refactor(sandbox): unified daemon across freestyle + docker + k8s

.github/workflows/test.yml

@@ -84,6 +84,9 @@ jobs:
       - name: Install dependencies
         run: bun install
 
+      - name: Build daemon bundle
+        run: bun run --cwd=packages/sandbox build
+
       - name: Start NATS with JetStream
         run: |
           docker run -d --name nats -p 4222:4222 nats:2.10 -js
@@ -135,3 +138,51 @@ jobs:
 
       - name: Run knip
         run: bun run knip
+
+  docker-smoke:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      - name: Setup Bun
+        uses: oven-sh/setup-bun@v2
+        with:
+          bun-version: "1.3.5"
+
+      - name: Install dependencies
+        run: bun install
+
+      - name: Build daemon bundle
+        run: bun run --cwd=packages/sandbox build
+
+      - name: Build sandbox image
+        run: |
+          docker build \
+            -t mesh-sandbox:ci \
+            -f packages/sandbox/image/Dockerfile \
+            packages/sandbox
+
+      - name: Smoke test
+        run: |
+          docker run -d --name sandbox-smoke -p 19999:9000 \
+            -e DAEMON_TOKEN="$(printf 't%.0s' {1..32})" \
+            -e DAEMON_BOOT_ID="ci-smoke" \
+            -e APP_ROOT=/app \
+            -e PROXY_PORT=9000 \
+            -e DAEMON_NO_AUTOSTART=1 \
+            mesh-sandbox:ci
+          for i in $(seq 1 30); do
+            if curl -fsS http://localhost:19999/health | grep -q '"bootId":"ci-smoke"'; then
+              echo "ok"
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "smoke test failed — daemon did not return /health with ci-smoke bootId"
+          docker logs sandbox-smoke
+          exit 1
+
+      - name: Tear down
+        if: always()
+        run: docker rm -f sandbox-smoke || true

apps/mesh/package.json

@@ -143,7 +143,7 @@
     "kysely-pglite": "^0.6.1",
     "lucide-react": "^0.468.0",
     "marked": "^15.0.6",
-    "mesh-plugin-user-sandbox": "workspace:*",
+    "@decocms/sandbox": "workspace:*",
     "mesh-plugin-workflows": "workspace:*",
     "nanoid": "^5.1.6",
     "pg": "^8.16.3",

apps/mesh/spec/monitoring-share-plugin.md

@@ -269,7 +269,7 @@ setup: (ctx) => {
 |------|--------|
 | `packages/bindings/src/core/plugins.ts` | Add `rootRoute` and `registerPublicRoutes` to context |
 | `apps/mesh/src/web/index.tsx` | Pass new context props, collect and mount public routes |
-| `packages/mesh-plugin-user-sandbox/client/index.ts` | Migrate connect route registration |
+| `packages/@decocms/sandbox/client/index.ts` | Migrate connect route registration |
 | `apps/mesh/src/web/routes/connect.tsx` | Remove (or keep as fallback) |
 
 ### Phase 2 & 3 (Plugin)

apps/mesh/src/api/app.ts

@@ -45,7 +45,6 @@ import oauthProxyRoutes, {
 } from "./routes/oauth-proxy";
 import openaiCompatRoutes from "./routes/openai-compat";
 import proxyRoutes from "./routes/proxy";
-import { createSandboxDaemonRoutes } from "./routes/sandbox-daemon";
 import { createKVRoutes } from "./routes/kv";
 import { createTriggerCallbackRoutes } from "./routes/trigger-callback";
 import publicConfigRoutes from "./routes/public-config";
@@ -1361,10 +1360,6 @@ export async function createApp(options: CreateAppOptions = {}) {
   });
   app.route("/api", decopilotRoutes);
 
-  // Daemon control-plane passthrough only — dev-server traffic bypasses
-  // mesh and hits pods' public URLs directly.
-  app.route("/", createSandboxDaemonRoutes());
-
   // Stable file redirect endpoint (resolves mesh-storage: URIs to presigned URLs)
   app.route("/api", filesRoutes);
 
