diff --git a/packages/aiCore/AI_SDK_ARCHITECTURE.md b/packages/aiCore/AI_SDK_ARCHITECTURE.md index 2e4b661e53..67af20c0a6 100644 --- a/packages/aiCore/AI_SDK_ARCHITECTURE.md +++ b/packages/aiCore/AI_SDK_ARCHITECTURE.md @@ -300,10 +300,7 @@ export interface AiPlugin { - 函数式设计，简化使用 ```typescript -export function wrapModelWithMiddlewares( - model: LanguageModel, - middlewares: LanguageModelV1Middleware[] -): LanguageModel +export function wrapModelWithMiddlewares(model: LanguageModel, middlewares: LanguageModelV1Middleware[]): LanguageModel ``` ### 4.5 Provider System (提供商系统) @@ -403,7 +400,7 @@ const model = await createModel({ import { Agent, run } from '@openai/agents' const agent = new Agent({ - model, // ✅ 直接兼容 LanguageModel 接口 + model, // ✅ 直接兼容 LanguageModel 接口 name: 'Assistant', instructions: '...', tools: [tool1, tool2] diff --git a/packages/aiCore/README.md b/packages/aiCore/README.md index a887097b61..4ca5ea6640 100644 --- a/packages/aiCore/README.md +++ b/packages/aiCore/README.md @@ -5,28 +5,33 @@ Cherry Studio AI Core 是一个基于 Vercel AI SDK 的统一 AI Provider 接口 ## ✨ 核心亮点 ### 🏗️ 优雅的架构设计 + - **简化分层**：`models`（模型层）→ `runtime`（运行时层），清晰的职责分离 - **函数式优先**：避免过度抽象，提供简洁直观的 API - **类型安全**：完整的 TypeScript 支持，直接复用 AI SDK 类型系统 - **最小包装**：直接使用 AI SDK 的接口，避免重复定义和性能损耗 ### 🔌 强大的插件系统 + - **生命周期钩子**：支持请求全生命周期的扩展点 - **流转换支持**：基于 AI SDK 的 `experimental_transform` 实现流处理 - **插件分类**：First、Sequential、Parallel 三种钩子类型，满足不同场景 - **内置插件**：webSearch、logging、toolUse 等开箱即用的功能 ### 🌐 统一多 Provider 接口 + - **扩展注册**：支持自定义 Provider 注册，无限扩展能力 - **配置统一**：统一的配置接口，简化多 Provider 管理 ### 🚀 多种使用方式 + - **函数式调用**：适合简单场景的直接函数调用 - **执行器实例**：适合复杂场景的可复用执行器 - **静态工厂**：便捷的静态创建方法 - **原生兼容**：完全兼容 AI SDK 原生 Provider Registry ### 🔮 面向未来 + - **Agent 就绪**：为 OpenAI Agents SDK 集成预留架构空间 - **模块化设计**：独立包结构，支持跨项目复用 - **渐进式迁移**：可以逐步从现有 AI SDK 代码迁移 @@ -181,6 +186,7 @@ AI Core 提供了强大的插件系统，支持请求全生命周期的扩展。 ### 内置插件 #### webSearchPlugin - 网络搜索插件 + 为不同 AI Provider 提供统一的网络搜索能力： ```typescript @@ -188,9 +194,13 @@ import { webSearchPlugin } from '@cherrystudio/ai-core/built-in/plugins' const executor = AiCore.create('openai', { apiKey: 'your-key' }, [ webSearchPlugin({ - openai: { /* OpenAI 搜索配置 */ }, + openai: { + /* OpenAI 搜索配置 */ + }, anthropic: { maxUses: 5 }, - google: { /* Google 搜索配置 */ }, + google: { + /* Google 搜索配置 */ + }, xai: { mode: 'on', returnCitations: true, @@ -202,6 +212,7 @@ const executor = AiCore.create('openai', { apiKey: 'your-key' }, [ ``` #### loggingPlugin - 日志插件 + 提供详细的请求日志记录： ```typescript @@ -217,24 +228,29 @@ const executor = AiCore.create('openai', { apiKey: 'your-key' }, [ ``` #### promptToolUsePlugin - 提示工具使用插件 + 为不支持原生 Function Call 的模型提供 prompt 方式的工具调用： ```typescript import { createPromptToolUsePlugin } from '@cherrystudio/ai-core/built-in/plugins' // 对于不支持 function call 的模型 -const executor = AiCore.create('providerId', { - apiKey: 'your-key', - baseURL: 'https://your-model-endpoint' -}, [ - createPromptToolUsePlugin({ - enabled: true, - // 可选：自定义系统提示符构建 - buildSystemPrompt: (userPrompt, tools) => { - return `${userPrompt}\n\nAvailable tools: ${Object.keys(tools).join(', ')}` - } - }) -]) +const executor = AiCore.create( + 'providerId', + { + apiKey: 'your-key', + baseURL: 'https://your-model-endpoint' + }, + [ + createPromptToolUsePlugin({ + enabled: true, + // 可选：自定义系统提示符构建 + buildSystemPrompt: (userPrompt, tools) => { + return `${userPrompt}\n\nAvailable tools: ${Object.keys(tools).join(', ')}` + } + }) + ] +) ``` ### 自定义插件 @@ -402,6 +418,7 @@ await client.streamObject({ - [AI SDK Providers](https://ai-sdk.dev/providers/ai-sdk-providers) ## 未来版本 + - 🔮 多 Agent 编排 - 🔮 可视化插件配置 - 🔮 实时监控和分析 diff --git a/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/StreamEventManager.ts b/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/StreamEventManager.ts new file mode 100644 index 0000000000..197b20e9b4 --- /dev/null +++ b/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/StreamEventManager.ts @@ -0,0 +1,139 @@ +/** + * 流事件管理器 + * + * 负责处理 AI SDK 流事件的发送和管理 + * 从 promptToolUsePlugin.ts 中提取出来以降低复杂度 + */ +import type { ModelMessage } from 'ai' + +import type { AiRequestContext } from '../../types' +import type { StreamController } from './ToolExecutor' + +/** + * 流事件管理器类 + */ +export class StreamEventManager { + /** + * 发送工具调用步骤开始事件 + */ + sendStepStartEvent(controller: StreamController): void { + controller.enqueue({ + type: 'start-step', + request: {}, + warnings: [] + }) + } + + /** + * 发送步骤完成事件 + */ + sendStepFinishEvent(controller: StreamController, chunk: any): void { + controller.enqueue({ + type: 'finish-step', + finishReason: 'stop', + response: chunk.response, + usage: chunk.usage, + providerMetadata: chunk.providerMetadata + }) + } + + /** + * 处理递归调用并将结果流接入当前流 + */ + async handleRecursiveCall( + controller: StreamController, + recursiveParams: any, + context: AiRequestContext, + stepId: string + ): Promise { + try { + console.log('[MCP Prompt] Starting recursive call after tool execution...') + + const recursiveResult = await context.recursiveCall(recursiveParams) + + if (recursiveResult && recursiveResult.fullStream) { + await this.pipeRecursiveStream(controller, recursiveResult.fullStream) + } else { + console.warn('[MCP Prompt] No fullstream found in recursive result:', recursiveResult) + } + } catch (error) { + this.handleRecursiveCallError(controller, error, stepId) + } + } + + /** + * 将递归流的数据传递到当前流 + */ + private async pipeRecursiveStream(controller: StreamController, recursiveStream: ReadableStream): Promise { + const reader = recursiveStream.getReader() + try { + while (true) { + const { done, value } = await reader.read() + if (done) { + break + } + if (value.type === 'finish') { + // 迭代的流不发finish + break + } + // 将递归流的数据传递到当前流 + controller.enqueue(value) + } + } finally { + reader.releaseLock() + } + } + + /** + * 处理递归调用错误 + */ + private handleRecursiveCallError(controller: StreamController, error: unknown, stepId: string): void { + console.error('[MCP Prompt] Recursive call failed:', error) + + // 使用 AI SDK 标准错误格式，但不中断流 + controller.enqueue({ + type: 'error', + error: { + message: error instanceof Error ? error.message : String(error), + name: error instanceof Error ? error.name : 'RecursiveCallError' + } + }) + + // 继续发送文本增量，保持流的连续性 + controller.enqueue({ + type: 'text-delta', + id: stepId, + text: '\n\n[工具执行后递归调用失败，继续对话...]' + }) + } + + /** + * 构建递归调用的参数 + */ + buildRecursiveParams(context: AiRequestContext, textBuffer: string, toolResultsText: string, tools: any): any { + // 构建新的对话消息 + const newMessages: ModelMessage[] = [ + ...