fix staging and production deploys

lalalune · lalalune · commit df446aac4e60 · 2026-05-16T01:24:42.000-07:00
diff --git a/.github/workflows/cloud-deploy-backend.yml b/.github/workflows/cloud-deploy-backend.yml
@@ -76,12 +76,12 @@ jobs:
         run: |
           for attempt in 1 2 3; do
             if [ "$attempt" -eq 3 ]; then
-              bun install --no-save --verbose && git diff --exit-code -- bun.lock && exit 0
+              bun install --frozen-lockfile --ignore-scripts --verbose && exit 0
             else
-              bun install --no-save && git diff --exit-code -- bun.lock && exit 0
+              bun install --frozen-lockfile --ignore-scripts && exit 0
             fi
 
-            echo "bun install attempt $attempt failed or changed bun.lock; retrying in 10s..."
+            echo "bun install attempt $attempt failed; retrying in 10s..."
             sleep 10
           done
 
@@ -148,12 +148,12 @@ jobs:
         run: |
           for attempt in 1 2 3; do
             if [ "$attempt" -eq 3 ]; then
-              bun install --no-save --verbose && git diff --exit-code -- bun.lock && exit 0
+              bun install --frozen-lockfile --ignore-scripts --verbose && exit 0
             else
-              bun install --no-save && git diff --exit-code -- bun.lock && exit 0
+              bun install --frozen-lockfile --ignore-scripts && exit 0
             fi
 
-            echo "bun install attempt $attempt failed or changed bun.lock; retrying in 10s..."
+            echo "bun install attempt $attempt failed; retrying in 10s..."
             sleep 10
           done
 
@@ -294,7 +294,7 @@ jobs:
               export PATH="$BUN_INSTALL/bin:$PATH"
             fi
             for attempt in 1 2 3; do
-              if bun install --frozen-lockfile; then
+              if bun install --frozen-lockfile --ignore-scripts; then
                 break
               fi
 
diff --git a/.github/workflows/deploy-eliza-provisioning-worker.yml b/.github/workflows/deploy-eliza-provisioning-worker.yml
@@ -87,7 +87,13 @@ jobs:
             rm -f .git/index.lock
 
             # Discard local drift, preserve env files.
-            git clean -fdx -e .env -e .env.local -e .env.* -e node_modules || true
+            git clean -fdx \
+              -e .env \
+              -e .env.local \
+              -e .env.* \
+              -e cloud/.env.local \
+              -e cloud/.env.* \
+              -e node_modules || true
             git checkout -- . 2>/dev/null || true
 
             # bun.lock became tracked at 4ae092dcda; the host's previous deploy
@@ -120,31 +126,20 @@ jobs:
               export PATH="$BUN_INSTALL/bin:$PATH"
             fi
 
-            # cloud/ is a separate bun project (not a sub-workspace of the
-            # outer monorepo, whose `workspaces` is `packages/*`). Install from
-            # /opt/eliza/cloud so the worker's deps end up in cloud/node_modules.
-            #
-            # No --frozen-lockfile: cloud/bun.lock currently has a duplicate
-            # package path for @elizaos/plugin-workflow (workspace + registry
-            # entries for the same name) that bun rejects under frozen mode.
-            # Cloud-CF deploys hit the same upstream bug; until the lockfile is
-            # cleaned at source, allow bun to recompute the deps tree.
-            cd /opt/eliza/cloud
+            cd /opt/eliza
             for attempt in 1 2 3; do
-              if bun install; then
+              if bun install --frozen-lockfile --ignore-scripts; then
                 break
               fi
               if [ "$attempt" -eq 3 ]; then exit 1; fi
               echo "bun install attempt $attempt failed; retrying in 10s..."
               sleep 10
             done
-            cd /opt/eliza
 
             # The daemons run under Node/tsx rather than Bun. Node resolves
             # linked workspace packages through their built `node` exports, so
             # refresh linked package dist files after installing dependencies.
-            bun run --cwd cloud build:linked-core
-            bun run --cwd cloud build:linked-billing
+            bun run build:core
             mkdir -p plugins/plugin-sql/node_modules/@elizaos
             rm -rf plugins/plugin-sql/node_modules/@elizaos/core
             ln -s ../../../../packages/core plugins/plugin-sql/node_modules/@elizaos/core
diff --git a/packages/scripts/cloud/admin/daemons/agent-router.ts b/packages/scripts/cloud/admin/daemons/agent-router.ts
@@ -18,8 +18,9 @@ import * as path from "node:path";
 import { fileURLToPath } from "node:url";
 import { loadLocalEnv } from "./shared/load-env";
 
-type Logger = typeof import("../../lib/utils/logger").logger;
-type Repo = typeof import("../../db/repositories/agent-sandboxes").agentSandboxesRepository;
+type Logger = typeof import("@elizaos/cloud-shared/lib/utils/logger").logger;
+type Repo =
+  typeof import("@elizaos/cloud-shared/db/repositories/agent-sandboxes").agentSandboxesRepository;
 
 interface RouterDeps {
   logger: Logger;
@@ -40,7 +41,9 @@ function parsePositiveInt(value: string | undefined, fallback: number): number {
   return Number.isFinite(parsed) && parsed > 0 ? parsed : fallback;
 }
 
