fix: align StackSelector endpoints + models with stack YAMLs

- Google: endpoint → /v1beta/openai/, model → gemini-2.0-flash
- Ollama: removed /v1 suffix from endpoint
- OpenRouter: model → anthropic/claude-sonnet-4
- Groq: model → llama-3.3-70b-versatile

Verified: stack switching via PUT /api/config re-registers provider and rebuilds orchestrator

Author: piotrlaczkowski

web/src/components/Settings/StackSelector.tsx

@@ -65,8 +65,8 @@ const STACKS: StackPreset[] = [
     description: 'Gemini 2.5 Pro — long context',
     icon: '💎',
     provider: 'google',
-    endpoint: 'https://generativelanguage.googleapis.com/v1beta',
-    model: 'gemini-2.5-pro-preview-03-25',
+    endpoint: 'https://generativelanguage.googleapis.com/v1beta/openai/',
+    model: 'gemini-2.0-flash',
     temperature: 0.15,
     max_tokens: 4096,
     max_parallel: 3,
@@ -103,7 +103,7 @@ const STACKS: StackPreset[] = [
     description: 'Run models locally via Ollama',
     icon: '🦙',
     provider: 'ollama',
-    endpoint: 'http://127.0.0.1:11434/v1',
+    endpoint: 'http://127.0.0.1:11434',
     model: 'qwen2.5-coder:14b',
     temperature: 0.15,
     max_tokens: 3072,
@@ -122,7 +122,7 @@ const STACKS: StackPreset[] = [
     icon: '🌐',
     provider: 'openrouter',
     endpoint: 'https://openrouter.ai/api/v1',
-    model: 'anthropic/claude-3.5-sonnet',
+    model: 'anthropic/claude-sonnet-4',
     temperature: 0.15,
     max_tokens: 4096,
     max_parallel: 3,
@@ -141,7 +141,7 @@ const STACKS: StackPreset[] = [
     icon: '⚙️',
     provider: 'groq',
     endpoint: 'https://api.groq.com/openai/v1',
-    model: 'llama-3.1-70b-versatile',
+    model: 'llama-3.3-70b-versatile',
     temperature: 0.15,
     max_tokens: 4096,
     max_parallel: 4,