Merge 8af0caaa40 into 44ecf41ca6

admin/app/services/ollama_service.ts

@@ -199,7 +199,7 @@ export class OllamaService {
       query: null,
       limit: 15,
     }
-  ): Promise<{ models: NomadOllamaModel[], hasMore: boolean } | null> {
+  ): Promise<{ models: NomadOllamaModel[], hasMore: boolean, source: 'api' | 'fallback' } | null> {
     try {
       const models = await this.retrieveAndRefreshModels(sort, force)
       if (!models) {
@@ -209,7 +209,8 @@ export class OllamaService {
         )
         return {
           models: FALLBACK_RECOMMENDED_OLLAMA_MODELS,
-          hasMore: false
+          hasMore: false,
+          source: 'fallback',
         }
       }
 
@@ -217,7 +218,8 @@ export class OllamaService {
         const filteredModels = query ? this.fuseSearchModels(models, query) : models
         return {
           models: filteredModels.slice(0, limit || 15),
-          hasMore: filteredModels.length > (limit || 15)
+          hasMore: filteredModels.length > (limit || 15),
+          source: 'api',
         }
       }
 
@@ -237,13 +239,15 @@ export class OllamaService {
         const filteredRecommendedModels = this.fuseSearchModels(recommendedModels, query)
         return {
           models: filteredRecommendedModels,
-          hasMore: filteredRecommendedModels.length > (limit || 15)
+          hasMore: filteredRecommendedModels.length > (limit || 15),
+          source: 'api',
         }
       }
 
       return {
         models: recommendedModels,
-        hasMore: recommendedModels.length > (limit || 15)
+        hasMore: recommendedModels.length > (limit || 15),
+        source: 'api',
       }
     } catch (error) {
       logger.error(

admin/app/services/system_service.ts

@@ -14,6 +14,7 @@ import env from '#start/env'
 import KVStore from '#models/kv_store'
 import { KV_STORE_SCHEMA, KVStoreKey } from '../../types/kv_store.js'
 import { isNewerVersion } from '../utils/version.js'
+import { NOMAD_API_DEFAULT_BASE_URL } from '../../constants/misc.js'
 
 
 @inject()
@@ -384,8 +385,9 @@ export class SystemService {
 
   async subscribeToReleaseNotes(email: string): Promise<{ success: boolean; message: string }> {
     try {
+      const baseUrl = env.get('NOMAD_API_URL') || NOMAD_API_DEFAULT_BASE_URL
       const response = await axios.post(
-        'https://api.projectnomad.us/api/v1/lists/release-notes/subscribe',
+        `${baseUrl}/api/v1/lists/release-notes/subscribe`,
         { email },
         { timeout: 5000 }
       )
@@ -399,13 +401,13 @@ export class SystemService {
 
       return {
         success: false,
-        message: `Failed to subscribe: ${response.statusText}`,
+        message: 'Failed to subscribe to release notes. Please try again later.',
       }
     } catch (error) {
-      logger.error('Error subscribing to release notes:', error)
+      logger.error({ err: error }, '[SystemService] Error subscribing to release notes')
       return {
         success: false,
-        message: `Failed to subscribe: ${error instanceof Error ? error.message : error}`,
+        message: 'Failed to subscribe to release notes. The service may be temporarily unavailable.',
       }
     }
   }