-export function readRouterConfig(env: NodeJS.ProcessEnv = process.env): AgentRouterConfig {
+export function readRouterConfig(
+  env: NodeJS.ProcessEnv = process.env,
+): AgentRouterConfig {
   return {
     port: parsePositiveInt(env.AGENT_ROUTER_PORT, DEFAULT_PORT),
     bindHost: env.AGENT_ROUTER_BIND_HOST?.trim() || DEFAULT_BIND_HOST,
@@ -52,8 +55,8 @@ let depsPromise: Promise<RouterDeps> | null = null;
 async function loadDeps(): Promise<RouterDeps> {
   if (!depsPromise) {
     depsPromise = Promise.all([
-      import("../../db/repositories/agent-sandboxes"),
-      import("../../lib/utils/logger"),
+      import("@elizaos/cloud-shared/db/repositories/agent-sandboxes"),
+      import("@elizaos/cloud-shared/lib/utils/logger"),
     ]).then(([repoModule, loggerModule]) => ({
       agentSandboxesRepository: repoModule.agentSandboxesRepository,
       logger: loggerModule.logger,
@@ -69,7 +72,9 @@ interface RoutingResponse {
   target: string;
 }
 
-export async function resolveAgentRouting(agentId: string): Promise<RoutingResponse | null> {
+export async function resolveAgentRouting(
+  agentId: string,
+): Promise<RoutingResponse | null> {
   const { agentSandboxesRepository } = await loadDeps();
   const sandbox = await agentSandboxesRepository.findById(agentId);
   if (!sandbox || sandbox.status !== "running") return null;
@@ -95,15 +100,18 @@ export async function resolveAgentRouting(agentId: string): Promise<RoutingRespo
   };
 }
 
-const AGENT_ID_RE = /^[a-f0-9]{8}-[a-f0-9]{4}-[a-f0-9]{4}-[a-f0-9]{4}-[a-f0-9]{12}$/i;
+const AGENT_ID_RE =
+  /^[a-f0-9]{8}-[a-f0-9]{4}-[a-f0-9]{4}-[a-f0-9]{4}-[a-f0-9]{12}$/i;
 
 async function handleRequest(url: URL): Promise<Response> {
   if (url.pathname === "/healthz") {
     return Response.json({ ok: true }, { status: 200 });
   }
   // /headscale-ip is the path nginx Lua already calls; /routing is the alias
   // for new callers.
-  const match = url.pathname.match(/^\/agents\/([^/]+)\/(headscale-ip|routing)$/);
+  const match = url.pathname.match(
+    /^\/agents\/([^/]+)\/(headscale-ip|routing)$/,
+  );
   if (!match) {
     return Response.json({ error: "not found" }, { status: 404 });
   }
@@ -113,7 +121,10 @@ async function handleRequest(url: URL): Promise<Response> {
   }
   const routing = await resolveAgentRouting(agentId);
   if (!routing) {
-    return Response.json({ error: "agent not found or not running" }, { status: 404 });
+    return Response.json(
+      { error: "agent not found or not running" },
+      { status: 404 },
+    );
   }
   return Response.json(routing, { status: 200 });
 }
@@ -127,11 +138,16 @@ async function main(): Promise<void> {
 
   const { createServer } = await import("node:http");
   server = createServer((req, res) => {
-    const url = new URL(req.url ?? "/", `http://${req.headers.host || "localhost"}`);
+    const url = new URL(
+      req.url ?? "/",
+      `http://${req.headers.host || "localhost"}`,
+    );
     handleRequest(url)
       .then((response) => {
         res.statusCode = response.status;
-        response.headers.forEach((v, k) => res.setHeader(k, v));
+        response.headers.forEach((v, k) => {
+          res.setHeader(k, v);
+        });
         return response.text();
       })
       .then((body) => {
@@ -206,7 +222,8 @@ process.on("unhandledRejection", (reason) => {
 });
 
 function isMainModule(): boolean {
-  return path.resolve(process.argv[1]) === fileURLToPath(import.meta.url);
+  const entry = process.argv[1];
+  return entry ? path.resolve(entry) === fileURLToPath(import.meta.url) : false;
 }
 
 if (isMainModule()) {
diff --git a/packages/scripts/cloud/admin/daemons/provisioning-worker.ts b/packages/scripts/cloud/admin/daemons/provisioning-worker.ts
@@ -14,11 +14,16 @@
 
 import * as path from "node:path";
 import { fileURLToPath } from "node:url";
-import type { HeartbeatResult, ProcessingResult } from "../../lib/services/provisioning-jobs";
+import type {
+  HeartbeatResult,
+  ProcessingResult,
+} from "@elizaos/cloud-shared/lib/services/provisioning-jobs";
 import { loadLocalEnv } from "./shared/load-env";
 
