feat: add MetaInstance dashboard UI and remove MlxIbv

Add MetaInstancePanel component showing meta-instance status cards with
create/delete support. Integrates into both welcome and chat sidebars
with status indicators, error display, and topology hover highlighting.
Remove deprecated MlxIbv runtime type from all dashboard components.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

bench/eval_tool_calls.py

@@ -20,6 +20,7 @@ from harness import (
     instance_id_from_instance,
     nodes_used_in_instance,
     resolve_model_short_id,
+    run_planning_phase,
     settle_and_fetch_placements,
     wait_for_instance_gone,
     wait_for_instance_ready,
@@ -962,6 +963,21 @@ Examples:
 
     selected.sort(key=_placement_sort_key)
     preview = selected[0]
+
+    settle_deadline = (
+        time.monotonic() + args.settle_timeout if args.settle_timeout > 0 else None
+    )
+
+    print("Planning phase: checking downloads...", file=log)
+    run_planning_phase(
+        exo,
+        full_model_id,
+        preview,
+        args.danger_delete_downloads,
+        args.timeout,
+        settle_deadline,
+    )
+
     instance = preview["instance"]
     instance_id = instance_id_from_instance(instance)
     sharding = str(preview["sharding"])

bench/exo_bench.py

@@ -35,6 +35,7 @@ from harness import (
     instance_id_from_instance,
     nodes_used_in_instance,
     resolve_model_short_id,
+    run_planning_phase,
     settle_and_fetch_placements,
     wait_for_instance_gone,
     wait_for_instance_ready,
@@ -332,6 +333,20 @@ def main() -> int:
     if args.dry_run:
         return 0
 
+    settle_deadline = (
+        time.monotonic() + args.settle_timeout if args.settle_timeout > 0 else None
+    )
+
+    logger.info("Planning phase: checking downloads...")
+    run_planning_phase(
+        client,
+        full_model_id,
+        selected[0],
+        args.danger_delete_downloads,
+        args.timeout,
+        settle_deadline,
+    )
+
     all_rows: list[dict[str, Any]] = []
 
     for preview in selected:

bench/harness.py

@@ -282,6 +282,151 @@ def settle_and_fetch_placements(
     return selected
 
 
+def run_planning_phase(
+    client: ExoClient,
+    full_model_id: str,
+    preview: dict[str, Any],
+    danger_delete: bool,
+    timeout: float,
+    settle_deadline: float | None,
+) -> None:
+    """Check disk space and ensure model is downloaded before benchmarking."""
+    # Get model size from /models
+    models = client.request_json("GET", "/models") or {}
+    model_bytes = 0
+    for m in models.get("data", []):
+        if m.get("hugging_face_id") == full_model_id:
+            model_bytes = m.get("storage_size_megabytes", 0) * 1024 * 1024
+            break
+
+    if not model_bytes:
+        logger.warning(
+            f"Could not determine size for {full_model_id}, skipping disk check"
+        )
+        return
+
+    # Get nodes from preview
+    inner = unwrap_instance(preview["instance"])
+    node_ids = list(inner["shardAssignments"]["nodeToRunner"].keys())
+    runner_to_shard = inner["shardAssignments"]["runnerToShard"]
+
+    state = client.request_json("GET", "/state")
+    downloads = state.get("downloads", {})
+    node_disk = state.get("nodeDisk", {})
+
+    for node_id in node_ids:
+        node_downloads = downloads.get(node_id, [])
+
+        # Check if model already downloaded on this node
+        already_downloaded = any(
+            "DownloadCompleted" in p
+            and unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
+                "modelId"
+            ]
+            == full_model_id
+            for p in node_downloads
+        )
+        if already_downloaded:
+            continue
+
+        # Wait for disk info if settle_deadline is set
+        disk_info = node_disk.get(node_id, {})
+        backoff = _SETTLE_INITIAL_BACKOFF_S
+        while not disk_info and settle_deadline and time.monotonic() < settle_deadline:
+            remaining = settle_deadline - time.monotonic()
+            logger.info(
+                f"Waiting for disk info on {node_id} ({remaining:.0f}s remaining)..."
+            )
+            time.sleep(min(backoff, remaining))
+            backoff = min(backoff * _SETTLE_BACKOFF_MULTIPLIER, _SETTLE_MAX_BACKOFF_S)
+            state = client.request_json("GET", "/state")
+            node_disk = state.get("nodeDisk", {})
+            disk_info = node_disk.get(node_id, {})
+
+        if not disk_info:
+            logger.warning(f"No disk info for {node_id}, skipping space check")
+            continue
+
+        avail = disk_info.get("available", {}).get("inBytes", 0)
+        if avail >= model_bytes:
+            continue
+
+        if not danger_delete:
+            raise RuntimeError(
+                f"Insufficient disk on {node_id}: need {model_bytes // (1024**3)}GB, "
+                f"have {avail // (1024**3)}GB. Use --danger-delete-downloads to free space."
+            )
+
+        # Delete from smallest to largest
+        completed = [
+            (
+                unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
+                    "modelId"
+                ],
+                p["DownloadCompleted"]["totalBytes"]["inBytes"],
+            )
+            for p in node_downloads
+            if "DownloadCompleted" in p
+        ]
+        for del_model, size in sorted(completed, key=lambda x: x[1]):
+            logger.info(f"Deleting {del_model} from {node_id} ({size // (1024**2)}MB)")
+            client.request_json("DELETE", f"/download/{node_id}/{del_model}")
+            avail += size
+            if avail >= model_bytes:
+                break
+
+        if avail < model_bytes:
+            raise RuntimeError(f"Could not free enough space on {node_id}")
+
+    # Start downloads (idempotent)
+    for node_id in node_ids:
+        runner_id = inner["shardAssignments"]["nodeToRunner"][node_id]
+        shard = runner_to_shard[runner_id]
+        client.request_json(
+            "POST",
+            "/download/start",
+            body={
+                "targetNodeId": node_id,
+                "shardMetadata": shard,
+            },
+        )
+        logger.info(f"Started download on {node_id}")
+
+    # Wait for downloads
+    start = time.time()
+    while time.time() - start < timeout:
+        state = client.request_json("GET", "/state")
+        downloads = state.get("downloads", {})
+        all_done = True
+        for node_id in node_ids:
+            done = any(
+                "DownloadCompleted" in p
+                and unwrap_instance(p["DownloadCompleted"]["shardMetadata"])[
+                    "modelCard"
+                ]["modelId"]
+                == full_model_id
+                for p in downloads.get(node_id, [])
+            )
+            failed = [
+                p["DownloadFailed"]["errorMessage"]
+                for p in downloads.get(node_id, [])
+                if "DownloadFailed" in p
+                and unwrap_instance(p["DownloadFailed"]["shardMetadata"])["modelCard"][
+                    "modelId"
+                ]
+                == full_model_id
+            ]
+            if failed:
+                raise RuntimeError(f"Download failed on {node_id}: {failed[0]}")
+            if not done:
+                all_done = False
+        if all_done:
+            return
+        time.sleep(1)
+
+    raise TimeoutError("Downloads did not complete in time")
+
+
 def add_common_instance_args(ap: argparse.ArgumentParser) -> None:
     ap.add_argument("--host", default=os.environ.get("EXO_HOST", "localhost"))
     ap.add_argument(
@@ -325,3 +470,8 @@ def add_common_instance_args(ap: argparse.ArgumentParser) -> None:
         default=0,
         help="Max seconds to wait for the cluster to produce valid placements (0 = try once).",
     )
+    ap.add_argument(
+        "--danger-delete-downloads",
+        action="store_true",
+        help="Delete existing models from smallest to largest to make room for benchmark model.",
+    )

dashboard/src/lib/components/ChatSidebar.svelte

@@ -185,11 +185,7 @@
 
     let instanceType: string | null = null;
     if (instanceTag === "MlxRingInstance") instanceType = "MLX Ring";
-    else if (
-      instanceTag === "MlxIbvInstance" ||
-      instanceTag === "MlxJacclInstance"
-    )
-      instanceType = "MLX RDMA";
+    else if (instanceTag === "MlxJacclInstance") instanceType = "MLX RDMA";
 
