Limit run-time model list waiting (#812)

package.json

@@ -1,6 +1,6 @@
 {
   "name": "hermes-web-ui",
-  "version": "0.5.27",
+  "version": "0.5.28",
   "description": "Self-hosted AI chat dashboard for Hermes Agent — multi-model web UI with multi-platform integration",
   "repository": {
     "type": "git",

packages/client/src/data/changelog.ts

@@ -6,7 +6,7 @@ export interface ChangelogEntry {
 
 export const changelog: ChangelogEntry[] = [
   {
-    version: '0.5.27',
+    version: '0.5.28',
     date: '2026-05-17',
     changes: [
       'changelog.new_0_5_27_1',

packages/client/src/stores/hermes/app.ts

@@ -44,7 +44,7 @@ export const useAppStore = defineStore('app', () => {
   const sessionPersistence = ref(true)
   const maxTokens = ref(4096)
   let modelsLoadPromise: Promise<void> | null = null
-  let modelsLoadedAt = 0
+  let modelsLastRequestedAt = 0
 
   async function doUpdate(): Promise<boolean> {
     updating.value = true
@@ -134,12 +134,12 @@ export const useAppStore = defineStore('app', () => {
   async function loadModels(force = false) {
     if (!hasApiKey()) return
     if (!force && modelsLoadPromise) return modelsLoadPromise
-    if (!force && modelGroups.value.length > 0 && Date.now() - modelsLoadedAt < MODELS_CACHE_TTL_MS) return
+    if (!force && modelsLastRequestedAt > 0 && Date.now() - modelsLastRequestedAt < MODELS_CACHE_TTL_MS) return
+    modelsLastRequestedAt = Date.now()
     modelsLoadPromise = (async () => {
       try {
         const res = await fetchAvailableModels()
         applyAvailableModelsResponse(res)
-        modelsLoadedAt = Date.now()
       } catch {
         // ignore
       } finally {
@@ -149,6 +149,16 @@ export const useAppStore = defineStore('app', () => {
     return modelsLoadPromise
   }
 
+  async function waitForModelsForRun(timeoutMs = 15000) {
+    if (!hasApiKey()) return
+    const pending = modelsLoadPromise || (modelsLastRequestedAt === 0 ? loadModels() : null)
+    if (!pending) return
+    await Promise.race([
+      pending,
+      new Promise<void>(resolve => setTimeout(resolve, timeoutMs)),
+    ])
+  }
+
   async function reloadModels() {
     return loadModels(true)
   }
@@ -300,6 +310,7 @@ export const useAppStore = defineStore('app', () => {
     maxTokens,
     checkConnection,
     loadModels,
+    waitForModelsForRun,
     reloadModels,
     applyAvailableModelsResponse,
     switchModel,

packages/client/src/stores/hermes/chat.ts

@@ -910,7 +910,7 @@ export const useChatStore = defineStore('chat', () => {
       }
 
       const appStore = useAppStore()
-      await appStore.loadModels()
+      await appStore.waitForModelsForRun()
       const sessionModel = activeSession.value?.model || appStore.selectedModel
       const isBridgeSource = activeSession.value?.source === 'cli'
       const sessionProvider = activeSession.value?.provider || appStore.selectedProvider

tests/client/app-store.test.ts

@@ -1,5 +1,5 @@
 // @vitest-environment jsdom
-import { beforeEach, describe, expect, it, vi } from 'vitest'
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
 import { createPinia, setActivePinia } from 'pinia'
 
 const mockSystemApi = vi.hoisted(() => ({
@@ -23,6 +23,10 @@ describe('App Store', () => {
     window.localStorage.clear()
   })
 
+  afterEach(() => {
+    vi.useRealTimers()
+  })
+
   it('persists desktop sidebar collapsed state to localStorage', () => {
     const store = useAppStore()
 
@@ -224,6 +228,40 @@ describe('App Store', () => {
     expect(store.displayModelName('unknown', 'deepseek')).toBe('unknown')
   })
 
+  it('does not refetch available models within the cache window after an empty response', async () => {
+    mockSystemApi.fetchAvailableModels.mockResolvedValue({
+      default: '',
+      default_provider: '',
+      groups: [],
+      allProviders: [],
+    })
+    const store = useAppStore()
+
+    await store.loadModels()
+    await store.loadModels()
+
+    expect(mockSystemApi.fetchAvailableModels).toHaveBeenCalledTimes(1)
+  })
+
+  it('waits only up to the run timeout for the first available models request', async () => {
+    vi.useFakeTimers()
+    mockSystemApi.fetchAvailableModels.mockReturnValue(new Promise(() => {}))
+    const store = useAppStore()
+    let resolved = false
+
+    const waitPromise = store.waitForModelsForRun(15000).then(() => {
+      resolved = true
+    })
+
+    expect(mockSystemApi.fetchAvailableModels).toHaveBeenCalledTimes(1)
+    await vi.advanceTimersByTimeAsync(14999)
+    expect(resolved).toBe(false)
+    await vi.advanceTimersByTimeAsync(1)
+    await waitPromise
+    expect(resolved).toBe(true)
+    expect(store.modelGroups).toEqual([])
+  })
+
   it('keeps aliases scoped to their provider when model IDs overlap', async () => {
     mockSystemApi.fetchAvailableModels.mockResolvedValue({
       default: 'shared-model',