diff --git a/apps/client/public/locales/en-US/translation.json b/apps/client/public/locales/en-US/translation.json
index b622087b..95fbfc0c 100644
--- a/apps/client/public/locales/en-US/translation.json
+++ b/apps/client/public/locales/en-US/translation.json
@@ -1307,5 +1307,9 @@
"Page tree (child pages, recursive)": "Page tree (child pages, recursive)",
"Render the full nested tree of all descendant pages": "Render the full nested tree of all descendant pages",
"Showing {{count}} subpages_one": "Showing {{count}} subpage",
- "Showing {{count}} subpages_other": "Showing {{count}} subpages"
+ "Showing {{count}} subpages_other": "Showing {{count}} subpages",
+ "Protocol": "Protocol",
+ "How chat requests are sent and how reasoning is surfaced": "How chat requests are sent and how reasoning is surfaced",
+ "OpenAI-compatible (surfaces reasoning)": "OpenAI-compatible (surfaces reasoning)",
+ "OpenAI (official)": "OpenAI (official)"
}
diff --git a/apps/client/public/locales/ru-RU/translation.json b/apps/client/public/locales/ru-RU/translation.json
index 038323a8..0d4926cd 100644
--- a/apps/client/public/locales/ru-RU/translation.json
+++ b/apps/client/public/locales/ru-RU/translation.json
@@ -1160,5 +1160,9 @@
"Render the full nested tree of all descendant pages": "Показать полное вложенное дерево всех дочерних страниц",
"Showing {{count}} subpages_one": "Показано {{count}} подстраница",
"Showing {{count}} subpages_few": "Показано {{count}} подстраницы",
- "Showing {{count}} subpages_many": "Показано {{count}} подстраниц"
+ "Showing {{count}} subpages_many": "Показано {{count}} подстраниц",
+ "Protocol": "Протокол",
+ "How chat requests are sent and how reasoning is surfaced": "Как отправляются запросы чата и как показывается reasoning",
+ "OpenAI-compatible (surfaces reasoning)": "OpenAI-совместимый (показывает reasoning)",
+ "OpenAI (official)": "OpenAI (официальный)"
}
diff --git a/apps/client/src/features/workspace/components/settings/components/ai-provider-settings.tsx b/apps/client/src/features/workspace/components/settings/components/ai-provider-settings.tsx
index 713d9b65..08348756 100644
--- a/apps/client/src/features/workspace/components/settings/components/ai-provider-settings.tsx
+++ b/apps/client/src/features/workspace/components/settings/components/ai-provider-settings.tsx
@@ -38,6 +38,7 @@ import {
AiTestCapability,
IAiSettingsUpdate,
SttApiStyle,
+ ChatApiStyle,
} from "@/features/workspace/services/ai-settings-service.ts";
import { useAiRolesQuery } from "@/features/ai-chat/queries/ai-chat-query.ts";
import { IAiRole } from "@/features/ai-chat/types/ai-chat.types.ts";
@@ -82,6 +83,8 @@ const STT_LANGUAGE_OPTIONS: { value: string; label: string }[] = [
// (empty means "leave unchanged" unless explicitly cleared).
const formSchema = z.object({
chatModel: z.string(),
+ // Chat provider implementation (reasoning surfacing). Default openai-compatible.
+ chatApiStyle: z.enum(["openai-compatible", "openai"]),
// Cheap model id for the anonymous public-share assistant; empty = use chatModel.
publicShareChatModel: z.string(),
// Agent-role id whose persona the public-share assistant adopts; empty =
@@ -308,6 +311,7 @@ export default function AiProviderSettings() {
validate: zod4Resolver(formSchema),
initialValues: {
chatModel: "",
+ chatApiStyle: "openai-compatible" as ChatApiStyle,
publicShareChatModel: "",
publicShareAssistantRoleId: "",
embeddingModel: "",
@@ -330,6 +334,7 @@ export default function AiProviderSettings() {
if (!settings) return;
form.setValues({
chatModel: settings.chatModel ?? "",
+ chatApiStyle: settings.chatApiStyle ?? "openai-compatible",
publicShareChatModel: settings.publicShareChatModel ?? "",
publicShareAssistantRoleId: settings.publicShareAssistantRoleId ?? "",
embeddingModel: settings.embeddingModel ?? "",
@@ -359,6 +364,7 @@ export default function AiProviderSettings() {
// Everything is OpenAI-compatible.