@@ -1591,7 +1586,7 @@ export async function createApp(options: CreateAppOptions = {}) {
     const dockerRunner = asDockerRunner(getSharedRunnerIfInit());
     if (dockerRunner) {
       const { sweepDockerOrphansOnShutdown } = await import(
-        "mesh-plugin-user-sandbox/runner"
+        "@decocms/sandbox/runner"
       );
       await sweepDockerOrphansOnShutdown(dockerRunner);
     }

apps/mesh/src/api/routes/decopilot/built-in-tools/index.ts

@@ -15,7 +15,7 @@ import { createReadResourceTool } from "./resources";
 import { createSandboxTool, type VirtualClient } from "./sandbox";
 import { createVmTools } from "./vm-tools";
 import { getRunnerByKind } from "@/sandbox/lifecycle";
-import type { RunnerKind } from "mesh-plugin-user-sandbox/runner";
+import type { RunnerKind } from "@decocms/sandbox/runner";
 import { createSubtaskTool } from "./subtask";
 import { userAskTool } from "./user-ask";
 import { proposePlanTool } from "./propose-plan";

apps/mesh/src/api/routes/decopilot/built-in-tools/vm-tools/index.ts

@@ -2,13 +2,13 @@
  * VM File Tools — runner-agnostic.
  *
  * Registers the six LLM-visible tools (read/write/edit/grep/glob/bash) on
- * top of any `SandboxRunner.proxyDaemonRequest`. Path scheme is Docker's
- * canonical `/_daemon/fs/<op>` + `/_daemon/bash`; non-Docker runners
- * translate inside `proxyDaemonRequest` (see Freestyle's `translateDaemonPath`).
+ * top of any `SandboxRunner.proxyDaemonRequest`. All runners speak the
+ * unified `/_decopilot_vm/*` surface with base64-wrapped JSON bodies
+ * (Cloudflare WAF bypass; harmless 33% overhead on non-CF paths).
  */
 
 import { tool, zodSchema } from "ai";
-import type { SandboxRunner } from "mesh-plugin-user-sandbox/runner";
+import type { SandboxRunner } from "@decocms/sandbox/runner";
 import { maybeTruncate } from "./common";
 import {
   BASH_DESCRIPTION,
@@ -37,10 +37,13 @@ async function daemonRequest(
 ): Promise<unknown> {
   let res: Response;
   try {
+    const b64Body = Buffer.from(JSON.stringify(body), "utf-8").toString(
+      "base64",
+    );
     res = await runner.proxyDaemonRequest(handle, path, {
       method: "POST",
       headers: new Headers({ "content-type": "application/json" }),
-      body: JSON.stringify(body),
+      body: b64Body,
     });
   } catch {
     throw new Error(
@@ -98,7 +101,7 @@ export function createVmTools(params: VmToolsParams) {
     description: READ_DESCRIPTION,
     inputSchema: zodSchema(ReadInputSchema),
     execute: async (input) => {
-      const result = await call("/_daemon/fs/read", input);
+      const result = await call("/_decopilot_vm/read", input);
       return maybeTruncate(result, toolOutputMap);
     },
   });
@@ -107,22 +110,22 @@ export function createVmTools(params: VmToolsParams) {
     needsApproval: approvalFor(TOOL_APPROVAL.write),
     description: WRITE_DESCRIPTION,
     inputSchema: zodSchema(WriteInputSchema),
-    execute: async (input) => call("/_daemon/fs/write", input),
+    execute: async (input) => call("/_decopilot_vm/write", input),
   });
 
   const edit = tool({
     needsApproval: approvalFor(TOOL_APPROVAL.edit),
     description: EDIT_DESCRIPTION,
     inputSchema: zodSchema(EditInputSchema),
-    execute: async (input) => call("/_daemon/fs/edit", input),
+    execute: async (input) => call("/_decopilot_vm/edit", input),
   });
 