     let sharding: string | null = null;
     const inst = instance as {

dashboard/src/lib/components/MetaInstancePanel.svelte

@@ -0,0 +1,256 @@
+<script lang="ts">
+  import type { MetaInstance } from "$lib/stores/app.svelte";
+
+  interface Props {
+    metaInstances: Record<string, MetaInstance>;
+    instances: Record<string, unknown>;
+    onDelete?: (metaInstanceId: string) => void;
+    onHoverNodes?: (nodeIds: Set<string>) => void;
+    onHoverEnd?: () => void;
+  }
+
+  let { metaInstances, instances, onDelete, onHoverNodes, onHoverEnd }: Props =
+    $props();
+
+  function getTagged(obj: unknown): [string | null, unknown] {
+    if (!obj || typeof obj !== "object") return [null, null];
+    const keys = Object.keys(obj as Record<string, unknown>);
+    if (keys.length === 1) {
+      return [keys[0], (obj as Record<string, unknown>)[keys[0]]];
+    }
+    return [null, null];
+  }
+
+  interface LinkedInstance {
+    instanceId: string;
+    modelId: string;
+    nodeIds: string[];
+  }
+
+  function findLinkedInstance(metaInstanceId: string): LinkedInstance | null {
+    for (const [instanceId, instanceWrapped] of Object.entries(instances)) {
+      const [, instance] = getTagged(instanceWrapped);
+      if (!instance || typeof instance !== "object") continue;
+      const inst = instance as {
+        metaInstanceId?: string;
+        shardAssignments?: {
+          modelId?: string;
+          nodeToRunner?: Record<string, string>;
+        };
+      };
+      if (inst.metaInstanceId === metaInstanceId) {
+        return {
+          instanceId,
+          modelId: inst.shardAssignments?.modelId || "Unknown",
+          nodeIds: Object.keys(inst.shardAssignments?.nodeToRunner ?? {}),
+        };
+      }
+    }
+    return null;
+  }
+
+  type MetaStatus = "active" | "provisioning" | "error" | "retrying";
+
+  function getStatus(
+    meta: MetaInstance,
+    linked: LinkedInstance | null,
+  ): MetaStatus {
+    if (meta.placementError || meta.lastFailureError) {
+      if (meta.consecutiveFailures > 0 && meta.consecutiveFailures < 3)
+        return "retrying";
+      return "error";
+    }
+    if (linked) return "active";
+    return "provisioning";
+  }
+
+  function statusLabel(status: MetaStatus): string {
+    switch (status) {
+      case "active":
+        return "ACTIVE";
+      case "provisioning":
+        return "PROVISIONING";
+      case "error":
+        return "ERROR";
+      case "retrying":
+        return "RETRYING";
+    }
+  }
+
+  function statusDotClass(status: MetaStatus): string {
+    switch (status) {
+      case "active":
+        return "bg-green-400 shadow-[0_0_6px_rgba(74,222,128,0.6)]";
+      case "provisioning":
+        return "bg-yellow-400 animate-pulse shadow-[0_0_6px_rgba(250,204,21,0.6)]";
+      case "error":
+        return "bg-red-400 shadow-[0_0_6px_rgba(248,113,113,0.6)]";
+      case "retrying":
+        return "bg-orange-400 animate-pulse shadow-[0_0_6px_rgba(251,146,60,0.6)]";
+    }
+  }
+
+  function statusTextClass(status: MetaStatus): string {
+    switch (status) {
+      case "active":
+        return "text-green-400";
+      case "provisioning":
+        return "text-yellow-400";
+      case "error":
+        return "text-red-400";
+      case "retrying":
+        return "text-orange-400";
+    }
+  }
+
+  function borderClass(status: MetaStatus): string {
+    switch (status) {
+      case "active":
+        return "border-green-500/30 border-l-green-400";
+      case "provisioning":
+        return "border-purple-500/30 border-l-purple-400";
+      case "error":
+        return "border-red-500/30 border-l-red-400";
+      case "retrying":
+        return "border-orange-500/30 border-l-orange-400";
+    }
+  }
+
+  function cornerClass(status: MetaStatus): string {
+    switch (status) {
+      case "active":
+        return "border-green-500/50";
+      case "provisioning":
+        return "border-purple-500/50";
+      case "error":
+        return "border-red-500/50";
+      case "retrying":
+        return "border-orange-500/50";
+    }
+  }
+
+  function handleHover(meta: MetaInstance, linked: LinkedInstance | null) {
+    if (!onHoverNodes) return;
+    if (linked && linked.nodeIds.length > 0) {
+      onHoverNodes(new Set(linked.nodeIds));
+    } else if (meta.nodeIds && meta.nodeIds.length > 0) {