driver: "openai",
chatModel: values.chatModel,
+ chatApiStyle: values.chatApiStyle,
// Cheap model id for the anonymous public-share assistant; empty falls
// back to chatModel server-side.
publicShareChatModel: values.publicShareChatModel,
@@ -761,6 +767,24 @@ export default function AiProviderSettings() {
{t("Resolves to {{url}}", { url: chatResolved })}
+
+
{/* Anonymous public-share assistant: a single master toggle + an
optional cheaper model id. Reuses this card's driver/URL/key. */}
diff --git a/apps/client/src/features/workspace/services/ai-settings-service.ts b/apps/client/src/features/workspace/services/ai-settings-service.ts
index 1814acd5..189589b0 100644
--- a/apps/client/src/features/workspace/services/ai-settings-service.ts
+++ b/apps/client/src/features/workspace/services/ai-settings-service.ts
@@ -9,6 +9,12 @@ export type AiDriver = "openai" | "gemini" | "ollama";
// - 'json' -> JSON body with base64-encoded audio (OpenRouter)
export type SttApiStyle = "multipart" | "json";
+// Chat provider implementation for the `openai` driver (chosen explicitly):
+// - 'openai-compatible' -> maps streamed reasoning_content to reasoning parts
+// (z.ai/GLM, DeepSeek, OpenRouter, ...). Default.
+// - 'openai' -> official provider; real-OpenAI reasoning-model shaping.
+export type ChatApiStyle = "openai-compatible" | "openai";
+
// Masked AI provider settings returned by the server.
// No API key is ever returned; only `hasApiKey` / `hasEmbeddingApiKey` indicate
// whether one is stored. `embeddingBaseUrl` is the RAW stored value (empty means
@@ -16,6 +22,7 @@ export type SttApiStyle = "multipart" | "json";
export interface IAiSettings {
driver?: AiDriver;
chatModel?: string;
+ chatApiStyle?: ChatApiStyle;
// Cheap model id for the anonymous public-share assistant; empty = chatModel.
publicShareChatModel?: string;
// Agent-role id whose persona the public-share assistant adopts; empty =
@@ -49,6 +56,7 @@ export interface IAiSettings {
export interface IAiSettingsUpdate {
driver?: AiDriver;
chatModel?: string;
+ chatApiStyle?: ChatApiStyle;
publicShareChatModel?: string;
// Agent-role id whose persona the public-share assistant adopts; empty =
// built-in locked persona.
diff --git a/apps/server/src/database/repos/workspace/workspace.repo.ts b/apps/server/src/database/repos/workspace/workspace.repo.ts
index 182a45f2..95e33aa9 100644
--- a/apps/server/src/database/repos/workspace/workspace.repo.ts
+++ b/apps/server/src/database/repos/workspace/workspace.repo.ts
@@ -239,7 +239,7 @@ export class WorkspaceRepo {
// is a real jsonb object, never a double-encoded string. The CASE self-heals
// workspaces whose settings.ai.provider was previously corrupted into an
// array/string.