admin/constants/ollama.ts

@@ -19,14 +19,14 @@ export const FALLBACK_RECOMMENDED_OLLAMA_MODELS: NomadOllamaModel[] = [
         context: '128k',
         input: 'Text',
         cloud: false,
-        thinking: false
+        thinking: false,
       },
     ],
   },
   {
     name: 'deepseek-r1',
     description:
-      'DeepSeek-R1 is a family of open reasoning models with performance approaching that of leading models, such as O3 and Gemini 2.5 Pro.',
+      'DeepSeek-R1 is a family of open reasoning models with performance approaching that of leading models.',
     estimated_pulls: '77.2M',
     id: '0b566560-68a6-4964-b0d4-beb3ab1ad694',
     first_seen: '2026-01-28T23:37:31.000+00:00',
@@ -38,7 +38,15 @@ export const FALLBACK_RECOMMENDED_OLLAMA_MODELS: NomadOllamaModel[] = [
         context: '128k',
         input: 'Text',
         cloud: false,
-        thinking: true
+        thinking: true,
+      },
+      {
+        name: 'deepseek-r1:8b',
+        size: '4.9 GB',
+        context: '128k',
+        input: 'Text',
+        cloud: false,
+        thinking: true,
       },
     ],
   },
@@ -56,7 +64,267 @@ export const FALLBACK_RECOMMENDED_OLLAMA_MODELS: NomadOllamaModel[] = [
         context: '128k',
         input: 'Text',
         cloud: false,
-        thinking: false
+        thinking: false,
+      },
+      {
+        name: 'llama3.2:3b',
+        size: '2.0 GB',
+        context: '128k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'mistral',
+    description:
+      'Mistral 7B is a compact yet powerful language model by Mistral AI, great for general-purpose tasks.',
+    estimated_pulls: '26.8M',
+    id: 'a2b3c4d5-e6f7-4890-abcd-ef1234567890',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '6 months ago',
+    tags: [
+      {
+        name: 'mistral:7b',
+        size: '4.1 GB',
+        context: '32k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'gemma2',
+    description:
+      'Gemma 2 is a family of lightweight, state-of-the-art open models by Google.',
+    estimated_pulls: '18.5M',
+    id: 'b3c4d5e6-f7a8-4901-bcde-f12345678901',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '8 months ago',
+    tags: [
+      {
+        name: 'gemma2:2b',
+        size: '1.6 GB',
+        context: '8k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+      {
+        name: 'gemma2:9b',
+        size: '5.4 GB',
+        context: '8k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'phi3',
+    description:
+      'Phi-3 is a family of small language models developed by Microsoft with strong reasoning capabilities.',
+    estimated_pulls: '12.4M',
+    id: 'c4d5e6f7-a8b9-4012-cdef-123456789012',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '10 months ago',
+    tags: [
+      {
+        name: 'phi3:mini',
+        size: '2.3 GB',
+        context: '128k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'codellama',
+    description:
+      'Code Llama is a model for generating and discussing code, built on top of Llama 2.',
+    estimated_pulls: '15.2M',
+    id: 'd5e6f7a8-b9c0-4123-defa-234567890123',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '1 year ago',
+    tags: [
+      {
+        name: 'codellama:7b',
+        size: '3.8 GB',
+        context: '16k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'qwen2.5-coder',
+    description:
+      'Qwen2.5-Coder is a series of code-specific large language models by Alibaba Cloud.',
+    estimated_pulls: '9.7M',
+    id: 'e6f7a8b9-c0d1-4234-efab-345678901234',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '4 months ago',
+    tags: [
+      {
+        name: 'qwen2.5-coder:3b',
+        size: '1.9 GB',
+        context: '32k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+      {
+        name: 'qwen2.5-coder:7b',
+        size: '4.7 GB',
+        context: '32k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'tinyllama',
+    description:
+      'TinyLlama is a compact 1.1B language model ideal for resource-constrained environments.',
+    estimated_pulls: '5.3M',
+    id: 'f7a8b9c0-d1e2-4345-fabc-456789012345',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '1 year ago',
+    tags: [
+      {
+        name: 'tinyllama:1.1b',
+        size: '638 MB',
+        context: '2k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'nomic-embed-text',
+    description:
+      'A high-performing open embedding model with a large token context window.',
+    estimated_pulls: '22.1M',
+    id: 'a8b9c0d1-e2f3-4456-abcd-567890123456',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '9 months ago',
+    tags: [
+      {
+        name: 'nomic-embed-text:latest',
+        size: '274 MB',
+        context: '8k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'qwen2.5',
+    description:
+      'Qwen2.5 is a series of large language models by Alibaba Cloud with strong multilingual support.',
+    estimated_pulls: '14.6M',
+    id: 'b9c0d1e2-f3a4-4567-bcde-678901234567',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '5 months ago',
+    tags: [
+      {
+        name: 'qwen2.5:3b',
+        size: '1.9 GB',
+        context: '32k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+      {
+        name: 'qwen2.5:7b',
+        size: '4.7 GB',
+        context: '32k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'llava',
+    description:
+      'LLaVA is a multimodal model that combines a vision encoder with Vicuna for visual and language understanding.',
+    estimated_pulls: '8.9M',
+    id: 'c0d1e2f3-a4b5-4678-cdef-789012345678',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '1 year ago',
+    tags: [
+      {
+        name: 'llava:7b',
+        size: '4.7 GB',
+        context: '4k',
+        input: 'Text, Images',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'starcoder2',
+    description:
+      'StarCoder2 is a family of code generation models trained on The Stack v2 dataset.',
+    estimated_pulls: '3.8M',
+    id: 'd1e2f3a4-b5c6-4789-defa-890123456789',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '1 year ago',
+    tags: [
+      {
+        name: 'starcoder2:3b',
+        size: '1.7 GB',
+        context: '16k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'dolphin-mistral',
+    description:
+      'Dolphin Mistral is an uncensored, fine-tuned Mistral model focused on helpfulness.',
+    estimated_pulls: '4.2M',
+    id: 'e2f3a4b5-c6d7-4890-efab-901234567890',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '1 year ago',
+    tags: [
+      {
+        name: 'dolphin-mistral:7b',
+        size: '4.1 GB',
+        context: '32k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
+      },
+    ],
+  },
+  {
+    name: 'phi3.5',
+    description:
+      'Phi-3.5 is an updated small language model by Microsoft with improved multilingual and reasoning capabilities.',
+    estimated_pulls: '6.1M',
+    id: 'f3a4b5c6-d7e8-4901-fabc-012345678901',
+    first_seen: '2026-01-28T23:37:31.000+00:00',
+    model_last_updated: '8 months ago',
+    tags: [
+      {
+        name: 'phi3.5:3.8b',
+        size: '2.2 GB',
+        context: '128k',
+        input: 'Text',
+        cloud: false,
+        thinking: false,
       },
     ],
   },

