@lobehub/chat 1.64.2 → 1.65.0

package/src/libs/agent-runtime/anthropic/index.ts

@@ -97,12 +97,29 @@ export class LobeAnthropicAI implements LobeRuntimeAI {
   }
 
   private async buildAnthropicPayload(payload: ChatStreamPayload) {
-    const { messages, model, max_tokens = 4096, temperature, top_p, tools } = payload;
+    const { messages, model, max_tokens, temperature, top_p, tools, thinking } = payload;
     const system_message = messages.find((m) => m.role === 'system');
     const user_messages = messages.filter((m) => m.role !== 'system');
 
+    if (!!thinking) {
+      const maxTokens =
+        max_tokens ?? (thinking?.budget_tokens ? thinking?.budget_tokens + 4096 : 4096);
+
+      // `temperature` may only be set to 1 when thinking is enabled.
+      // `top_p` must be unset when thinking is enabled.
+      return {
+        max_tokens: maxTokens,
+        messages: await buildAnthropicMessages(user_messages),
+        model,
+        system: system_message?.content as string,
+
+        thinking,
+        tools: buildAnthropicTools(tools),
+      } satisfies Anthropic.MessageCreateParams;
+    }
+
     return {
-      max_tokens,
+      max_tokens: max_tokens ?? 4096,
       messages: await buildAnthropicMessages(user_messages),
       model,
       system: system_message?.content as string,
@@ -124,29 +141,30 @@ export class LobeAnthropicAI implements LobeRuntimeAI {
       method: 'GET',
     });
     const json = await response.json();
-  
+
     const modelList: AnthropicModelCard[] = json['data'];
-  
+
     return modelList
       .map((model) => {
-        const knownModel = LOBE_DEFAULT_MODEL_LIST.find((m) => model.id.toLowerCase() === m.id.toLowerCase());
+        const knownModel = LOBE_DEFAULT_MODEL_LIST.find(
+          (m) => model.id.toLowerCase() === m.id.toLowerCase(),
+        );
 
         return {
           contextWindowTokens: knownModel?.contextWindowTokens ?? undefined,
           displayName: model.display_name,
           enabled: knownModel?.enabled || false,
           functionCall:
-            model.id.toLowerCase().includes('claude-3')
-            || knownModel?.abilities?.functionCall
-            || false,
+            model.id.toLowerCase().includes('claude-3') ||
+            knownModel?.abilities?.functionCall ||
+            false,
           id: model.id,
-          reasoning:
-            knownModel?.abilities?.reasoning
-            || false,
+          reasoning: knownModel?.abilities?.reasoning || false,
           vision:
-            model.id.toLowerCase().includes('claude-3') && !model.id.toLowerCase().includes('claude-3-5-haiku')
-            || knownModel?.abilities?.vision
-            || false,
+            (model.id.toLowerCase().includes('claude-3') &&
+              !model.id.toLowerCase().includes('claude-3-5-haiku')) ||
+            knownModel?.abilities?.vision ||
+            false,
         };
       })
       .filter(Boolean) as ChatModelCard[];

package/src/libs/agent-runtime/types/chat.ts

@@ -88,8 +88,14 @@ export interface ChatStreamPayload {
    * @default 1
    */
   temperature: number;
+  /**
+   * use for Claude
+   */
+  thinking?: {
+    budget_tokens: number;
+    type: 'enabled' | 'disabled';
+  };
   tool_choice?: string;
-
   tools?: ChatCompletionTool[];
   /**
    * @title 控制生成文本中最高概率的单个令牌

package/src/libs/agent-runtime/utils/streams/anthropic.test.ts

@@ -384,6 +384,132 @@ describe('AnthropicStream', () => {
     expect(onToolCallMock).toHaveBeenCalledTimes(6);
   });
 