- const ALLOWED = ['driver', 'chatModel', 'embeddingModel', 'baseUrl', 'embeddingBaseUrl', 'sttModel', 'sttBaseUrl', 'sttApiStyle', 'sttLanguage', 'systemPrompt', 'publicShareChatModel', 'publicShareAssistantRoleId'];
+ const ALLOWED = ['driver', 'chatModel', 'chatApiStyle', 'embeddingModel', 'baseUrl', 'embeddingBaseUrl', 'sttModel', 'sttBaseUrl', 'sttApiStyle', 'sttLanguage', 'systemPrompt', 'publicShareChatModel', 'publicShareAssistantRoleId'];
const entries = Object.entries(provider).filter(
([k, v]) => v !== undefined && ALLOWED.includes(k),
);
diff --git a/apps/server/src/integrations/ai/ai-settings.service.ts b/apps/server/src/integrations/ai/ai-settings.service.ts
index e556c0d0..0717c3c4 100644
--- a/apps/server/src/integrations/ai/ai-settings.service.ts
+++ b/apps/server/src/integrations/ai/ai-settings.service.ts
@@ -14,6 +14,7 @@ import {
MaskedAiSettings,
ResolvedAiConfig,
SttApiStyle,
+ ChatApiStyle,
} from './ai.types';
/**
@@ -24,6 +25,7 @@ import {
export interface UpdateAiSettingsInput {
driver?: AiDriver;
chatModel?: string;
+ chatApiStyle?: ChatApiStyle;
embeddingModel?: string;
baseUrl?: string;
embeddingBaseUrl?: string;
@@ -157,6 +159,8 @@ export class AiSettingsService {
const config: ResolvedAiConfig = {
driver: provider.driver,
chatModel: provider.chatModel,
+ // Plain passthrough; getChatModel defaults unset to 'openai-compatible'.
+ chatApiStyle: provider.chatApiStyle,
// Cheap model id for the anonymous public-share assistant; reuses the chat
// driver/baseUrl/apiKey. Empty/unset → callers fall back to chatModel.
publicShareChatModel: provider.publicShareChatModel,
@@ -238,6 +242,7 @@ export class AiSettingsService {
return {
driver: provider.driver,
chatModel: provider.chatModel,
+ chatApiStyle: provider.chatApiStyle,
embeddingModel: provider.embeddingModel,
baseUrl: provider.baseUrl,
embeddingBaseUrl: provider.embeddingBaseUrl,
@@ -278,6 +283,7 @@ export class AiSettingsService {
for (const key of [
'driver',
'chatModel',
+ 'chatApiStyle',
'embeddingModel',
'baseUrl',
'embeddingBaseUrl',
diff --git a/apps/server/src/integrations/ai/ai.service.spec.ts b/apps/server/src/integrations/ai/ai.service.spec.ts
index ef44a59d..b3c7f6f0 100644
--- a/apps/server/src/integrations/ai/ai.service.spec.ts
+++ b/apps/server/src/integrations/ai/ai.service.spec.ts
@@ -285,3 +285,64 @@ describe('AiService.getChatModel role model override', () => {
);
});
});
+
+/**
+ * Chat provider selection by the EXPLICIT `chatApiStyle` (NOT inferred from
+ * baseUrl): 'openai-compatible' (default) uses @ai-sdk/openai-compatible, which
+ * maps streamed reasoning_content to reasoning parts; 'openai' uses the official
+ * provider; and openai-compatible without a baseURL safely falls back to the
+ * official provider (it has no default endpoint). Asserted via `.provider`.