-type WorkerLogger = typeof import("../../lib/utils/logger").logger;
-type WorkerService = typeof import("../../lib/services/provisioning-jobs").provisioningJobService;
+type WorkerLogger =
+  typeof import("@elizaos/cloud-shared/lib/utils/logger").logger;
+type WorkerService =
+  typeof import("@elizaos/cloud-shared/lib/services/provisioning-jobs").provisioningJobService;
 
 interface WorkerDeps {
   logger: WorkerLogger;
@@ -49,7 +54,10 @@ export function readWorkerConfig(
   argv: readonly string[] = process.argv.slice(2),
 ): ProvisioningWorkerConfig {
   return {
-    pollIntervalMs: parsePositiveInt(env.WORKER_POLL_INTERVAL, DEFAULT_POLL_INTERVAL_MS),
+    pollIntervalMs: parsePositiveInt(
+      env.WORKER_POLL_INTERVAL,
+      DEFAULT_POLL_INTERVAL_MS,
+    ),
     batchSize: parsePositiveInt(env.WORKER_BATCH_SIZE, DEFAULT_BATCH_SIZE),
     runOnce: env.WORKER_RUN_ONCE === "1" || hasFlag(argv, "--once"),
   };
@@ -60,8 +68,8 @@ let depsPromise: Promise<WorkerDeps> | null = null;
 async function loadDeps(): Promise<WorkerDeps> {
   if (!depsPromise) {
     depsPromise = Promise.all([
-      import("../../lib/services/provisioning-jobs"),
-      import("../../lib/utils/logger"),
+      import("@elizaos/cloud-shared/lib/services/provisioning-jobs"),
+      import("@elizaos/cloud-shared/lib/utils/logger"),
     ]).then(([jobsModule, loggerModule]) => ({
       provisioningJobService: jobsModule.provisioningJobService,
       logger: loggerModule.logger,
@@ -86,7 +94,9 @@ export async function processProvisioningWorkerCycle(
   return provisioningJobService.processPendingJobs(batchSize);
 }
 
-export async function processHeartbeatCycle(concurrency = 5): Promise<HeartbeatResult> {
+export async function processHeartbeatCycle(
+  concurrency = 5,
+): Promise<HeartbeatResult> {
   const { provisioningJobService } = await loadDeps();
   return provisioningJobService.processRunningHeartbeats(concurrency);
 }
@@ -97,11 +107,17 @@ async function sleep(ms: number): Promise<void> {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
 
-async function pollCycle(logger: WorkerLogger, config: ProvisioningWorkerConfig): Promise<void> {
+async function pollCycle(
+  logger: WorkerLogger,
+  config: ProvisioningWorkerConfig,
+): Promise<void> {
   try {
     const result = await processProvisioningWorkerCycle(config.batchSize);
     if (result.claimed > 0 || result.failed > 0) {
-      logger.info("[provisioning-worker] cycle complete", resultContext(result));
+      logger.info(
+        "[provisioning-worker] cycle complete",
+        resultContext(result),
+      );
     }
   } catch (error) {
     logger.error("[provisioning-worker] cycle failed", {
diff --git a/packages/scripts/cloud/admin/eliza-agent-router.service b/packages/scripts/cloud/admin/eliza-agent-router.service
@@ -8,11 +8,11 @@ Wants=network-online.target
 Type=simple
 User=deploy
 Group=deploy
-WorkingDirectory=/opt/eliza/cloud
+WorkingDirectory=/opt/eliza
 EnvironmentFile=/opt/eliza/cloud/.env.local
 # Use the installed workspace binary directly to avoid `npx` startup overhead.
 Environment=NODE_OPTIONS=--max-old-space-size=512
-ExecStart=/opt/eliza/cloud/node_modules/.bin/tsx packages/scripts/cloud/admin/daemons/agent-router.ts
+ExecStart=/opt/eliza/node_modules/.bin/tsx --tsconfig /opt/eliza/packages/cloud-shared/tsconfig.json /opt/eliza/packages/scripts/cloud/admin/daemons/agent-router.ts
 
 Restart=always
 RestartSec=5
diff --git a/packages/scripts/cloud/admin/eliza-provisioning-worker.service b/packages/scripts/cloud/admin/eliza-provisioning-worker.service
@@ -8,13 +8,13 @@ Wants=network-online.target
 Type=simple
 User=deploy
 Group=deploy
-WorkingDirectory=/opt/eliza/cloud
+WorkingDirectory=/opt/eliza
 EnvironmentFile=/opt/eliza/cloud/.env.local
 # ssh2 + dockerode are Node-only, so the daemon runs under tsx (not bun).
 # Use the installed workspace binary directly; `npx tsx` adds npm/esbuild
 # startup overhead and has hit Node worker heap limits on the deploy host.
 Environment=NODE_OPTIONS=--max-old-space-size=1024
-ExecStart=/opt/eliza/cloud/node_modules/.bin/tsx packages/scripts/cloud/admin/daemons/provisioning-worker.ts
+ExecStart=/opt/eliza/node_modules/.bin/tsx --tsconfig /opt/eliza/packages/cloud-shared/tsconfig.json /opt/eliza/packages/scripts/cloud/admin/daemons/provisioning-worker.ts
 
 Restart=always
 RestartSec=5
