chore(DELEG-002): update delegation pricing constants and fixture model names

src/components/dashboard/delegation/DelegationModelRoutingWidget.test.tsx

@@ -46,7 +46,7 @@ describe('DelegationModelRoutingWidget', () => {
         <DelegationModelRoutingWidget config={{}} />
       </DataSourceTestProvider>,
     );
-    expect(await screen.findByText('Qwen3-Coder-30B')).toBeInTheDocument();
+    expect(await screen.findByText('qwen3-coder-30b')).toBeInTheDocument();
     expect(screen.getByText('claude-sonnet-4-6')).toBeInTheDocument();
   });
 
@@ -57,7 +57,7 @@ describe('DelegationModelRoutingWidget', () => {
         <DelegationModelRoutingWidget config={{}} />
       </DataSourceTestProvider>,
     );
-    await screen.findByText('Qwen3-Coder-30B');
+    await screen.findByText('qwen3-coder-30b');
     expect(screen.getByText(/312 total delegations/i)).toBeInTheDocument();
   });
 
@@ -68,7 +68,7 @@ describe('DelegationModelRoutingWidget', () => {
         <DelegationModelRoutingWidget config={{}} />
       </DataSourceTestProvider>,
     );
-    await screen.findByText('Qwen3-Coder-30B');
+    await screen.findByText('qwen3-coder-30b');
     expect(screen.getByText('Model')).toBeInTheDocument();
     expect(screen.getByText('N')).toBeInTheDocument();
     expect(screen.getByText('Frequency')).toBeInTheDocument();
@@ -85,7 +85,7 @@ describe('DelegationModelRoutingWidget', () => {
         <DelegationModelRoutingWidget config={{}} />
       </DataSourceTestProvider>,
     );
-    await screen.findByText('Qwen3-Coder-30B');
+    await screen.findByText('qwen3-coder-30b');
     // Fixture provides task_types including 'code-review'
     expect(screen.getAllByText('code-review').length).toBeGreaterThan(0);
   });
@@ -97,7 +97,7 @@ describe('DelegationModelRoutingWidget', () => {
         <DelegationModelRoutingWidget config={{}} />
       </DataSourceTestProvider>,
     );
-    await screen.findByText('Qwen3-Coder-30B');
+    await screen.findByText('qwen3-coder-30b');
     expect(screen.getByText(/upstream-blocked/i)).toBeInTheDocument();
   });
 });