+ */
+describe('AiService.getChatModel chatApiStyle provider selection', () => {
+ function serviceWith(opts: {
+ baseUrl?: string;
+ chatApiStyle?: 'openai-compatible' | 'openai';
+ }) {
+ const aiSettings = {
+ resolve: jest.fn().mockResolvedValue({
+ driver: 'openai',
+ chatModel: 'glm-5.2',
+ apiKey: 'key',
+ baseUrl: opts.baseUrl,
+ chatApiStyle: opts.chatApiStyle,
+ }),
+ };
+ return new AiService(
+ // eslint-disable-next-line @typescript-eslint/no-explicit-any
+ aiSettings as any,
+ { find: jest.fn() } as never,
+ { decryptSecret: jest.fn() } as never,
+ );
+ }
+
+ const providerOf = async (svc: AiService) =>
+ (
+ (await svc.getChatModel('ws-1')) as { provider: string }
+ ).provider;
+
+ it("'openai-compatible' + baseURL -> openai-compatible provider", async () => {
+ expect(
+ await providerOf(
+ serviceWith({ baseUrl: 'https://api.z.ai/v4', chatApiStyle: 'openai-compatible' }),
+ ),
+ ).toContain('openai-compatible');
+ });
+
+ it("'openai' + baseURL -> official openai provider", async () => {
+ expect(
+ await providerOf(serviceWith({ baseUrl: 'https://api.z.ai/v4', chatApiStyle: 'openai' })),
+ ).toBe('openai.chat');
+ });
+
+ it('unset + baseURL -> defaults to openai-compatible', async () => {
+ expect(
+ await providerOf(serviceWith({ baseUrl: 'https://api.z.ai/v4' })),
+ ).toContain('openai-compatible');
+ });
+
+ it("'openai-compatible' WITHOUT baseURL -> safe fallback to official openai", async () => {
+ expect(
+ await providerOf(serviceWith({ chatApiStyle: 'openai-compatible' })),
+ ).toBe('openai.chat');
+ });
+});
diff --git a/apps/server/src/integrations/ai/ai.service.ts b/apps/server/src/integrations/ai/ai.service.ts
index 2a524f2c..18f15b5d 100644
--- a/apps/server/src/integrations/ai/ai.service.ts
+++ b/apps/server/src/integrations/ai/ai.service.ts
@@ -7,6 +7,7 @@ import {
type LanguageModel,
} from 'ai';
import { createOpenAI } from '@ai-sdk/openai';
+import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
import { createGoogleGenerativeAI } from '@ai-sdk/google';
import { createOllama } from 'ai-sdk-ollama';
import { AiSettingsService } from './ai-settings.service';
@@ -95,6 +96,10 @@ export class AiService {
let apiKey = cfg.apiKey;
let baseUrl = cfg.baseUrl;
+ // Chat provider implementation, chosen EXPLICITLY by the admin (not inferred
+ // from baseUrl). Unset → 'openai-compatible' so reasoning is surfaced by
+ // default for this fork's openai+baseUrl setups.
+ const chatApiStyle = cfg.chatApiStyle ?? 'openai-compatible';
// A driver override that differs from the workspace driver needs that
// driver's own creds (the workspace driver's key would be wrong/absent).
@@ -145,19 +150,41 @@ export class AiService {
}
switch (driver) {
- case 'openai':
- // baseURL (when set) covers openai-compatible endpoints. Use Chat
- // Completions (/chat/completions) — the portable OpenAI-compatible
- // endpoint. The default callable createOpenAI(...)(model) targets the
- // Responses API (/responses), which OpenAI-compatible gateways
- // (OpenRouter, etc.) reject on multi-turn requests (history with
- // assistant messages) → 400. The provider fetch is the instrumented
- // streaming fetch (finite-but-generous stream timeouts, #175).
+ case 'openai': {
+ // The provider implementation is chosen by the admin's `chatApiStyle`
+ // (NOT inferred from baseUrl — a custom URL can front real OpenAI too).
+ // Both branches hit Chat Completions (/chat/completions); the provider
+ // fetch is the instrumented streaming fetch (finite-but-generous stream
+ // timeouts, #175).
+ //
+ // 'openai-compatible' (default) maps the third-party provider's streamed
+ // `reasoning_content` to reasoning parts (z.ai/GLM, DeepSeek, ...) — the
+ // point of #175. It has no default endpoint, so it requires a baseURL;
+ // when there is none (real OpenAI, or a role's cross-driver override that
+ // cleared baseUrl) we fall back to the official provider.
+ if (chatApiStyle === 'openai-compatible' && baseUrl) {
+ return createOpenAICompatible({
+ name: 'openai-compatible',
+ apiKey,
+ baseURL: baseUrl,
+ // Keep streamed token usage (stream_options.include_usage): without
+ // it @ai-sdk/openai-compatible omits usage, zeroing the live token
+ // counter and reasoning-token metadata. The official provider always
+ // sent it, so this preserves parity.