+  it('should handle thinking ', async () => {
+    const streams = [
+      {
+        type: 'message_start',
+        message: {
+          id: 'msg_01MNsLe7n1uVLtu6W8rCFujD',
+          type: 'message',
+          role: 'assistant',
+          model: 'claude-3-7-sonnet-20250219',
+          content: [],
+          stop_reason: null,
+          stop_sequence: null,
+          usage: {
+            input_tokens: 46,
+            cache_creation_input_tokens: 0,
+            cache_read_input_tokens: 0,
+            output_tokens: 11,
+          },
+        },
+      },
+      {
+        type: 'content_block_start',
+        index: 0,
+        content_block: { type: 'thinking', thinking: '', signature: '' },
+      },
+      {
+        type: 'content_block_delta',
+        index: 0,
+        delta: { type: 'thinking_delta', thinking: '我需要比较两个数字的' },
+      },
+      {
+        type: 'content_block_delta',
+        index: 0,
+        delta: { type: 'thinking_delta', thinking: '大小：9.8和9' },
+      },
+      {
+        type: 'content_block_delta',
+        index: 0,
+        delta: { type: 'thinking_delta', thinking: '11\n\n所以9.8比9.11大。' },
+      },
+      {
+        type: 'content_block_delta',
+        index: 0,
+        delta: {
+          type: 'signature_delta',
+          signature:
+            'EuYBCkQYAiJAHnHRJG4nPBrdTlo6CmXoyE8WYoQeoPiLnXaeuaM8ExdiIEkVvxK1DYXOz5sCubs2s/G1NsST8A003Zb8XmuhYBIMwDGMZSZ3+gxOEBpVGgzdpOlDNBTxke31SngiMKUk6WcSiA11OSVBuInNukoAhnRd5jPAEg7e5mIoz/qJwnQHV8I+heKUreP77eJdFipQaM3FHn+avEHuLa/Z/fu0O9BftDi+caB1UWDwJakNeWX1yYTvK+N1v4gRpKbj4AhctfYHMjq8qX9XTnXme5AGzCYC6HgYw2/RfalWzwNxI6k=',
+        },
+      },
+      { type: 'content_block_stop', index: 0 },
+      { type: 'content_block_start', index: 1, content_block: { type: 'text', text: '' } },
+      {
+        type: 'content_block_delta',
+        index: 1,
+        delta: { type: 'text_delta', text: '9.8比9.11大。' },
+      },
+      { type: 'content_block_stop', index: 1 },
+      {
+        type: 'message_delta',
+        delta: { stop_reason: 'end_turn', stop_sequence: null },
+        usage: { output_tokens: 354 },
+      },
+      { type: 'message_stop' },
+    ];
+
+    const mockReadableStream = new ReadableStream({
+      start(controller) {
+        streams.forEach((chunk) => {
+          controller.enqueue(chunk);
+        });
+        controller.close();
+      },
+    });
+
+    const protocolStream = AnthropicStream(mockReadableStream);
+
+    const decoder = new TextDecoder();
+    const chunks = [];
+
+    // @ts-ignore
+    for await (const chunk of protocolStream) {
+      chunks.push(decoder.decode(chunk, { stream: true }));
+    }
+
+    expect(chunks).toEqual(
+      [
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: data',
+        'data: {"id":"msg_01MNsLe7n1uVLtu6W8rCFujD","type":"message","role":"assistant","model":"claude-3-7-sonnet-20250219","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":46,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":11}}\n',
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: reasoning',
+        'data: ""\n',
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: reasoning',
+        'data: "我需要比较两个数字的"\n',
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: reasoning',
+        'data: "大小：9.8和9"\n',
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: reasoning',
+        'data: "11\\n\\n所以9.8比9.11大。"\n',
+        // Tool calls
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: reasoning_signature',
+        `data: "EuYBCkQYAiJAHnHRJG4nPBrdTlo6CmXoyE8WYoQeoPiLnXaeuaM8ExdiIEkVvxK1DYXOz5sCubs2s/G1NsST8A003Zb8XmuhYBIMwDGMZSZ3+gxOEBpVGgzdpOlDNBTxke31SngiMKUk6WcSiA11OSVBuInNukoAhnRd5jPAEg7e5mIoz/qJwnQHV8I+heKUreP77eJdFipQaM3FHn+avEHuLa/Z/fu0O9BftDi+caB1UWDwJakNeWX1yYTvK+N1v4gRpKbj4AhctfYHMjq8qX9XTnXme5AGzCYC6HgYw2/RfalWzwNxI6k="\n`,
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: data',
+        `data: {"type":"content_block_stop","index":0}\n`,
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: data',
+        `data: ""\n`,
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: text',
+        `data: "9.8比9.11大。"\n`,
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: data',
+        `data: {"type":"content_block_stop","index":1}\n`,
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: stop',
+        'data: "end_turn"\n',
+        'id: msg_01MNsLe7n1uVLtu6W8rCFujD',
+        'event: stop',
+        'data: "message_stop"\n',
+      ].map((item) => `${item}\n`),
+    );
+  });
   it('should handle ReadableStream input', async () => {
     const mockReadableStream = new ReadableStream({
       start(controller) {

