[comp] Production Deploy

apps/api/package.json

@@ -4,9 +4,9 @@
     "version": "0.0.1",
     "author": "",
     "dependencies": {
-        "@ai-sdk/anthropic": "^2.0.53",
-        "@ai-sdk/groq": "^2.0.32",
-        "@ai-sdk/openai": "^2.0.65",
+        "@ai-sdk/anthropic": "^3.0.75",
+        "@ai-sdk/groq": "^3.0.38",
+        "@ai-sdk/openai": "^3.0.62",
         "@aws-sdk/client-acm": "^3.948.0",
         "@aws-sdk/client-api-gateway": "^3.948.0",
         "@aws-sdk/client-apigatewayv2": "^3.948.0",
@@ -88,7 +88,7 @@
         "@upstash/redis": "^1.34.2",
         "@upstash/vector": "^1.2.2",
         "adm-zip": "^0.5.16",
-        "ai": "^5.0.60",
+        "ai": "^6.0.175",
         "archiver": "^7.0.1",
         "axios": "^1.12.2",
         "better-auth": "^1.4.22",

apps/api/src/assistant-chat/assistant-chat.controller.ts

@@ -127,7 +127,7 @@ Important:
       const result = streamText({
         model: openai('gpt-5'),
         system: systemPrompt,
-        messages: convertToModelMessages(messages),
+        messages: await convertToModelMessages(messages),
         tools,
         stopWhen: stepCountIs(5),
       });

apps/api/src/frameworks/frameworks-scores.helper.ts

@@ -29,7 +29,7 @@ export async function getOverviewScores(organizationId: string) {
       }),
       db.onboarding.findUnique({
         where: { organizationId },
-        select: { triggerJobId: true },
+        select: { triggerJobId: true, triggerJobCompleted: true },
       }),
       db.organization.findUnique({
         where: { id: organizationId },
@@ -90,7 +90,7 @@ export async function getOverviewScores(organizationId: string) {
       incompleteTasks,
     },
     people,
-    onboardingTriggerJobId: onboarding?.triggerJobId ?? null,
+    onboardingTriggerJobId: onboarding?.triggerJobCompleted ? null : (onboarding?.triggerJobId ?? null),
     documents: await computeDocumentsScore(organizationId),
     findings: await getOrganizationFindings(organizationId),
   };

apps/api/src/policies/policies.controller.ts

@@ -1321,7 +1321,7 @@ Keep responses helpful and focused on the policy editing task.`;
     const result = streamText({
       model: openai('gpt-5.5'),
       system: systemPrompt,
-      messages: convertToModelMessages(messages),
+      messages: await convertToModelMessages(messages),
     });
 
     return result.pipeTextStreamToResponse(res);

apps/app/next.config.ts

@@ -140,7 +140,7 @@ export default withSentryConfig(withBotId(config), {
   // https://docs.sentry.io/platforms/javascript/guides/nextjs/manual-setup/
 
   // Upload a larger set of source maps for prettier stack traces (increases build time)
-  widenClientFileUpload: true,
+  widenClientFileUpload: false,
 
   // Route browser requests to Sentry through a Next.js rewrite to circumvent ad-blockers.
   // This can increase your server load as well as your hosting bill.

apps/app/package.json

@@ -3,12 +3,14 @@
     "version": "0.1.0",
     "type": "module",
     "dependencies": {
-        "@ai-sdk/anthropic": "^3.0.0",
-        "@ai-sdk/groq": "^3.0.0",
-        "@ai-sdk/openai": "^3.0.0",
-        "@ai-sdk/provider": "^3.0.0",
-        "@ai-sdk/react": "^3.0.0",
-        "@ai-sdk/rsc": "^2.0.0",
+        "@ai-sdk/anthropic": "^3.0.75",
+        "@ai-sdk/gateway": "^3.0.110",
+        "@ai-sdk/google": "^3.0.68",
+        "@ai-sdk/groq": "^3.0.38",
+        "@ai-sdk/openai": "^3.0.62",
+        "@ai-sdk/provider": "^3.0.10",
+        "@ai-sdk/react": "^3.0.177",
+        "@ai-sdk/rsc": "^2.0.175",
         "@aws-sdk/client-ec2": "^3.948.0",
         "@aws-sdk/client-lambda": "^3.948.0",
         "@aws-sdk/client-s3": "3.1013.0",
@@ -84,7 +86,7 @@
         "@vercel/sandbox": "^0.0.21",
         "@vercel/sdk": "^1.7.1",
         "@xyflow/react": "^12.10.0",
-        "ai": "^6.0.116",
+        "ai": "^6.0.175",
         "ai-elements": "^1.6.1",
         "axios": "^1.9.0",
         "better-auth": "^1.4.22",

apps/app/src/app/(app)/[orgId]/frameworks/[frameworkInstanceId]/components/FrameworkDetailContent.tsx