+ includeUsage: true,
+ fetch: this.aiProviderFetch,
+ })(chatModel);
+ }
+ // Official @ai-sdk/openai: real-OpenAI reasoning-model request shaping;
+ // `.chat()` targets Chat Completions (the default callable targets the
+ // Responses API, which openai-compatible gateways 400 on multi-turn
+ // history). In this fork baseUrl is normally set; undefined = real OpenAI.
return createOpenAI({
apiKey,
baseURL: baseUrl,
fetch: this.aiProviderFetch,
}).chat(chatModel);
+ }
case 'gemini':
return createGoogleGenerativeAI({ apiKey })(chatModel);
case 'ollama':
diff --git a/apps/server/src/integrations/ai/ai.types.ts b/apps/server/src/integrations/ai/ai.types.ts
index 0a3d925e..5cdb6d1d 100644
--- a/apps/server/src/integrations/ai/ai.types.ts
+++ b/apps/server/src/integrations/ai/ai.types.ts
@@ -16,6 +16,15 @@ export const AI_DRIVERS: AiDriver[] = ['openai', 'gemini', 'ollama'];
export type SttApiStyle = 'multipart' | 'json';
export const STT_API_STYLES: SttApiStyle[] = ['multipart', 'json'];
+// Chat provider implementation for the `openai` driver. Chosen explicitly by the
+// admin (NOT inferred from baseUrl — a custom URL can front real OpenAI too).
+// 'openai-compatible' = @ai-sdk/openai-compatible: maps streamed
+// `reasoning_content` to reasoning parts (z.ai/GLM, DeepSeek, OpenRouter, ...).
+// 'openai' = official @ai-sdk/openai: real-OpenAI reasoning-model request shaping
+// (max_completion_tokens, the 'developer' role), no third-party reasoning map.
+export type ChatApiStyle = 'openai-compatible' | 'openai';
+export const CHAT_API_STYLES: ChatApiStyle[] = ['openai-compatible', 'openai'];
+
/**
* Non-secret provider settings persisted under `settings.ai.provider`.
* The API key is intentionally absent here.
@@ -23,6 +32,9 @@ export const STT_API_STYLES: SttApiStyle[] = ['multipart', 'json'];
export interface AiProviderSettings {
driver: AiDriver;
chatModel: string;
+ // Chat provider implementation for the `openai` driver. Unset → defaults to
+ // 'openai-compatible' (so reasoning is surfaced by default). See ChatApiStyle.
+ chatApiStyle?: ChatApiStyle;
embeddingModel?: string;
baseUrl?: string;
// Embedding-specific base URL. Falls back to `baseUrl` when empty/unset.
@@ -76,6 +88,7 @@ export interface ResolvedAiConfig extends Partial {
export interface MaskedAiSettings {
driver?: AiDriver;
chatModel?: string;
+ chatApiStyle?: ChatApiStyle;
embeddingModel?: string;
baseUrl?: string;
embeddingBaseUrl?: string;
diff --git a/apps/server/src/integrations/ai/dto/update-ai-settings.dto.ts b/apps/server/src/integrations/ai/dto/update-ai-settings.dto.ts
index 37fe8143..53aa8220 100644
--- a/apps/server/src/integrations/ai/dto/update-ai-settings.dto.ts
+++ b/apps/server/src/integrations/ai/dto/update-ai-settings.dto.ts
@@ -1,5 +1,12 @@
import { IsIn, IsOptional, IsString } from 'class-validator';
-import { AI_DRIVERS, AiDriver, STT_API_STYLES, SttApiStyle } from '../ai.types';
+import {
+ AI_DRIVERS,
+ AiDriver,
+ CHAT_API_STYLES,
+ ChatApiStyle,
+ STT_API_STYLES,
+ SttApiStyle,
+} from '../ai.types';
/**
* Admin update payload for the workspace AI provider settings.
@@ -18,6 +25,10 @@ export class UpdateAiSettingsDto {
@IsString()
chatModel?: string;
+ @IsOptional()
+ @IsIn(CHAT_API_STYLES)
+ chatApiStyle?: ChatApiStyle;
+
@IsOptional()
@IsString()
embeddingModel?: string;