+      onHoverNodes(new Set(meta.nodeIds));
+    }
+  }
+
+  function formatModelId(modelId: string): string {
+    return modelId.split("/").pop() || modelId;
+  }
+</script>
+
+<!-- Panel Header -->
+<div class="flex items-center gap-2 mb-4">
+  <div
+    class="w-2 h-2 bg-purple-400 rounded-full shadow-[0_0_8px_rgba(168,85,247,0.6)] animate-pulse"
+  ></div>
+  <h3 class="text-xs text-purple-400 font-mono tracking-[0.2em] uppercase">
+    Meta-Instances
+  </h3>
+  <div
+    class="flex-1 h-px bg-gradient-to-r from-purple-400/30 to-transparent"
+  ></div>
+</div>
+
+<div
+  class="space-y-3 max-h-72 xl:max-h-96 overflow-y-auto overflow-x-hidden py-px"
+>
+  {#each Object.entries(metaInstances) as [id, meta]}
+    {@const linked = findLinkedInstance(meta.metaInstanceId)}
+    {@const status = getStatus(meta, linked)}
+    {@const corners = cornerClass(status)}
+    <div
+      class="relative group cursor-default"
+      role="group"
+      onmouseenter={() => handleHover(meta, linked)}
+      onmouseleave={() => onHoverEnd?.()}
+    >
+      <!-- Corner accents -->
+      <div
+        class="absolute -top-px -left-px w-2 h-2 border-l border-t {corners}"
+      ></div>
+      <div
+        class="absolute -top-px -right-px w-2 h-2 border-r border-t {corners}"
+      ></div>
+      <div
+        class="absolute -bottom-px -left-px w-2 h-2 border-l border-b {corners}"
+      ></div>
+      <div
+        class="absolute -bottom-px -right-px w-2 h-2 border-r border-b {corners}"
+      ></div>
+
+      <div
+        class="bg-exo-dark-gray/60 border border-l-2 {borderClass(status)} p-3"
+      >
+        <div class="flex justify-between items-start mb-2 pl-2">
+          <div class="flex items-center gap-2">
+            <div
+              class="w-1.5 h-1.5 {statusDotClass(status)} rounded-full"
+            ></div>
+            <span class="text-exo-light-gray font-mono text-sm tracking-wider">
+              {meta.metaInstanceId.slice(0, 8).toUpperCase()}
+            </span>
+            <span
+              class="{statusTextClass(
+                status,
+              )} text-[10px] font-mono tracking-wider"
+            >
+              {statusLabel(status)}
+            </span>
+          </div>
+          <button
+            onclick={() => onDelete?.(meta.metaInstanceId)}
+            class="text-xs px-2 py-1 font-mono tracking-wider uppercase border border-red-500/30 text-red-400 hover:bg-red-500/20 hover:text-red-400 hover:border-red-500/50 transition-all duration-200 cursor-pointer"
+          >
+            DELETE
+          </button>
+        </div>
+        <div class="pl-2">
+          <div class="text-exo-yellow text-xs font-mono tracking-wide truncate">
+            {formatModelId(meta.modelId)}
+          </div>
+          <div class="text-white/60 text-xs font-mono">
+            {meta.sharding} &middot; {meta.instanceMeta} &middot; min {meta.minNodes}
+            node{meta.minNodes !== 1 ? "s" : ""}
+          </div>
+          {#if meta.nodeIds && meta.nodeIds.length > 0}
+            <div class="text-white/50 text-[10px] font-mono mt-0.5">
+              Pinned: {meta.nodeIds.map((n) => n.slice(0, 8)).join(", ")}
+            </div>
+          {/if}
+          {#if meta.placementError}
+            <div
+              class="text-red-400/80 text-[10px] font-mono mt-1 truncate"
+              title={meta.placementError}
+            >
+              {meta.placementError}
+            </div>
+          {/if}
+          {#if meta.lastFailureError}
+            <div
+              class="text-orange-400/80 text-[10px] font-mono mt-0.5 truncate"
+              title={meta.lastFailureError}
+            >
+              Failure: {meta.lastFailureError}
+            </div>
+          {/if}
+          {#if meta.consecutiveFailures > 0}
+            <div class="text-orange-400/60 text-[10px] font-mono mt-0.5">
+              Retries: {meta.consecutiveFailures}/3
+            </div>
+          {/if}
+          {#if linked}
+            <div class="text-purple-400/60 text-[10px] font-mono mt-1">
+              Instance: {linked.instanceId.slice(0, 8)} &middot; {linked.nodeIds
+                .length} node{linked.nodeIds.length !== 1 ? "s" : ""}
+            </div>
+          {/if}
+        </div>
+      </div>
+    </div>
+  {/each}
+</div>