package/src/libs/agent-runtime/utils/streams/anthropic.ts

@@ -14,12 +14,12 @@ import {
 
 export const transformAnthropicStream = (
   chunk: Anthropic.MessageStreamEvent,
-  stack: StreamContext,
+  context: StreamContext,
 ): StreamProtocolChunk => {
   // maybe need another structure to add support for multiple choices
   switch (chunk.type) {
     case 'message_start': {
-      stack.id = chunk.message.id;
+      context.id = chunk.message.id;
       return { data: chunk.message, id: chunk.message.id, type: 'data' };
     }
     case 'content_block_start': {
@@ -27,12 +27,12 @@ export const transformAnthropicStream = (
         const toolChunk = chunk.content_block;
 
         // if toolIndex is not defined, set it to 0
-        if (typeof stack.toolIndex === 'undefined') {
-          stack.toolIndex = 0;
+        if (typeof context.toolIndex === 'undefined') {
+          context.toolIndex = 0;
         }
         // if toolIndex is defined, increment it
         else {
-          stack.toolIndex += 1;
+          context.toolIndex += 1;
         }
 
         const toolCall: StreamToolCallChunkData = {
@@ -41,22 +41,36 @@ export const transformAnthropicStream = (
             name: toolChunk.name,
           },
           id: toolChunk.id,
-          index: stack.toolIndex,
+          index: context.toolIndex,
           type: 'function',
         };
 
-        stack.tool = { id: toolChunk.id, index: stack.toolIndex, name: toolChunk.name };
+        context.tool = { id: toolChunk.id, index: context.toolIndex, name: toolChunk.name };
 
-        return { data: [toolCall], id: stack.id, type: 'tool_calls' };
+        return { data: [toolCall], id: context.id, type: 'tool_calls' };
       }
 
-      return { data: chunk.content_block.text, id: stack.id, type: 'data' };
+      if (chunk.content_block.type === 'thinking') {
+        const thinkingChunk = chunk.content_block;
+
+        return { data: thinkingChunk.thinking, id: context.id, type: 'reasoning' };
+      }
+
+      if (chunk.content_block.type === 'redacted_thinking') {
+        return {
+          data: chunk.content_block.data,
+          id: context.id,
+          type: 'reasoning',
+        };
+      }
+
+      return { data: chunk.content_block.text, id: context.id, type: 'data' };
     }
 
     case 'content_block_delta': {
       switch (chunk.delta.type) {
         case 'text_delta': {
-          return { data: chunk.delta.text, id: stack.id, type: 'text' };
+          return { data: chunk.delta.text, id: context.id, type: 'text' };
         }
 
         case 'input_json_delta': {
@@ -64,34 +78,50 @@ export const transformAnthropicStream = (
 
           const toolCall: StreamToolCallChunkData = {
             function: { arguments: delta },
-            index: stack.toolIndex || 0,
+            index: context.toolIndex || 0,
             type: 'function',
           };
 
           return {
             data: [toolCall],
-            id: stack.id,
+            id: context.id,
             type: 'tool_calls',
           } as StreamProtocolToolCallChunk;
         }
 
+        case 'signature_delta': {
+          return {
+            data: chunk.delta.signature,
+            id: context.id,
+            type: 'reasoning_signature' as any,
+          };
+        }
+
+        case 'thinking_delta': {
+          return {
+            data: chunk.delta.thinking,
+            id: context.id,
+            type: 'reasoning',
+          };
+        }
+
         default: {
           break;
         }
       }
-      return { data: chunk, id: stack.id, type: 'data' };
+      return { data: chunk, id: context.id, type: 'data' };
     }
 
     case 'message_delta': {
-      return { data: chunk.delta.stop_reason, id: stack.id, type: 'stop' };
+      return { data: chunk.delta.stop_reason, id: context.id, type: 'stop' };
     }
 
     case 'message_stop': {
-      return { data: 'message_stop', id: stack.id, type: 'stop' };
+      return { data: 'message_stop', id: context.id, type: 'stop' };
     }
 