src/components/dashboard/delegation/DelegationQualityGateWidget.test.tsx

@@ -174,9 +174,9 @@ describe('DelegationQualityGateWidget', () => {
   describe('recent checks table', () => {
     it('renders recent checks table with task type, model, and pass/fail when decisions data is available', async () => {
       const decisions = [
-        { id: '1', task_type: 'code-review', model_name: 'Qwen3-Coder-30B', delegated_to: null, quality_gate_passed: 1, quality_gate_detail: 'schema_valid', created_at: '2026-05-05T10:00:00Z' },
-        { id: '2', task_type: 'summarization', model_name: 'glm-4-plus', delegated_to: null, quality_gate_passed: 0, quality_gate_detail: 'output_too_short', created_at: '2026-05-05T09:00:00Z' },
-        { id: '3', task_type: 'classification', model_name: null, delegated_to: 'codex-cli', quality_gate_passed: 1, quality_gate_detail: null, created_at: '2026-05-05T08:00:00Z' },
+        { id: '1', task_type: 'code-review', model_name: 'qwen3-coder-30b', delegated_to: null, quality_gate_passed: 1, quality_gate_detail: 'schema_valid', created_at: '2026-05-05T10:00:00Z' },
+        { id: '2', task_type: 'summarization', model_name: 'deepseek-r1-14b', delegated_to: null, quality_gate_passed: 0, quality_gate_detail: 'output_too_short', created_at: '2026-05-05T09:00:00Z' },
+        { id: '3', task_type: 'classification', model_name: null, delegated_to: 'openrouter-glm-flash', quality_gate_passed: 1, quality_gate_detail: null, created_at: '2026-05-05T08:00:00Z' },
       ];
       mockFetchForQualityGateWithDecisions([buildDelegationQualityGate()], decisions);
       render(
@@ -191,7 +191,7 @@ describe('DelegationQualityGateWidget', () => {
       expect(screen.getByText('classification')).toBeInTheDocument();
       // Model names and failure categories may appear in both the recent
       // checks table and other sections, so use getAllByText.
-      expect(screen.getAllByText('Qwen3-Coder-30B').length).toBeGreaterThanOrEqual(1);
+      expect(screen.getAllByText('qwen3-coder-30b').length).toBeGreaterThanOrEqual(1);
       expect(screen.getAllByText('output_too_short').length).toBeGreaterThanOrEqual(1);
       expect(screen.getByText('schema_valid')).toBeInTheDocument();
     });
@@ -243,11 +243,11 @@ describe('DelegationQualityGateWidget', () => {
       );
       await screen.findByText('Automated checks');
       expect(screen.getByText('Tokens-to-compliance by model')).toBeInTheDocument();
-      // The fixture seeds Qwen3-Coder-30B at 3,120 tokens (lowest, first row).
-      const models = screen.getAllByText(/Qwen3-Coder-30B|glm-4-plus|codex-cli|gemini-cli/);
+      // The fixture seeds qwen3-coder-30b at 3,120 tokens (lowest, first row).
+      const models = screen.getAllByText(/qwen3-coder-30b|deepseek-r1-14b|openrouter-glm-flash|claude-sonnet-4-6/);
       // Sorted ascending → most efficient first.
-      expect(models[0].textContent).toBe('Qwen3-Coder-30B');
-      expect(models[models.length - 1].textContent).toBe('gemini-cli');
+      expect(models[0].textContent).toBe('qwen3-coder-30b');
+      expect(models[models.length - 1].textContent).toBe('claude-sonnet-4-6');
     });
   });
 });

src/components/dashboard/delegation/DelegationSavingsWidget.tsx

@@ -56,6 +56,11 @@ export interface DelegationSavingsConfig {
 // Used when cost_savings_usd / savings_usd is 0 but token counts are present.
 // Prices in USD per token.
 
+// Client-side estimation fallback — used only when the backend projection
+// returns zero for savings_usd / cloud_cost_usd / local_cost_usd.
+// Authoritative savings are computed server-side via the pricing manifest
+// (omnimarket/src/omnimarket/pricing.py). These constants should NOT be
+// treated as the source of truth for cost calculations.
 const OPUS_INPUT_PRICE  = 15.0  / 1_000_000;  // $15/M tokens
 const OPUS_OUTPUT_PRICE = 75.0  / 1_000_000;  // $75/M tokens
 const SONNET_INPUT_PRICE  = 3.0  / 1_000_000;

src/components/dashboard/delegation/DelegationTokenUsageWidget.test.tsx

@@ -60,8 +60,8 @@ describe('DelegationTokenUsageWidget', () => {
       </DataSourceTestProvider>,
     );
     await screen.findByText('Total tokens');
-    expect(screen.getByText('Qwen3-Coder-30B')).toBeInTheDocument();
-    expect(screen.getByText('claude-sonnet-4-6')).toBeInTheDocument();
+    expect(screen.getByText('qwen3-coder-30b')).toBeInTheDocument();
+    expect(screen.getByText('deepseek-r1-14b')).toBeInTheDocument();
   });
 
   it('renders provenance badges when showProvenance is true', async () => {

src/storybook/fixtures/delegation-routing.ts

@@ -23,7 +23,7 @@ const SESSION_IDS = [
 ];
 
 const SESSION_TASK_TYPES = ['code-review', 'pattern-match', 'document-summarize', 'classification', 'code-review'];
-const SESSION_MODELS = ['Qwen3-Coder-30B', 'glm-4-plus', 'Qwen3-Coder-30B', 'codex-cli', 'claude-sonnet-4-6'];
+const SESSION_MODELS = ['qwen3-coder-30b', 'deepseek-r1-14b', 'qwen3-coder-30b', 'openrouter-glm-flash', 'claude-sonnet-4-6'];
 
 export function buildDelegationSavings(opts: BuildDelegationSavingsOptions = {}): DelegationSavingsProjection {
   const {
@@ -87,7 +87,7 @@ export interface BuildDelegationModelRoutingOptions {
   provisioned?: boolean;
 }
 