   const grep = tool({
     needsApproval: approvalFor(TOOL_APPROVAL.grep),
     description: GREP_DESCRIPTION,
     inputSchema: zodSchema(GrepInputSchema),
     execute: async (input) => {
-      const result = await call("/_daemon/fs/grep", input);
+      const result = await call("/_decopilot_vm/grep", input);
       return maybeTruncate(result, toolOutputMap);
     },
   });
@@ -132,7 +135,7 @@ export function createVmTools(params: VmToolsParams) {
     description: GLOB_DESCRIPTION,
     inputSchema: zodSchema(GlobInputSchema),
     execute: async (input) => {
-      const result = await call("/_daemon/fs/glob", input);
+      const result = await call("/_decopilot_vm/glob", input);
       return maybeTruncate(result, toolOutputMap);
     },
   });
@@ -142,7 +145,7 @@ export function createVmTools(params: VmToolsParams) {
     description: BASH_DESCRIPTION,
     inputSchema: zodSchema(BashInputSchema),
     execute: async (input) => {
-      const result = await call("/_daemon/bash", input);
+      const result = await call("/_decopilot_vm/bash", input);
       return maybeTruncate(result, toolOutputMap);
     },
   });

apps/mesh/src/api/routes/decopilot/built-in-tools/vm-tools/types.ts

@@ -1,4 +1,4 @@
-import type { SandboxRunner } from "mesh-plugin-user-sandbox/runner";
+import type { SandboxRunner } from "@decocms/sandbox/runner";
 
 export interface VmToolsParams {
   readonly runner: SandboxRunner;

apps/mesh/src/api/routes/decopilot/memory.test.ts

@@ -0,0 +1,45 @@
+import { describe, it, expect, beforeAll, afterAll } from "bun:test";
+import { createMemory } from "./memory";
+import {
+  buildThreadTestContext,
+  type ThreadTestEnv,
+} from "../../../tools/thread/test-helpers";
+
+describe("createMemory", () => {
+  let env: ThreadTestEnv;
+
+  beforeAll(async () => {
+    env = await buildThreadTestContext();
+  });
+  afterAll(async () => {
+    await env.close();
+  });
+
+  it("returns Memory when thread exists", async () => {
+    const thread = await env.ctx.storage.threads.create({
+      id: "thrd_existing",
+      organization_id: env.orgId,
+      title: "ok",
+      created_by: env.userId,
+      virtual_mcp_id: "vmcp_x",
+    });
+
+    const memory = await createMemory(env.ctx.storage.threads, {
+      thread_id: thread.id,
+      organization_id: env.orgId,
+      userId: env.userId,
+    });
+
+    expect(memory.thread.id).toBe("thrd_existing");
+  });
+
+  it("throws when thread_id is provided but thread does not exist", async () => {
+    await expect(
+      createMemory(env.ctx.storage.threads, {
+        thread_id: "thrd_does_not_exist",
+        organization_id: env.orgId,
+        userId: env.userId,
+      }),
+    ).rejects.toThrow(/thread.*not.*found/i);
+  });
+});

apps/mesh/src/api/routes/decopilot/memory.ts

@@ -7,14 +7,13 @@
 
 import type { OrgScopedThreadStorage } from "@/storage/threads";
 import type { Thread, ThreadMessage } from "@/storage/types";
-import { generatePrefixedId } from "@/shared/utils/generate-id";
 
 /**
  * Configuration for creating a Memory instance
  */
 export interface MemoryConfig {
-  /** Thread ID (creates new if not found) */
-  thread_id?: string | null;
+  /** Thread ID (required — thread must exist) */
+  thread_id: string;
 
   /** Organization scope */
   organization_id: string;
@@ -24,19 +23,6 @@ export interface MemoryConfig {
 
