feat(server): adapt gpt5 (#13478)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

- New Features
- Added GPT-5 family and made GPT-5/-mini the new defaults for Copilot
scenarios and prompts.

- Bug Fixes
- Improved streaming chunk formats and reasoning/text semantics,
consistent attachment mediaType handling, and more reliable reranking
via log-prob handling.

- Refactor
- Unified maxOutputTokens usage; removed per-call step caps and migrated
several tools to a unified inputSchema shape.

- Chores
- Upgraded AI SDK dependencies and bumped an internal dependency
version.

- Tests
- Updated mocks and tests to reference GPT-5 variants and new stream
formats.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

packages/backend/server/src/__tests__/copilot-provider.spec.ts

@@ -118,11 +118,11 @@ test.serial.before(async t => {
         enabled: true,
         scenarios: {
           image: 'flux-1/schnell',
-          rerank: 'gpt-4.1-mini',
-          complex_text_generation: 'gpt-4.1-mini',
-          coding: 'gpt-4.1-mini',
-          quick_decision_making: 'gpt-4.1-mini',
-          quick_text_generation: 'gpt-4.1-mini',
+          rerank: 'gpt-5-mini',
+          complex_text_generation: 'gpt-5-mini',
+          coding: 'gpt-5-mini',
+          quick_decision_making: 'gpt-5-mini',
+          quick_text_generation: 'gpt-5-mini',
           polish_and_summarize: 'gemini-2.5-flash',
         },
       },

packages/backend/server/src/__tests__/copilot.spec.ts

@@ -5,6 +5,7 @@ import { ProjectRoot } from '@affine-tools/utils/path';
 import { PrismaClient } from '@prisma/client';
 import type { TestFn } from 'ava';
 import ava from 'ava';
+import { nanoid } from 'nanoid';
 import Sinon from 'sinon';
 
 import { EventBus, JobQueue } from '../base';
@@ -1340,16 +1341,16 @@ test('TextStreamParser should format different types of chunks correctly', t =>
     textDelta: {
       chunk: {
         type: 'text-delta' as const,
-        textDelta: 'Hello world',
-      } as any,
+        text: 'Hello world',
+      },
       expected: 'Hello world',
       description: 'should format text-delta correctly',
     },
     reasoning: {
       chunk: {
-        type: 'reasoning' as const,
-        textDelta: 'I need to think about this',
-      } as any,
+        type: 'reasoning-delta' as const,
+        text: 'I need to think about this',
+      },
       expected: '\n> [!]\n> I need to think about this',
       description: 'should format reasoning as callout',
     },
@@ -1358,8 +1359,8 @@ test('TextStreamParser should format different types of chunks correctly', t =>
         type: 'tool-call' as const,
         toolName: 'web_search_exa' as const,
         toolCallId: 'test-id-1',
-        args: { query: 'test query', mode: 'AUTO' as const },
-      } as any,
+        input: { query: 'test query', mode: 'AUTO' as const },
+      },
       expected: '\n> [!]\n> \n> Searching the web "test query"\n> ',
       description: 'should format web search tool call correctly',
     },
@@ -1368,8 +1369,8 @@ test('TextStreamParser should format different types of chunks correctly', t =>
         type: 'tool-call' as const,
         toolName: 'web_crawl_exa' as const,
         toolCallId: 'test-id-2',
-        args: { url: 'https://example.com' },
-      } as any,
+        input: { url: 'https://example.com' },
+      },
       expected: '\n> [!]\n> \n> Crawling the web "https://example.com"\n> ',
       description: 'should format web crawl tool call correctly',
     },
@@ -1378,8 +1379,8 @@ test('TextStreamParser should format different types of chunks correctly', t =>
         type: 'tool-result' as const,
         toolName: 'web_search_exa' as const,
         toolCallId: 'test-id-1',
-        args: { query: 'test query', mode: 'AUTO' as const },
-        result: [
+        input: { query: 'test query', mode: 'AUTO' as const },
+        output: [
           {
             title: 'Test Title',
             url: 'https://test.com',
@@ -1406,7 +1407,7 @@ test('TextStreamParser should format different types of chunks correctly', t =>
       chunk: {
         type: 'error' as const,
         error: { type: 'testError', message: 'Test error message' },
-      } as any,
+      },
       errorMessage: 'Test error message',
       description: 'should throw error for error chunks',
     },
@@ -1436,78 +1437,85 @@ test('TextStreamParser should process a sequence of message chunks', t => {
     chunks: [
       // Reasoning chunks
       {
-        type: 'reasoning' as const,
-        textDelta: 'The user is asking about',
-      } as any,
+        id: nanoid(),
+        type: 'reasoning-delta' as const,
+        text: 'The user is asking about',
+      },
       {
-        type: 'reasoning' as const,
-        textDelta: ' recent advances in quantum computing',
-      } as any,
+        id: nanoid(),
+        type: 'reasoning-delta' as const,
+        text: ' recent advances in quantum computing',
+      },
       {
-        type: 'reasoning' as const,
-        textDelta: ' and how it might impact',
-      } as any,
+        id: nanoid(),
+        type: 'reasoning-delta' as const,
+        text: ' and how it might impact',
+      },
       {
-        type: 'reasoning' as const,
-        textDelta: ' cryptography and data security.',
-      } as any,
+        id: nanoid(),
+        type: 'reasoning-delta' as const,
+        text: ' cryptography and data security.',
+      },
       {
-        type: 'reasoning' as const,
-        textDelta:
-          ' I should provide information on quantum supremacy achievements',
-      } as any,
+        id: nanoid(),
+        type: 'reasoning-delta' as const,
+        text: ' I should provide information on quantum supremacy achievements',
+      },
 