@@ -53,7 +53,6 @@ export function FrameworkDetailContent({
   const pathname = usePathname();
   const searchParams = useSearchParams();
   const { hasPermission, permissions } = usePermissions();
-  const versioningEnabled = useFeatureFlag('is-framework-versioning-enabled');
   const complianceTimelineEnabled = useFeatureFlag('is-timeline-enabled');
   const [deleteDialogOpen, setDeleteDialogOpen] = useState(false);
   const [dropdownOpen, setDropdownOpen] = useState(false);
@@ -82,7 +81,7 @@ export function FrameworkDetailContent({
   const validTabsList: string[] = [];
   if (complianceTimelineEnabled) validTabsList.push('progress');
   validTabsList.push('requirements');
-  if (versioningEnabled) validTabsList.push('history');
+  validTabsList.push('history');
   const validTabs = new Set(validTabsList);
   const activeTab = tabParam && validTabs.has(tabParam) ? tabParam : DEFAULT_TAB;
 
@@ -158,7 +157,7 @@ export function FrameworkDetailContent({
                 <TabsTrigger value="requirements">
                   Requirements <TabBadge>{requirementsCount}</TabBadge>
                 </TabsTrigger>
-                {versioningEnabled && <TabsTrigger value="history">History</TabsTrigger>}
+                <TabsTrigger value="history">History</TabsTrigger>
               </TabsList>
             }
           >
@@ -197,14 +196,12 @@ export function FrameworkDetailContent({
           />
         </TabsContent>
 
-        {versioningEnabled && (
-          <TabsContent value="history">
-            <SyncHistorySection
-              frameworkInstanceId={frameworkInstanceId}
-              permissions={permissions}
-            />
-          </TabsContent>
-        )}
+        <TabsContent value="history">
+          <SyncHistorySection
+            frameworkInstanceId={frameworkInstanceId}
+            permissions={permissions}
+          />
+        </TabsContent>
       </PageLayout>
 
       <FrameworkDeleteDialog

apps/app/src/app/(app)/[orgId]/frameworks/[frameworkInstanceId]/components/FrameworkVersioningSection.tsx

@@ -1,6 +1,5 @@
 'use client';
 
-import { useFeatureFlag } from '@trycompai/analytics';
 import { useParams, useRouter } from 'next/navigation';
 import { useFrameworkUpdateStatus } from '@/hooks/use-framework-update-status';
 import { usePermissions } from '@/hooks/use-permissions';
@@ -18,20 +17,13 @@ export function FrameworkVersioningSection({
   initialStatus,
   hasActiveAudit,
 }: FrameworkVersioningSectionProps) {
-  const enabled = useFeatureFlag('is-framework-versioning-enabled');
-  // Thread the flag into SWR so the update-status request doesn't fire for
-  // orgs that don't have versioning enabled. Without this the request runs
-  // every mount and we just throw the response away.
   const { data } = useFrameworkUpdateStatus(frameworkInstanceId, {
     fallbackData: initialStatus,
-    enabled,
   });
   const { hasPermission } = usePermissions();
   const router = useRouter();
   const { orgId } = useParams<{ orgId: string }>();
 
-  if (!enabled) return null;
-
   const canUpdate = hasPermission('framework', 'update');
 
   return (

apps/app/src/app/(app)/[orgId]/overview/components/ToDoOverview.tsx

@@ -5,7 +5,6 @@ import { Card, CardContent, CardHeader, CardTitle } from '@trycompai/ui/card';
 import { ScrollArea } from '@trycompai/ui/scroll-area';
 import { Tabs, TabsContent, TabsList, TabsTrigger } from '@trycompai/ui/tabs';
 import { Policy, Task } from '@db';
-import { useRealtimeRun } from '@trigger.dev/react-hooks';
 import {
   ArrowRight,
   CheckCircle2,
@@ -49,25 +48,7 @@ export function ToDoOverview({
   const [isConfirmDialogOpen, setIsConfirmDialogOpen] = useState(false);
   const [isLoading, setIsLoading] = useState(false);
 