admin/inertia/lib/api.ts

@@ -249,6 +249,7 @@ class API {
       const response = await this.client.get<{
         models: NomadOllamaModel[]
         hasMore: boolean
+        source?: 'api' | 'fallback'
       }>('/ollama/models', {
         params: { sort: 'pulls', ...params },
       })

admin/inertia/pages/settings/models.tsx

@@ -131,12 +131,20 @@ export default function ModelsPage(props: {
     initialData: { models: props.models.availableModels, hasMore: false },
   })
 
+  const isUsingFallback = availableModelData?.source === 'fallback'
+
   async function handleForceRefresh() {
     forceRefreshRef.current = true
     setIsForceRefreshing(true)
-    await refetch()
+    const result = await refetch()
     setIsForceRefreshing(false)
-    addNotification({ message: 'Model list refreshed from remote.', type: 'success' })
+    const refreshSource = result.data?.source
+    addNotification({
+      message: refreshSource === 'fallback'
+        ? 'Could not reach model service. Showing offline list.'
+        : 'Model list refreshed from remote.',
+      type: refreshSource === 'fallback' ? 'warning' : 'success',
+    })
   }
 
   async function handleInstallModel(modelName: string) {
@@ -312,6 +320,11 @@ export default function ModelsPage(props: {
               Refresh Models
             </StyledButton>
           </div>
+          {isUsingFallback && (
+            <Alert variant="info">
+              Showing offline model list. Connect to the internet and refresh for the full catalog.
+            </Alert>
+          )}
           <StyledTable<NomadOllamaModel>
             className="font-semibold mt-4"
             rowLines={true}