   /** Default window size for pruning */
   defaultWindowSize?: number;
-
-  /** Optional trigger ID for automation-created threads */
-  triggerId?: string;
-
-  /** Virtual MCP ID to associate with the thread */
-  virtualMcpId?: string;
-
-  /**
-   * Git branch to pin this thread to. Only meaningful for GitHub-linked
-   * virtualmcps. When set on a brand-new thread, it's persisted on the
-   * thread row and propagates to VM_START.
-   */
-  branch?: string | null;
 }
 
 /**
@@ -88,53 +74,23 @@ export class Memory {
 }
 
 /**
- * Create or get a thread, returning a Memory instance
+ * Get an existing thread by id, returning a Memory instance.
+ * Throws if the thread does not exist — the route loader is responsible for
+ * creating threads up-front via COLLECTION_THREADS_CREATE.
  */
 export async function createMemory(
   storage: OrgScopedThreadStorage,
   config: MemoryConfig,
 ): Promise<Memory> {
-  const {
-    thread_id,
-    organization_id,
-    userId,
-    defaultWindowSize,
-    triggerId,
-    virtualMcpId,
-    branch,
-  } = config;
-
-  let thread: Thread;
+  const { thread_id, defaultWindowSize } = config;
 
   if (!thread_id) {
-    // Create new thread
-    thread = await storage.create({
-      id: generatePrefixedId("thrd"),
-      organization_id,
-      created_by: userId,
-      trigger_id: triggerId ?? null,
-      virtual_mcp_id: virtualMcpId ?? "",
-      branch: branch ?? null,
-    });
-  } else {
-    // Try to get existing thread scoped to this org
-    const existing = await storage.get(thread_id);
+    throw new Error("createMemory: thread_id is required");
+  }
 
-    if (existing) {
-      thread = existing;
-    } else {
-      // Thread not found — create using the client-provided ID so the
-      // frontend and server stay in sync (avoids a thread-ID switch in
-      // onFinish which causes a full re-render cascade).
-      thread = await storage.create({
-        id: thread_id,
-        organization_id,
-        created_by: userId,
-        trigger_id: triggerId ?? null,
-        virtual_mcp_id: virtualMcpId ?? "",
-        branch: branch ?? null,
-      });
-    }
+  const thread = await storage.get(thread_id);
+  if (!thread) {
+    throw new Error(`Thread not found: ${thread_id}`);
   }
 
   return new Memory({

apps/mesh/src/api/routes/decopilot/run-registry.test.ts

@@ -421,7 +421,7 @@ describe("RunRegistry", () => {
             created_at: "",
             updated_at: "",
             created_by: "u1",
-            updated_by: null,
+            updated_by: undefined,
             hidden: false,
             context_start_message_id: null,
             run_owner_pod: null,
@@ -455,7 +455,7 @@ describe("RunRegistry", () => {
             created_at: "",
             updated_at: "",
             created_by: "u1",
-            updated_by: null,
+            updated_by: undefined,
             hidden: false,
             context_start_message_id: null,
             run_owner_pod: null,
@@ -489,7 +489,7 @@ describe("RunRegistry", () => {
             created_at: "",
             updated_at: "",
             created_by: "u1",
-            updated_by: null,
+            updated_by: undefined,
             hidden: false,
             context_start_message_id: null,
             run_owner_pod: null,
@@ -523,7 +523,7 @@ describe("RunRegistry", () => {
             created_at: "",
             updated_at: "",
             created_by: "u1",
-            updated_by: null,
+            updated_by: undefined,
             hidden: false,
             context_start_message_id: null,
             run_owner_pod: null,
@@ -637,7 +637,7 @@ describe("RunRegistry", () => {
       created_at: "",
       updated_at: "",
       created_by: "u1",
-      updated_by: null,
+      updated_by: undefined,
       hidden: false,
       context_start_message_id: null,
       run_owner_pod: "dead-pod",