fix(app): stop spawning providers per client by warming the snapshot store

The workspace screen was prefetching per-provider model lists via
`useProviderModels`, which fanned out N parallel
`list_provider_models_request`s that bypassed `ProviderSnapshotManager`
and spawned each provider fresh on every call (e.g. repeated copilot
ENOENT errors). The return value was never consumed.

Replace it with `useProvidersSnapshot(serverId, workspaceDirectory)` so
warmup goes through the shared, deduped, TTL-cached snapshot path that
consumers already read from.

packages/app/src/hooks/use-provider-models.ts

@@ -1,49 +0,0 @@
-import { useMemo } from "react";
-import { useQueries } from "@tanstack/react-query";
-import {
-  AGENT_PROVIDER_DEFINITIONS,
-  type AgentProviderDefinition,
-} from "@server/server/agent/provider-manifest";
-import type { AgentModelDefinition, AgentProvider } from "@server/server/agent/agent-sdk-types";
-import { useHostRuntimeClient, useHostRuntimeIsConnected } from "@/runtime/host-runtime";
-
-const STALE_TIME = 5 * 60 * 1000;
-
-export function useProviderModels(serverId: string) {
-  const client = useHostRuntimeClient(serverId);
-  const isConnected = useHostRuntimeIsConnected(serverId);
-  const enabled = Boolean(serverId && client && isConnected);
-
-  const queries = useQueries({
-    queries: AGENT_PROVIDER_DEFINITIONS.map((def) => ({
-      queryKey: ["providerModels", serverId, def.id] as const,
-      enabled,
-      staleTime: STALE_TIME,
-      queryFn: async () => {
-        if (!client) {
-          throw new Error("Host is not connected");
-        }
-        const payload = await client.listProviderModels(def.id as AgentProvider);
-        if (payload.error) {
-          throw new Error(payload.error);
-        }
-        return payload.models ?? [];
-      },
-    })),
-  });
-
-  const allProviderModels = useMemo(() => {
-    const map = new Map<string, AgentModelDefinition[]>();
-    for (let i = 0; i < AGENT_PROVIDER_DEFINITIONS.length; i++) {
-      const query = queries[i];
-      if (query?.data) {
-        map.set(AGENT_PROVIDER_DEFINITIONS[i]!.id, query.data);
-      }
-    }
-    return map;
-  }, [queries]);
-
-  const isLoading = queries.some((q) => q.isLoading);
-
-  return { allProviderModels, isLoading };
-}

packages/app/src/screens/workspace/workspace-screen.tsx

@@ -64,7 +64,7 @@ import {
   workspaceTabTargetsEqual,
 } from "@/utils/workspace-tab-identity";
 import { useHostRuntimeClient, useHostRuntimeIsConnected } from "@/runtime/host-runtime";
-import { useProviderModels } from "@/hooks/use-provider-models";
+import { useProvidersSnapshot } from "@/hooks/use-providers-snapshot";
 import { useWorkspaceSetupStore } from "@/stores/workspace-setup-store";
 import { useWorkspaceTerminalSessionRetention } from "@/terminal/hooks/use-workspace-terminal-session-retention";
 import {
@@ -612,9 +612,6 @@ function WorkspaceScreenContent({ serverId, workspaceId }: WorkspaceScreenProps)
 
   const normalizedServerId = trimNonEmpty(decodeSegment(serverId)) ?? "";
 
-  // Prefetch provider models early so the model picker is warm by the time it opens
-  useProviderModels(normalizedServerId);
-
   const normalizedWorkspaceId =
     resolveWorkspaceRouteId({
       routeWorkspaceId: workspaceId,
@@ -645,6 +642,10 @@ function WorkspaceScreenContent({ serverId, workspaceId }: WorkspaceScreenProps)
   );
   const workspaceDirectory = workspaceAuthority?.workspaceDirectory ?? null;
   const isMissingWorkspaceExecutionAuthority = Boolean(workspaceDescriptor && !workspaceAuthority);
+
+  // Warm the server-side provider snapshot for this workspace cwd so the model
+  // picker is ready when opened. Consumers share the same query cache key.
+  useProvidersSnapshot(normalizedServerId, workspaceDirectory);
   const [pendingTerminalCreateInput, setPendingTerminalCreateInput] = useState<{
     paneId?: string;
   } | null>(null);