slog.String("container_id", containerID))
 			sess.closeWithError(err)
-		} else {
-			sess.closeWithError(io.EOF)
+			return
 		}
+		sess.closeWithError(io.EOF)
 	}()
 
 	return sess, nil
@@ -273,11 +280,15 @@ func (h *ContainerdHandler) startLimaMCPSession(containerID string) (*mcpSession
 	go sess.readLoop()
 	go func() {
 		if err := cmd.Wait(); err != nil {
+			if isBenignMCPSessionExit(err) {
+				sess.closeWithError(io.EOF)
+				return
+			}
 			h.logger.Error("mcp session exited", slog.Any("error", err), slog.String("container_id", containerID))
 			sess.closeWithError(err)
-		} else {
-			sess.closeWithError(io.EOF)
+			return
 		}
+		sess.closeWithError(io.EOF)
 	}()
 
 	return sess, nil
@@ -320,11 +331,32 @@ func (h *ContainerdHandler) startMCPStderrLogger(stderr io.ReadCloser, container
 			h.logger.Warn("mcp stderr", slog.String("container_id", containerID), slog.String("message", line))
 		}
 		if err := scanner.Err(); err != nil {
+			if errors.Is(err, io.EOF) || errors.Is(err, io.ErrClosedPipe) || strings.Contains(err.Error(), "closed pipe") {
+				return
+			}
 			h.logger.Error("mcp stderr read failed", slog.Any("error", err), slog.String("container_id", containerID))
 		}
 	}()
 }
 
+func isBenignMCPSessionExit(err error) bool {
+	if err == nil {
+		return false
+	}
+	if errors.Is(err, context.Canceled) || errors.Is(err, io.EOF) || errors.Is(err, io.ErrClosedPipe) {
+		return true
+	}
+	msg := strings.ToLower(err.Error())
+	return strings.Contains(msg, "code = canceled") || strings.Contains(msg, "context canceled") || strings.Contains(msg, "closed pipe")
+}
+
+func (h *ContainerdHandler) mcpFIFODir() string {
+	if root := strings.TrimSpace(h.cfg.DataRoot); root != "" {
+		return filepath.Join(root, ".containerd-fifo")
+	}
+	return "/tmp/memoh-containerd-fifo"
+}
+
 func (s *mcpSession) readLoop() {
 	scanner := bufio.NewScanner(s.stdout)
 	scanner.Buffer(make([]byte, 0, 64*1024), 8*1024*1024)

internal/handlers/mcp_stdio.go

@@ -178,6 +178,7 @@ func (h *ContainerdHandler) startContainerdMCPCommandSession(ctx context.Context
 		Args:    args,
 		Env:     env,
 		WorkDir: strings.TrimSpace(req.Cwd),
+		FIFODir: h.mcpFIFODir(),
 	})
 	if err != nil {
 		return nil, err
@@ -195,11 +196,15 @@ func (h *ContainerdHandler) startContainerdMCPCommandSession(ctx context.Context
 	go func() {
 		_, err := execSession.Wait()
 		if err != nil {
+			if isBenignMCPSessionExit(err) {
+				sess.closeWithError(io.EOF)
+				return
+			}
 			h.logger.Error("mcp stdio session exited", slog.Any("error", err), slog.String("container_id", containerID))
 			sess.closeWithError(err)
-		} else {
-			sess.closeWithError(io.EOF)
+			return
 		}
+		sess.closeWithError(io.EOF)
 	}()
 	return sess, nil
 }
@@ -342,11 +347,15 @@ func (h *ContainerdHandler) startLimaMCPCommandSession(containerID string, req M
 	go sess.readLoop()
 	go func() {
 		if err := cmd.Wait(); err != nil {
+			if isBenignMCPSessionExit(err) {
+				sess.closeWithError(io.EOF)
+				return
+			}
 			h.logger.Error("mcp stdio session exited", slog.Any("error", err), slog.String("container_id", containerID))
 			sess.closeWithError(err)
-		} else {
-			sess.closeWithError(io.EOF)
+			return
 		}
+		sess.closeWithError(io.EOF)
 	}()
 
 	return sess, nil

internal/handlers/ping.go

@@ -17,6 +17,7 @@ func NewPingHandler(log *slog.Logger) *PingHandler {
 
 func (h *PingHandler) Register(e *echo.Echo) {
 	e.GET("/ping", h.Ping)
+	e.HEAD("/health", h.PingHead)
 }
 
 func (h *PingHandler) Ping(c echo.Context) error {
@@ -24,3 +25,7 @@ func (h *PingHandler) Ping(c echo.Context) error {
 		"status": "ok",
 	})
 }
+
+func (h *PingHandler) PingHead(c echo.Context) error {
+	return c.NoContent(http.StatusOK)
+}

internal/server/server.go

@@ -42,7 +42,7 @@ func NewServer(log *slog.Logger, addr string, jwtSecret string, pingHandler *han
 	}))
 	e.Use(auth.JWTMiddleware(jwtSecret, func(c echo.Context) bool {
 		path := c.Request().URL.Path
-		if path == "/ping" || path == "/api/swagger.json" || path == "/auth/login" {
+		if path == "/ping" || path == "/health" || path == "/api/swagger.json" || path == "/auth/login" {
 			return true
 		}
 		if strings.HasPrefix(path, "/api/docs") {

scripts/containerd-install.sh

@@ -7,7 +7,10 @@ if [ "$(uname -s)" = "Darwin" ]; then
   exit $?
 fi
 
-if command -v containerd >/dev/null 2>&1 && command -v nerdctl >/dev/null 2>&1 && command -v buildctl >/dev/null 2>&1 && command -v buildkitd >/dev/null 2>&1; then
+if command -v containerd >/dev/null 2>&1 \
+  && command -v nerdctl >/dev/null 2>&1 \
+  && command -v buildctl >/dev/null 2>&1 \
+  && command -v buildkitd >/dev/null 2>&1; then
   containerd --version
   nerdctl --version
   buildctl --version