-  const { run: onboardingRun } = useRealtimeRun(onboardingTriggerJobId || '', {
-    enabled: !!onboardingTriggerJobId,
-  });
-
-  const IN_PROGRESS_STATUSES = [
-    'QUEUED',
-    'EXECUTING',
-    'WAITING_FOR_DEPLOY',
-    'REATTEMPTING',
-    'FROZEN',
-    'DELAYED',
-    'WAITING',
-    'PENDING_VERSION',
-    'DEQUEUED',
-  ];
-
-  const isOnboardingInProgress = onboardingRun
-    ? IN_PROGRESS_STATUSES.includes(onboardingRun.status)
-    : false;
+  const isOnboardingInProgress = !!onboardingTriggerJobId;
 
   const formatStatus = (status: string) => {
     return status.replace('_', ' ').replace(/\b\w/g, (l) => l.toUpperCase());

apps/app/src/app/(app)/[orgId]/tasks/[taskId]/automation/[automationId]/tools/gateway.ts

@@ -1,6 +1,6 @@
 import { createGatewayProvider } from '@ai-sdk/gateway';
 import type { OpenAIResponsesProviderOptions } from '@ai-sdk/openai';
-import type { LanguageModelV2 } from '@ai-sdk/provider';
+import type { LanguageModelV3 } from '@ai-sdk/provider';
 import type { JSONValue } from 'ai';
 
 export async function getAvailableModels() {
@@ -10,7 +10,7 @@ export async function getAvailableModels() {
 }
 
 export interface ModelOptions {
-  model: LanguageModelV2;
+  model: LanguageModelV3;
   providerOptions?: Record<string, Record<string, JSONValue>>;
   headers?: Record<string, string>;
 }

apps/app/src/app/(app)/setup/go/[id]/components/onboarding-status.tsx

@@ -16,8 +16,13 @@ export function OnboardingStatus({ runId }: { runId: string }) {
   useEffect(() => {
     if (run?.status === 'COMPLETED') {
       router.replace('/');
+      return;
     }
-  }, [run?.status, router]);
+    const meta = run?.metadata as Record<string, unknown> | undefined;
+    if (meta?.readyForDashboard === true) {
+      router.replace('/');
+    }
+  }, [run?.status, run?.metadata, router]);
 
   return (
     <div className="flex flex-col items-center justify-center">

apps/app/src/hooks/use-framework-update-status.ts

@@ -6,17 +6,15 @@ import type { FrameworkUpdateStatus } from '@/types/framework-versioning';
 
 interface UseFrameworkUpdateStatusOptions {
   fallbackData?: FrameworkUpdateStatus;
-  enabled?: boolean;
 }
 
 export function useFrameworkUpdateStatus(
   frameworkInstanceId: string,
   options?: UseFrameworkUpdateStatusOptions,
 ) {
-  const key =
-    frameworkInstanceId && options?.enabled !== false
-      ? `/v1/frameworks/${frameworkInstanceId}/update-status`
-      : null;
+  const key = frameworkInstanceId
+    ? `/v1/frameworks/${frameworkInstanceId}/update-status`
+    : null;
 
   return useSWR<FrameworkUpdateStatus>(
     key,
@@ -30,10 +28,9 @@ export function useFrameworkUpdateStatus(
       // Always revalidate on mount, even when fallbackData is provided.
       // fallbackData is only a fast first paint — without this, SWR treats
       // the server-rendered snapshot as authoritative forever and skips the
-      // client fetch, so users don't see newly-available upgrades after the
-      // feature flag is flipped, after a sync on a sibling framework, or
-      // whenever the Next.js router cache serves a stale RSC. Short of
-      // signing out and back in.
+      // client fetch, so users don't see newly-available upgrades after a
+      // sync on a sibling framework or whenever the Next.js router cache
+      // serves a stale RSC.
       revalidateOnMount: true,
       revalidateOnFocus: true,
     },

apps/app/src/lib/embedding/run-linkage.ts

@@ -109,19 +109,10 @@ const AUTONOMOUS_FINAL_TOP_K = 8;
 const AUTONOMOUS_MIN_RERANK_SCORE = 5;
 const AUTONOMOUS_MIN_LINKS_FLOOR = 3;
 
-// How many risks/vendors to match concurrently in the bulk onboarding path.
-// Each iteration makes 1 vector query (Upstash) + 1 OpenAI rerank call + 1
-// Prisma update — typical wall-clock per iteration is 3–10 seconds (the
-// rerank LLM call dominates). With 32 in-flight at once a 20-entity
-// onboarding finishes in roughly one batch, well within gpt-5-mini /
-// Upstash rate limits.
-//
-// NOTE: this is in-process concurrency on a single trigger.dev task. The
-// natural next step (true fan-out per entity using `task.batchTrigger`)
-// would unlock trigger.dev's queue-level concurrency (50), but requires
-// passing the embedded-task metadata to children rather than rebuilding
-// taskById per child. Filed as a follow-up.
-const MATCH_CONCURRENCY = 32;
+// Risk and vendor matching run in parallel, so this limit applies to
+// EACH side independently. Keep it at 16 so both sides combined stay
+// under ~32 concurrent LLM rerank calls.
+const MATCH_CONCURRENCY = 16;
 
 async function mapWithConcurrency<T, R>(
   items: T[],
@@ -604,14 +595,22 @@ export async function runLinkage({
 
   let suggestions: RunLinkageOutput['suggestions'];
 