-const ROUTING_MODELS = ['Qwen3-Coder-30B', 'claude-sonnet-4-6', 'glm-4-plus', 'codex-cli'];
+const ROUTING_MODELS = ['qwen3-coder-30b', 'deepseek-r1-14b', 'openrouter-glm-flash', 'claude-sonnet-4-6'];
 const TASK_TYPES = ['code-review', 'classification', 'summarization', 'pattern-match'];
 
 export function buildDelegationModelRouting(
@@ -218,10 +218,10 @@ export function buildDelegationQualityGate(
           median_tokens_to_compliance: 3_950,
           avg_compliance_attempts: 1.32,
           tokens_to_compliance_by_model: [
-            { model_name: 'Qwen3-Coder-30B', avg_tokens: 3_120, avg_attempts: 1.18, sample_count: 142 },
-            { model_name: 'glm-4-plus', avg_tokens: 4_640, avg_attempts: 1.34, sample_count: 96 },
-            { model_name: 'codex-cli', avg_tokens: 5_280, avg_attempts: 1.51, sample_count: 64 },
-            { model_name: 'gemini-cli', avg_tokens: 6_910, avg_attempts: 1.78, sample_count: 38 },
+            { model_name: 'qwen3-coder-30b', avg_tokens: 3_120, avg_attempts: 1.18, sample_count: 142 },
+            { model_name: 'deepseek-r1-14b', avg_tokens: 4_640, avg_attempts: 1.34, sample_count: 96 },
+            { model_name: 'openrouter-glm-flash', avg_tokens: 5_280, avg_attempts: 1.51, sample_count: 64 },
+            { model_name: 'claude-sonnet-4-6', avg_tokens: 6_910, avg_attempts: 1.78, sample_count: 38 },
           ],
         }
       : {}),
@@ -237,10 +237,10 @@ export interface BuildDelegationTokenUsageOptions {
 }
 
 const TOKEN_MODELS: Array<{ id: string; name: string; promptRatio: number; costPerToken: number; provenance: TokenProvenance }> = [
-  { id: 'qwen3-coder-30b', name: 'Qwen3-Coder-30B', promptRatio: 0.78, costPerToken: 0, provenance: 'measured' },
-  { id: 'claude-sonnet-4-6', name: 'claude-sonnet-4-6', promptRatio: 0.72, costPerToken: 0.000003, provenance: 'measured' },
-  { id: 'glm-4-plus', name: 'glm-4-plus', promptRatio: 0.68, costPerToken: 0.0000015, provenance: 'estimated' },
-  { id: 'codex-cli', name: 'codex-cli', promptRatio: 0.65, costPerToken: 0.000002, provenance: 'estimated' },
+  { id: 'qwen3-coder-30b', name: 'qwen3-coder-30b', promptRatio: 0.78, costPerToken: 0, provenance: 'measured' },
+  { id: 'deepseek-r1-14b', name: 'deepseek-r1-14b', promptRatio: 0.72, costPerToken: 0.000003, provenance: 'measured' },
+  { id: 'openrouter-glm-flash', name: 'openrouter-glm-flash', promptRatio: 0.68, costPerToken: 0.0000015, provenance: 'estimated' },
+  { id: 'claude-sonnet-4-6', name: 'claude-sonnet-4-6', promptRatio: 0.65, costPerToken: 0.000002, provenance: 'estimated' },
 ];
 
 export function buildDelegationTokenUsage(

src/storybook/fixtures/delegation.ts

@@ -21,7 +21,7 @@ export interface BuildDelegationMetricsOptions {
 }
 
 const TASK_TYPES = ['code-review', 'pattern-match', 'document-summarize', 'classification'];
-const MODELS = ['Qwen3-Coder-30B', 'glm-4-plus', 'codex-cli', 'gemini-cli'];
+const MODELS = ['qwen3-coder-30b', 'deepseek-r1-14b', 'openrouter-glm-flash', 'claude-sonnet-4-6'];
 
 /**
  * Build a single `DelegationSummary` record. Story callers should