Expose Assistant fallback model route

docs/release-control/v6/internal/subsystems/ai-runtime.md

@@ -141,7 +141,11 @@ runtime cost control, and shared AI transport surfaces.
    as realtime, audio, moderation, embedding, and content-safety catalog
    entries. Once visible output has streamed, Pulse must not silently switch
    providers for that turn; the error belongs to the visible attempt and is
-   surfaced through normal failed-turn recovery.
+   surfaced through normal failed-turn recovery. Assistant completion events
+   must carry the effective model route that actually completed the turn, and
+   the drawer must update the in-flight transcript row when `provider_fallback`
+   names the next route so message labels, cost context, retry decisions, and
+   model-route recovery do not continue to point at the failed provider.
    Streamed provider startup must be bounded by the configured Assistant request
    timeout and the OpenAI-compatible SSE response-header guard; transient
    startup failures may retry once before surfacing failed-turn recovery, but a

docs/release-control/v6/internal/subsystems/api-contracts.md

@@ -448,7 +448,10 @@ payload shape change when the portal presents compact client rows.
     create-session request. The generated frontend union, stream parser tests,
     and backend JSON snapshot proof must stay in lockstep with that payload;
     `done.session_id` and `question.session_id` remain compatibility payloads,
-    not the primary cold-session creation contract.
+    not the primary cold-session creation contract. Assistant `done` events
+    must also carry the effective `model` route that completed the stream so
+    provider fallback, transcript labels, retries, and model-route recovery stay
+    tied to the actual responding provider rather than the failed request route.
 34. `internal/api/ai_handlers.go` shared with `ai-runtime`: AI settings and remediation handlers are both an AI runtime control surface and a canonical API payload contract boundary.
     Provider test responses from `/api/ai/test` and provider-specific
     `/api/ai/test/{provider}` preflight responses must return one safe

frontend-modern/src/api/__tests__/aiChatEvents.test.ts

@@ -2,6 +2,7 @@ import { describe, expect, it } from 'vitest';
 import aiChatEventsSource from '@/api/generated/aiChatEvents.ts?raw';
 import type {
   AIChatStreamEvent,
+  DoneData,
   SessionData,
   WorkflowStateData,
 } from '@/api/generated/aiChatEvents';
@@ -38,4 +39,17 @@ describe('AI chat stream event contract', () => {
     expect(aiChatEventsSource).toContain('failed_model?: string');
     expect(aiChatEventsSource).toContain('next_model?: string');
   });
+
+  it('exposes the effective completion model on done events', () => {
+    const done: DoneData = {
+      session_id: 'sess-stream',
+      model: 'deepseek:deepseek-v4-pro',
+      input_tokens: 904,
+      output_tokens: 30,
+    };
+    const event: AIChatStreamEvent = { type: 'done', data: done };
+
+    expect(event.data?.model).toBe('deepseek:deepseek-v4-pro');
+    expect(aiChatEventsSource).toContain('model?: string');
+  });
 });

frontend-modern/src/api/generated/aiChatEvents.ts

@@ -55,6 +55,7 @@ export interface ContentData {
 
 export interface DoneData {
   session_id?: string;
+  model?: string;
   input_tokens?: number;
   output_tokens?: number;
 }

frontend-modern/src/components/AI/Chat/__tests__/useChat.test.ts

@@ -934,6 +934,34 @@ describe('useChat', () => {
       dispose();
     });
 
+    it('updates the in-flight message model when provider fallback starts', async () => {
+      const { getFireEvent } = setupWithEventCapture();
+      const { value: chat, dispose } = withRoot(() =>
+        useChat({ sessionId: 's', model: 'openrouter:openai/gpt-4o-mini' }),
+      );
+
+      await chat.sendMessage('hi');
+      const fire = getFireEvent();
+
+      fire({
+        type: 'workflow_state',
+        data: {
+          phase: 'provider_fallback',
+          message: 'OpenRouter did not start a response; trying Gemini.',
+          failed_model: 'openrouter:openai/gpt-4o-mini',
+          next_model: 'gemini:gemini-3.1-flash-lite',
+        },
+      });
+
+      const assistant = chat.messages().find((m) => m.role === 'assistant')!;
+      expect(assistant.model).toBe('gemini:gemini-3.1-flash-lite');
+      expect(assistant.workflowStatus).toEqual({
+        phase: 'provider_fallback',
+        message: 'OpenRouter did not start a response; trying Gemini.',
+      });
+      dispose();
+    });
+
     it('processes tool_start events', async () => {
       const { getFireEvent } = setupWithEventCapture();
       const { value: chat, dispose } = withRoot(() => useChat({ sessionId: 's' }));
@@ -1238,10 +1266,18 @@ describe('useChat', () => {
       const fire = getFireEvent();
 
       fire({ type: 'content', data: 'response' });
-      fire({ type: 'done', data: { input_tokens: 100, output_tokens: 50 } });
+      fire({
+        type: 'done',
+        data: {
+          model: 'gemini:gemini-3.1-flash-lite',
+          input_tokens: 100,
+          output_tokens: 50,
+        },
+      });
 
       const assistant = chat.messages().find((m) => m.role === 'assistant')!;
       expect(assistant.isStreaming).toBe(false);
+      expect(assistant.model).toBe('gemini:gemini-3.1-flash-lite');
       expect(assistant.tokens).toEqual({ input: 100, output: 50 });
       expect(assistant.pendingTools).toHaveLength(0);
       dispose();

frontend-modern/src/components/AI/Chat/hooks/useChat.ts

@@ -321,6 +321,20 @@ export function useChat(options: UseChatOptions = {}) {
     };
   };
 