-  // Risk matching — fan out with bounded concurrency. Each iteration is one
-  // vector query + (rerank | DB update). Order doesn't matter here; we sum
-  // riskLinks at the end and emit `current` based on completion count.
+  // Emit initial matching phases before the parallel fan-out so the UI shows
+  // both phases starting at once.
   if (risks.length > 0) {
     onPhase?.({ name: 'matching-risks', current: 0, total: risks.length });
   }
+  if (vendors.length > 0) {
+    onPhase?.({ name: 'matching-vendors', current: 0, total: vendors.length });
+  }
+
+  // Risk + vendor matching run in parallel — they write to separate DB
+  // tables (Risk.tasks vs Vendor.tasks) and the shared taskById is read-only.
   let completedRisks = 0;
-  const riskOutcomes = await mapWithConcurrency(risks, MATCH_CONCURRENCY, async (risk) => {
+  let completedVendors = 0;
+
+  const [riskOutcomes, vendorOutcomes] = await Promise.all([
+    mapWithConcurrency(risks, MATCH_CONCURRENCY, async (risk) => {
     const similar = await findSimilarTasks({
       organizationId,
       queryText: riskQueryText(risk),
@@ -687,23 +686,8 @@ export async function runLinkage({
     completedRisks += 1;
     onPhase?.({ name: 'matching-risks', current: completedRisks, total: risks.length });
     return { count, perEntitySuggestions };
-  });
-  const riskLinks = riskOutcomes.reduce((sum, r) => sum + r.count, 0);
-  // suggestionsOnly endpoints always pass a single riskId, so at most one
-  // outcome carries suggestions — pick the first non-null.
-  for (const r of riskOutcomes) {
-    if (r.perEntitySuggestions) {
-      suggestions = r.perEntitySuggestions;
-      break;
-    }
-  }
-
-  // Vendor matching — same pattern.
-  if (vendors.length > 0) {
-    onPhase?.({ name: 'matching-vendors', current: 0, total: vendors.length });
-  }
-  let completedVendors = 0;
-  const vendorOutcomes = await mapWithConcurrency(vendors, MATCH_CONCURRENCY, async (vendor) => {
+    }),
+    mapWithConcurrency(vendors, MATCH_CONCURRENCY, async (vendor) => {
     const similar = await findSimilarTasks({
       organizationId,
       queryText: vendorQueryText(vendor),
@@ -751,8 +735,17 @@ export async function runLinkage({
     completedVendors += 1;
     onPhase?.({ name: 'matching-vendors', current: completedVendors, total: vendors.length });
     return { count, perEntitySuggestions };
-  });
+  }),
+  ]);
+
+  const riskLinks = riskOutcomes.reduce((sum, r) => sum + r.count, 0);
   const vendorLinks = vendorOutcomes.reduce((sum, v) => sum + v.count, 0);
+  for (const r of riskOutcomes) {
+    if (r.perEntitySuggestions) {
+      suggestions = r.perEntitySuggestions;
+      break;
+    }
+  }
   if (!suggestions) {
     for (const v of vendorOutcomes) {
       if (v.perEntitySuggestions) {

apps/app/src/lib/rerank-suggestions.ts

@@ -1,4 +1,4 @@
-import { openai } from '@ai-sdk/openai';
+import { createGatewayProvider } from '@ai-sdk/gateway';
 import { generateObject, jsonSchema } from 'ai';
 
 /**
@@ -39,7 +39,11 @@ export interface RerankedCandidate {
   rerankScore: number;
 }
 
-const RERANK_MODEL = 'gpt-5-mini';
+const gateway = createGatewayProvider({
+  baseURL: process.env.AI_GATEWAY_BASE_URL,
+});
+
+const RERANK_MODEL = 'google/gemini-3.1-flash-lite-preview' as const;
 
 const SYSTEM_PROMPT = `You are a GRC analyst evaluating which compliance tasks would meaningfully reduce a specific risk or vendor exposure.
 
@@ -105,7 +109,7 @@ export async function rerankSuggestions({
     .join('\n');
 
   const result = await generateObject({
-    model: openai(RERANK_MODEL),
+    model: gateway(RERANK_MODEL),
     system: SYSTEM_PROMPT,
     prompt: userPrompt,
     schema: rerankSchema,