     default: {
-      return { data: chunk, id: stack.id, type: 'data' };
+      return { data: chunk, id: context.id, type: 'data' };
     }
   }
 };

package/src/libs/agent-runtime/utils/streams/protocol.ts

@@ -12,6 +12,10 @@ export interface StreamContext {
    * this flag is used to check if the pplx citation is returned,and then not return it again
    */
   returnedPplxCitation?: boolean;
+  thinking?: {
+    id: string;
+    name: string;
+  };
   tool?: {
     id: string;
     index: number;

package/src/locales/default/chat.ts

@@ -32,7 +32,13 @@ export default {
   },
   duplicateTitle: '{{title}} 副本',
   emptyAgent: '暂无助手',
-  extendControls: {
+  extendParams: {
+    enableReasoning: {
+      title: '开启深度思考',
+    },
+    reasoningBudgetToken: {
+      title: '思考消耗 Token',
+    },
     title: '模型扩展功能',
   },
   historyRange: '历史范围',

package/src/services/chat.ts

@@ -214,9 +214,35 @@ class ChatService {
 
     const tools = shouldUseTools ? filterTools : undefined;
 
+    // ============  3. process extend params   ============ //
+
+    let extendParams: Record<string, any> = {};
+
+    const isModelHasExtendParams = aiModelSelectors.isModelHasExtendParams(
+      payload.model,
+      payload.provider!,
+    )(useAiInfraStore.getState());
+
+    // model
+    if (isModelHasExtendParams) {
+      const modelExtendParams = aiModelSelectors.modelExtendParams(
+        payload.model,
+        payload.provider!,
+      )(useAiInfraStore.getState());
+      // if model has extended params, then we need to check if the model can use reasoning
+
+      if (modelExtendParams!.includes('enableReasoning') && chatConfig.enableReasoning) {
+        extendParams.thinking = {
+          budget_tokens: chatConfig.reasoningBudgetToken || 1024,
+          type: 'enabled',
+        };
+      }
+    }
+
     return this.getChatCompletion(
       {
         ...params,
+        ...extendParams,
         enabledSearch: enabledSearch && isModelHasBuiltinSearch ? true : undefined,
         messages: oaiMessages,
         tools,

package/src/store/agent/slices/chat/__snapshots__/selectors.test.ts.snap

@@ -8,7 +8,9 @@ exports[`agentSelectors > defaultAgentConfig > should merge DEFAULT_AGENT_CONFIG
     "enableAutoCreateTopic": true,
     "enableCompressHistory": true,
     "enableHistoryCount": true,
+    "enableReasoning": true,
     "historyCount": 8,
+    "reasoningBudgetToken": 1024,
     "searchMode": "off",
   },
   "model": "gpt-3.5-turbo",

package/src/store/aiInfra/slices/aiModel/selectors.ts

@@ -70,14 +70,14 @@ const modelContextWindowTokens = (id: string, provider: string) => (s: AIProvide
   return model?.contextWindowTokens;
 };
 
-const modelExtendControls = (id: string, provider: string) => (s: AIProviderStoreState) => {
+const modelExtendParams = (id: string, provider: string) => (s: AIProviderStoreState) => {
   const model = getEnabledModelById(id, provider)(s);
 
-  return model?.settings?.extendControls;
+  return model?.settings?.extendParams;
 };
 
-const isModelHasExtendControls = (id: string, provider: string) => (s: AIProviderStoreState) => {
-  const controls = modelExtendControls(id, provider)(s);
+const isModelHasExtendParams = (id: string, provider: string) => (s: AIProviderStoreState) => {
+  const controls = modelExtendParams(id, provider)(s);
 
   return !!controls && controls.length > 0;
 };
@@ -119,13 +119,13 @@ export const aiModelSelectors = {
   isModelHasBuiltinSearch,
   isModelHasBuiltinSearchConfig,
   isModelHasContextWindowToken,
-  isModelHasExtendControls,
+  isModelHasExtendParams,
   isModelLoading,
   isModelSupportReasoning,
   isModelSupportToolUse,
   isModelSupportVision,
   modelBuiltinSearchImpl,
   modelContextWindowTokens,
-  modelExtendControls,
+  modelExtendParams,
   totalAiProviderModelList,
 };