dashboard/src/lib/components/ModelCard.svelte

@@ -21,7 +21,7 @@
     } | null;
     nodes?: Record<string, NodeInfo>;
     sharding?: "Pipeline" | "Tensor";
-    runtime?: "MlxRing" | "MlxIbv" | "MlxJaccl";
+    runtime?: "MlxRing" | "MlxJaccl";
     onLaunch?: () => void;
     tags?: string[];
     apiPreview?: PlacementPreview | null;
@@ -348,7 +348,7 @@
   // Debug mode state
   const isDebugMode = $derived(debugMode());
   const topology = $derived(topologyData());
-  const isRdma = $derived(runtime === "MlxIbv" || runtime === "MlxJaccl");
+  const isRdma = $derived(runtime === "MlxJaccl");
 
   // Get interface name for an IP from node data
   function getInterfaceForIp(nodeId: string, ip?: string): string | null {
@@ -575,7 +575,7 @@
       >
         {runtime === "MlxRing"
           ? "MLX Ring"
-          : runtime === "MlxIbv" || runtime === "MlxJaccl"
+          : runtime === "MlxJaccl"
             ? "MLX RDMA"
             : runtime}
       </span>

dashboard/src/lib/components/index.ts

@@ -11,4 +11,5 @@ export { default as FamilySidebar } from "./FamilySidebar.svelte";
 export { default as HuggingFaceResultItem } from "./HuggingFaceResultItem.svelte";
 export { default as ModelFilterPopover } from "./ModelFilterPopover.svelte";
 export { default as ModelPickerGroup } from "./ModelPickerGroup.svelte";
+export { default as MetaInstancePanel } from "./MetaInstancePanel.svelte";
 export { default as ModelPickerModal } from "./ModelPickerModal.svelte";

dashboard/src/lib/stores/app.svelte.ts

@@ -74,6 +74,18 @@ export interface Instance {
   };
 }
 
+export interface MetaInstance {
+  metaInstanceId: string;
+  modelId: string;
+  sharding: "Pipeline" | "Tensor";
+  instanceMeta: "MlxRing" | "MlxJaccl";
+  minNodes: number;
+  nodeIds: string[] | null;
+  placementError: string | null;
+  consecutiveFailures: number;
+  lastFailureError: string | null;
+}
+
 // Granular node state types from the new state structure
 interface RawNodeIdentity {
   modelId?: string;
@@ -168,7 +180,7 @@ export interface ModelDownloadStatus {
 export interface PlacementPreview {
   model_id: string;
   sharding: "Pipeline" | "Tensor";
-  instance_meta: "MlxRing" | "MlxIbv" | "MlxJaccl";
+  instance_meta: "MlxRing" | "MlxJaccl";
   instance: unknown | null;
   memory_delta_by_node: Record<string, number> | null;
   error: string | null;
@@ -219,10 +231,10 @@ interface RawStateResponse {
     string,
     {
       MlxRingInstance?: Instance;
-      MlxIbvInstance?: Instance;
       MlxJacclInstance?: Instance;
     }
   >;
+  metaInstances?: Record<string, MetaInstance>;
   runners?: Record<string, unknown>;
   downloads?: Record<string, unknown[]>;
   // New granular node state fields
@@ -533,6 +545,7 @@ class AppStore {
   // Topology state
   topologyData = $state<TopologyData | null>(null);
   instances = $state<Record<string, unknown>>({});
+  metaInstances = $state<Record<string, MetaInstance>>({});
   runners = $state<Record<string, unknown>>({});
   downloads = $state<Record<string, unknown[]>>({});
   nodeDisk = $state<
@@ -905,11 +918,7 @@ class AppStore {
 