+  const extractCompletedModel = (data: unknown): string => {
+    if (!data || typeof data !== 'object') return '';
+    const record = data as Record<string, unknown>;
+    const modelRoute = record.model ?? record.model_route ?? record.modelRoute;
+    return typeof modelRoute === 'string' ? modelRoute.trim() : '';
+  };
+
+  const extractWorkflowNextModel = (data: unknown): string => {
+    if (!data || typeof data !== 'object') return '';
+    const record = data as Record<string, unknown>;
+    const nextModel = record.next_model ?? record.nextModel;
+    return typeof nextModel === 'string' ? nextModel.trim() : '';
+  };
+
   const extractErrorMessage = (data: unknown): string => {
     if (typeof data === 'string') return data;
     if (data && typeof data === 'object') {
@@ -519,7 +533,13 @@ export function useChat(options: UseChatOptions = {}) {
 
             case 'workflow_state': {
               const workflowStatus = extractWorkflowStatus(event.data);
-              return workflowStatus ? { ...msg, workflowStatus } : msg;
+              const nextModel = extractWorkflowNextModel(event.data);
+              if (!workflowStatus && !nextModel) return msg;
+              return {
+                ...msg,
+                ...(workflowStatus ? { workflowStatus } : {}),
+                ...(nextModel ? { model: nextModel } : {}),
+              };
             }
 
             case 'tool_start': {
@@ -791,10 +811,12 @@ export function useChat(options: UseChatOptions = {}) {
                   }
                 : msg;
               const tokens = extractTokens(event.data);
+              const completedModel = extractCompletedModel(event.data);
               if (tokens && (tokens.input > 0 || tokens.output > 0)) {
                 return {
                   ...flushedMsg,
                   isStreaming: false,
+                  ...(completedModel ? { model: completedModel } : {}),
                   pendingTools: [],
                   tokens,
                   workflowStatus: undefined,
@@ -803,6 +825,7 @@ export function useChat(options: UseChatOptions = {}) {
               return {
                 ...flushedMsg,
                 isStreaming: false,
+                ...(completedModel ? { model: completedModel } : {}),
                 pendingTools: [],
                 workflowStatus: undefined,
               };

internal/ai/chat/service.go

@@ -942,6 +942,7 @@ func (s *Service) ExecuteStream(ctx context.Context, req ExecuteRequest, callbac
 	// Send done event with token usage for this request.
 	doneData, _ := json.Marshal(DoneData{
 		SessionID:    session.ID,
+		Model:        selectedModel,
 		InputTokens:  loop.GetTotalInputTokens(),
 		OutputTokens: loop.GetTotalOutputTokens(),
 	})
@@ -2304,6 +2305,7 @@ func (s *Service) ExecutePatrolStream(ctx context.Context, req PatrolRequest, ca
 	// Send done event
 	doneData, _ := json.Marshal(DoneData{
 		SessionID:    session.ID,
+		Model:        patrolModel,
 		InputTokens:  tempLoop.GetTotalInputTokens(),
 		OutputTokens: tempLoop.GetTotalOutputTokens(),
 	})

internal/ai/chat/service_execute_additional_test.go

@@ -431,6 +431,7 @@ func TestService_ExecuteStream_FallsBackWhenPrimaryProviderFailsBeforeVisibleOut
 	var content strings.Builder
 	var errorEvents int
 	var fallbackEvents int
+	var doneModel string
 	err = service.ExecuteStream(context.Background(), ExecuteRequest{
 		SessionID: "fallback-before-visible-output",
 		Prompt:    "reply",
@@ -452,6 +453,12 @@ func TestService_ExecuteStream_FallsBackWhenPrimaryProviderFailsBeforeVisibleOut
 			if data.Phase == "provider_fallback" {
 				fallbackEvents++
 			}
+		case "done":
+			var data DoneData
+			if err := json.Unmarshal(event.Data, &data); err != nil {
+				t.Fatalf("unmarshal done: %v", err)
+			}
+			doneModel = data.Model
 		}
 	})
 	if err != nil {
@@ -466,6 +473,9 @@ func TestService_ExecuteStream_FallsBackWhenPrimaryProviderFailsBeforeVisibleOut
 	if fallbackEvents != 1 {
 		t.Fatalf("fallback workflow events = %d, want 1", fallbackEvents)
 	}
+	if doneModel != "gemini:gemini-test" {
+		t.Fatalf("done model = %q, want fallback model", doneModel)
+	}
 
 	messages, err := store.GetMessages("fallback-before-visible-output")
 	if err != nil {

internal/ai/chat/types.go

@@ -359,6 +359,7 @@ type ErrorData struct {
 // DoneData is the data for "done" events
 type DoneData struct {
 	SessionID    string `json:"session_id,omitempty"`
+	Model        string `json:"model,omitempty"`
 	InputTokens  int    `json:"input_tokens,omitempty"`
 	OutputTokens int    `json:"output_tokens,omitempty"`
 }