       // Text delta
       {
+        id: nanoid(),
         type: 'text-delta' as const,
-        textDelta:
-          'Let me search for the latest breakthroughs in quantum computing and their ',
-      } as any,
+        text: 'Let me search for the latest breakthroughs in quantum computing and their ',
+      },
 
       // Tool call
       {
         type: 'tool-call' as const,
         toolCallId: 'toolu_01ABCxyz123456789',
         toolName: 'web_search_exa' as const,
-        args: {
+        input: {
           query: 'latest quantum computing breakthroughs cryptography impact',
         },
-      } as any,
+      },
 
       // Tool result
       {
         type: 'tool-result' as const,
         toolCallId: 'toolu_01ABCxyz123456789',
         toolName: 'web_search_exa' as const,
-        args: {
+        input: {
           query: 'latest quantum computing breakthroughs cryptography impact',
         },
-        result: [
+        output: [
           {
             title: 'IBM Unveils 1000-Qubit Quantum Processor',
             url: 'https://example.com/tech/quantum-computing-milestone',
           },
         ],
-      } as any,
+      },
 
       // More text deltas
       {
+        id: nanoid(),
         type: 'text-delta' as const,
-        textDelta: 'implications for security.',
-      } as any,
+        text: 'implications for security.',
+      },
       {
+        id: nanoid(),
         type: 'text-delta' as const,
-        textDelta: '\n\nQuantum computing has made ',
-      } as any,
+        text: '\n\nQuantum computing has made ',
+      },
       {
+        id: nanoid(),
         type: 'text-delta' as const,
-        textDelta: 'remarkable progress in the past year. ',
-      } as any,
+        text: 'remarkable progress in the past year. ',
+      },
       {
+        id: nanoid(),
         type: 'text-delta' as const,
-        textDelta:
-          'The development of more stable qubits has accelerated research significantly.',
-      } as any,
+        text: 'The development of more stable qubits has accelerated research significantly.',
+      },
     ],
     expected:
       '\n> [!]\n> The user is asking about recent advances in quantum computing and how it might impact cryptography and data security. I should provide information on quantum supremacy achievements\n\nLet me search for the latest breakthroughs in quantum computing and their \n> [!]\n> \n> Searching the web "latest quantum computing breakthroughs cryptography impact"\n> \n> \n> \n> [IBM Unveils 1000-Qubit Quantum Processor](https://example.com/tech/quantum-computing-milestone)\n> \n> \n> \n\nimplications for security.\n\nQuantum computing has made remarkable progress in the past year. The development of more stable qubits has accelerated research significantly.',

packages/backend/server/src/__tests__/mocks/copilot.mock.ts

@@ -57,15 +57,6 @@ export class MockCopilotProvider extends OpenAIProvider {
         },
       ],
     },
-    {
-      id: 'gpt-4.1',
-      capabilities: [
-        {
-          input: [ModelInputType.Text, ModelInputType.Image],
-          output: [ModelOutputType.Text, ModelOutputType.Object],
-        },
-      ],
-    },
     {
       id: 'gpt-4.1-2025-04-14',
       capabilities: [
@@ -76,7 +67,25 @@ export class MockCopilotProvider extends OpenAIProvider {
       ],
     },
     {
-      id: 'gpt-4.1-mini',
+      id: 'gpt-5',
+      capabilities: [
+        {
+          input: [ModelInputType.Text, ModelInputType.Image],
+          output: [ModelOutputType.Text, ModelOutputType.Object],
+        },
+      ],
+    },
+    {
+      id: 'gpt-5-2025-08-07',
+      capabilities: [
+        {
+          input: [ModelInputType.Text, ModelInputType.Image],
+          output: [ModelOutputType.Text, ModelOutputType.Object],
+        },
+      ],
+    },
+    {
+      id: 'gpt-5-mini',
       capabilities: [
         {
           input: [ModelInputType.Text, ModelInputType.Image],

packages/backend/server/src/__tests__/models/copilot-context.spec.ts

@@ -48,7 +48,7 @@ let docId = 'doc1';
 
 test.beforeEach(async t => {
   await t.context.module.initTestingDB();
-  await t.context.copilotSession.createPrompt('prompt-name', 'gpt-4.1');
+  await t.context.copilotSession.createPrompt('prompt-name', 'gpt-5-mini');
   user = await t.context.user.create({
     email: 'test@affine.pro',
   });

packages/backend/server/src/__tests__/models/copilot-session.spec.ts

@@ -58,9 +58,9 @@ const createTestPrompts = async (
   copilotSession: CopilotSessionModel,
   db: PrismaClient
 ) => {
-  await copilotSession.createPrompt(TEST_PROMPTS.NORMAL, 'gpt-4.1');
+  await copilotSession.createPrompt(TEST_PROMPTS.NORMAL, 'gpt-5-mini');
   await db.aiPrompt.create({
-    data: { name: TEST_PROMPTS.ACTION, model: 'gpt-4.1', action: 'edit' },
+    data: { name: TEST_PROMPTS.ACTION, model: 'gpt-5-mini', action: 'edit' },
   });
 };
 
@@ -116,7 +116,7 @@ const addMessagesToSession = async (
   await copilotSession.updateMessages({
     sessionId,
     userId: user.id,
-    prompt: { model: 'gpt-4.1' },
+    prompt: { model: 'gpt-5-mini' },
     messages: [
       {
         role: 'user',
@@ -807,7 +807,7 @@ test('should handle fork and session attachment operations', async t => {
       pinned: forkConfig.pinned,
       title: null,
       parentSessionId,
-      prompt: { name: TEST_PROMPTS.NORMAL, action: null, model: 'gpt-4.1' },
+      prompt: { name: TEST_PROMPTS.NORMAL, action: null, model: 'gpt-5-mini' },
       messages: [
         {
           role: 'user',