     let instanceType: string | null = null;
     if (instanceTag === "MlxRingInstance") instanceType = "MLX Ring";
-    else if (
-      instanceTag === "MlxIbvInstance" ||
-      instanceTag === "MlxJacclInstance"
-    )
-      instanceType = "MLX RDMA";
+    else if (instanceTag === "MlxJacclInstance") instanceType = "MLX RDMA";
 
     let sharding: string | null = null;
     const inst = instance as {
@@ -1271,6 +1280,9 @@ class AppStore {
       if (data.runners) {
         this.runners = data.runners;
       }
+      if (data.metaInstances) {
+        this.metaInstances = data.metaInstances;
+      }
       if (data.downloads) {
         this.downloads = data.downloads;
       }
@@ -3112,6 +3124,7 @@ export const totalTokens = () => appStore.totalTokens;
 export const prefillProgress = () => appStore.prefillProgress;
 export const topologyData = () => appStore.topologyData;
 export const instances = () => appStore.instances;
+export const metaInstances = () => appStore.metaInstances;
 export const runners = () => appStore.runners;
 export const downloads = () => appStore.downloads;
 export const nodeDisk = () => appStore.nodeDisk;

dashboard/src/routes/+page.svelte

@@ -5,6 +5,7 @@
     ChatMessages,
     ChatSidebar,
     ModelCard,
+    MetaInstancePanel,
     ModelPickerModal,
   } from "$lib/components";
   import {
@@ -24,6 +25,7 @@
     lastUpdate,
     clearChat,
     instances,
+    metaInstances,
     runners,
     downloads,
     placementPreviews,
@@ -60,6 +62,7 @@
   const data = $derived(topologyData());
   const update = $derived(lastUpdate());
   const instanceData = $derived(instances());
+  const metaInstancesData = $derived(metaInstances());
   const runnersData = $derived(runners());
   const downloadsData = $derived(downloads());
   const previewsData = $derived(placementPreviews());
@@ -224,7 +227,7 @@
     return model.tasks.includes("ImageToImage");
   }
   let selectedSharding = $state<"Pipeline" | "Tensor">("Pipeline");
-  type InstanceMeta = "MlxRing" | "MlxIbv" | "MlxJaccl";
+  type InstanceMeta = "MlxRing" | "MlxJaccl";
 
   // Launch defaults persistence
   const LAUNCH_DEFAULTS_KEY = "exo-launch-defaults";
@@ -481,7 +484,7 @@
   const matchesSelectedRuntime = (runtime: InstanceMeta): boolean =>
     selectedInstanceType === "MlxRing"
       ? runtime === "MlxRing"
-      : runtime === "MlxIbv" || runtime === "MlxJaccl";
+      : runtime === "MlxJaccl";
 
   // Helper to check if a model can be launched (has valid placement with >= minNodes)
   function canModelFit(modelId: string): boolean {
@@ -1248,6 +1251,44 @@
     }
   }
 
+  async function deleteMetaInstance(metaInstanceId: string) {
+    if (!confirm(`Delete meta-instance ${metaInstanceId.slice(0, 8)}...?`))
+      return;
+    try {
+      const response = await fetch(`/meta_instance/${metaInstanceId}`, {
+        method: "DELETE",
+        headers: { "Content-Type": "application/json" },
+      });
+      if (!response.ok) {
+        console.error("Failed to delete meta-instance:", response.status);
+      }
+    } catch (error) {
+      console.error("Error deleting meta-instance:", error);
+    }
+  }
+
+  async function createMetaInstance() {
+    if (!selectedModelId) return;
+    try {
+      const response = await fetch("/meta_instance", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          model_id: selectedModelId,
+          sharding: selectedSharding,
+          instance_meta: selectedInstanceType,
+          min_nodes: selectedMinNodes,
+          node_ids: nodeFilter.size > 0 ? Array.from(nodeFilter) : null,
+        }),
+      });
+      if (!response.ok) {
+        console.error("Failed to create meta-instance:", await response.text());
+      }
+    } catch (error) {
+      console.error("Error creating meta-instance:", error);
+    }
+  }
+
   // Helper to unwrap tagged unions like { MlxRingInstance: {...} }
   function getTagged(obj: unknown): [string | null, unknown] {
     if (!obj || typeof obj !== "object") return [null, null];
@@ -1288,11 +1329,7 @@
     // Instance type from tag
     let instanceType = "Unknown";
     if (instanceTag === "MlxRingInstance") instanceType = "MLX Ring";
-    else if (
-      instanceTag === "MlxIbvInstance" ||
-      instanceTag === "MlxJacclInstance"
-    )
-      instanceType = "MLX RDMA";
+    else if (instanceTag === "MlxJacclInstance") instanceType = "MLX RDMA";
 
     const inst = instance as {
       shardAssignments?: {
@@ -1641,6 +1678,7 @@
 
   const nodeCount = $derived(data ? Object.keys(data.nodes).length : 0);
   const instanceCount = $derived(Object.keys(instanceData).length);
+  const metaInstanceCount = $derived(Object.keys(metaInstancesData).length);
 
   // Helper to get the number of nodes in a placement preview
   function getPreviewNodeCount(preview: PlacementPreview): number {
@@ -2286,6 +2324,19 @@
         <aside
           class="w-80 border-l border-exo-yellow/10 bg-exo-dark-gray flex flex-col flex-shrink-0"
         >
+          <!-- Meta-Instances Panel -->
+          {#if metaInstanceCount > 0}
+            <div class="p-4 flex-shrink-0">
+              <MetaInstancePanel
+                metaInstances={metaInstancesData}
+                instances={instanceData}
+                onDelete={deleteMetaInstance}
+                onHoverNodes={(nodes) => (hoveredPreviewNodes = nodes)}
+                onHoverEnd={() => (hoveredPreviewNodes = new Set())}
+              />
+            </div>
+          {/if}
+
           <!-- Running Instances Panel (only shown when instances exist) - Scrollable -->
           {#if instanceCount > 0}
             <div class="p-4 flex-shrink-0">
@@ -2877,21 +2928,21 @@
                   </button>
                   <button
                     onclick={() => {
-                      selectedInstanceType = "MlxIbv";
+                      selectedInstanceType = "MlxJaccl";
                       saveLaunchDefaults();
                     }}
                     class="flex items-center gap-2 py-2 px-4 text-sm font-mono border rounded transition-all duration-200 cursor-pointer {selectedInstanceType ===
-                    'MlxIbv'
+                    'MlxJaccl'
                       ? 'bg-transparent text-exo-yellow border-exo-yellow'
                       : 'bg-transparent text-white/70 border-exo-medium-gray/50 hover:border-exo-yellow/50'}"
                   >
                     <span
                       class="w-4 h-4 rounded-full border-2 flex items-center justify-center {selectedInstanceType ===
-                      'MlxIbv'
+                      'MlxJaccl'
                         ? 'border-exo-yellow'
                         : 'border-exo-medium-gray'}"
                     >
-                      {#if selectedInstanceType === "MlxIbv"}
+                      {#if selectedInstanceType === "MlxJaccl"}
                         <span class="w-2 h-2 rounded-full bg-exo-yellow"></span>
                       {/if}
                     </span>
@@ -3018,6 +3069,15 @@
                         />
                       </div>
                     {/each}
+                    {#if selectedModelId}
+                      <button
+                        type="button"
+                        onclick={createMetaInstance}
+                        class="w-full mt-3 py-2 px-4 text-xs font-mono tracking-wider uppercase border border-purple-500/30 text-purple-400 hover:bg-purple-500/20 hover:border-purple-500/50 transition-all duration-200 cursor-pointer"
+                      >
+                        CREATE PERSISTENT
+                      </button>
+                    {/if}
                   </div>
                 {:else if selectedModel}
                   <div class="text-center py-4">
@@ -3101,6 +3161,19 @@
               </div>
             </button>
 
+            <!-- Meta-Instances Section -->
+            {#if metaInstanceCount > 0}
+              <div class="p-4 flex-shrink-0">
+                <MetaInstancePanel
+                  metaInstances={metaInstancesData}
+                  instances={instanceData}
+                  onDelete={deleteMetaInstance}
+                  onHoverNodes={(nodes) => (hoveredPreviewNodes = nodes)}
+                  onHoverEnd={() => (hoveredPreviewNodes = new Set())}
+                />
+              </div>
+            {/if}
+
             <!-- Instances Section (only shown when instances exist) -->
             {#if instanceCount > 0}
               <div class="p-4 flex-1">