(context.originalParams.messages || []), + { + role: 'assistant', + content: textBuffer + }, + { + role: 'user', + content: toolResultsText + } + ] + + // 递归调用，继续对话，重新传递 tools + const recursiveParams = { + ...context.originalParams, + messages: newMessages, + tools: tools + } + + // 更新上下文中的消息 + context.originalParams.messages = newMessages + + return recursiveParams + } +} diff --git a/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/ToolExecutor.ts b/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/ToolExecutor.ts new file mode 100644 index 0000000000..ec174fa2ea --- /dev/null +++ b/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/ToolExecutor.ts @@ -0,0 +1,156 @@ +/** + * 工具执行器 + * + * 负责工具的执行、结果格式化和相关事件发送 + * 从 promptToolUsePlugin.ts 中提取出来以降低复杂度 + */ +import type { ToolSet } from 'ai' + +import type { ToolUseResult } from './type' + +/** + * 工具执行结果 + */ +export interface ExecutedResult { + toolCallId: string + toolName: string + result: any + isError?: boolean +} + +/** + * 流控制器类型（从 AI SDK 提取） + */ +export interface StreamController { + enqueue(chunk: any): void +} + +/** + * 工具执行器类 + */ +export class ToolExecutor { + /** + * 执行多个工具调用 + */ + async executeTools( + toolUses: ToolUseResult[], + tools: ToolSet, + controller: StreamController + ): Promise { + const executedResults: ExecutedResult[] = [] + + for (const toolUse of toolUses) { + try { + const tool = tools[toolUse.toolName] + if (!tool || typeof tool.execute !== 'function') { + throw new Error(`Tool "${toolUse.toolName}" has no execute method`) + } + + // 发送工具调用开始事件 + this.sendToolStartEvents(controller, toolUse) + + console.log(`[MCP Prompt Stream] Executing tool: ${toolUse.toolName}`, toolUse.arguments) + + // 发送 tool-call 事件 + controller.enqueue({ + type: 'tool-call', + toolCallId: toolUse.id, + toolName: toolUse.toolName, + input: tool.inputSchema + }) + + const result = await tool.execute(toolUse.arguments, { + toolCallId: toolUse.id, + messages: [], + abortSignal: new AbortController().signal + }) + + // 发送 tool-result 事件 + controller.enqueue({ + type: 'tool-result', + toolCallId: toolUse.id, + toolName: toolUse.toolName, + input: toolUse.arguments, + output: result + }) + + executedResults.push({ + toolCallId: toolUse.id, + toolName: toolUse.toolName, + result, + isError: false + }) + } catch (error) { + console.error(`[MCP Prompt Stream] Tool execution failed: ${toolUse.toolName}`, error) + + // 处理错误情况 + const errorResult = this.handleToolError(toolUse, error, controller) + executedResults.push(errorResult) + } + } + + return executedResults + } + + /** + * 格式化工具结果为 Cherry Studio 标准格式 + */ + formatToolResults(executedResults: ExecutedResult[]): string { + return executedResults + .map((tr) => { + if (!tr.isError) { + return `\n ${tr.toolName}\n ${JSON.stringify(tr.result)}\n` + } else { + const error = tr.result || 'Unknown error' + return `\n ${tr.toolName}\n ${error}\n` + } + }) + .join('\n\n') + } + + /** + * 发送工具调用开始相关事件 + */ + private sendToolStartEvents(controller: StreamController, toolUse: ToolUseResult): void { + // 发送 tool-input-start 事件 + controller.enqueue({ + type: 'tool-input-start', + id: toolUse.id, + toolName: toolUse.toolName + }) + } + + /** + * 处理工具执行错误 + */ + private handleToolError( + toolUse: ToolUseResult, + error: unknown, + controller: StreamController + // _tools: ToolSet + ): ExecutedResult { + // 使用 AI SDK 标准错误格式 + // const toolError: TypedToolError = { + // type: 'tool-error', + // toolCallId: toolUse.id, + // toolName: toolUse.toolName, + // input: toolUse.arguments, + // error: error instanceof Error ? error.message : String(error) + // } + + // controller.enqueue(toolError) + + // 发送标准错误事件 + controller.enqueue({ + type: 'error', + error: error instanceof Error ? error.message : String(error) + }) + + return { + toolCallId: toolUse.id, + toolName: toolUse.toolName, + result: error instanceof Error ? error.message : String(error), + isError: true + } + } +} diff --git a/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/promptToolUsePlugin.ts b/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/promptToolUsePlugin.ts index 7eb2bb3e09..1d795c94a3 100644 --- a/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/promptToolUsePlugin.ts +++ b/packages/aiCore/src/core/plugins/built-in/toolUsePlugin/promptToolUsePlugin.ts @@ -3,29 +3,14 @@ * 为不支持原生 Function Call 的模型提供 prompt 方式的工具调用 * 内置默认逻辑，支持自定义覆盖 */ -import type { ModelMessage, TextStreamPart, ToolSet, TypedToolError } from 'ai' +import type { TextStreamPart, ToolSet } from 'ai' import { definePlugin } from '../../index' import type { AiRequestContext } from '../../types' +import { StreamEventManager } from './StreamEventManager' +import { ToolExecutor } from './ToolExecutor' import { PromptToolUseConfig, ToolUseResult } from './type' -/** - * 使用 AI SDK 的 Tool 类型，更通用 - */ -// export interface Tool { -// type: 'function' -// function: { -// name: string -// description?: string -// parameters?: { -// type: 'object' -// properties: Record -// required?: string[] -// additionalProperties?: boolean -// } -// } -// } - /** * 默认系统提示符模板（提取自 Cherry Studio） */ @@ -289,34 +274,32 @@ export const createPromptToolUsePlugin = (config: PromptToolUseConfig = {}) => { transformStream: (_: any, context: AiRequestContext) => () => { let textBuffer = '' let stepId = '' - let executedResults: { toolCallId: string; toolName: string; result: any; isError?: boolean }[] = [] + if (!context.mcpTools) { throw new Error('No tools available') } + + // 创建工具执行器和流事件管理器 + const toolExecutor = new ToolExecutor() + const streamEventManager = new StreamEventManager() + type TOOLS = NonNullable return new TransformStream, TextStreamPart>({ async transform( chunk: TextStreamPart, controller: TransformStreamDefaultController> ) { - // console.log('chunk', chunk) // 收集文本内容 if (chunk.type === 'text-delta') { textBuffer += chunk.text || '' stepId = chunk.id || '' - // console.log('textBuffer', textBuffer) controller.enqueue(chunk) return } if (chunk.type === 'text-end' || chunk.type === 'finish-step') { - // console.log('[MCP Prompt Stream] Received step-finish, checking for tool use...') - - // 从 context 获取工具信息 const tools = context.mcpTools - // console.log('tools from context', tools) if (!tools || Object.keys(tools).length === 0) { - // console.log('[MCP Prompt Stream] No tools available, passing through') controller.enqueue(chunk) return } @@ -324,14 +307,13 @@ export const createPromptToolUsePlugin = (config: PromptToolUseConfig = {}) => { // 解析工具调用 const { results: parsedTools, content: parsedContent } = parseToolUse(textBuffer, tools) const validToolUses = parsedTools.filter((t) => t.status === 'pending') - // console.log('parsedTools', parsedTools) // 如果没有有效的工具调用，直接传递原始事件 if (validToolUses.length === 0) { - // console.log('[MCP Prompt Stream] No valid tool uses found, passing through') controller.enqueue(chunk) return } + if (chunk.type === 'text-end') { controller.enqueue({ type: 'text-end', @@ -349,195 +331,31 @@ export const createPromptToolUsePlugin = (config: PromptToolUseConfig = {}) => { ...chunk, finishReason: 'tool-calls' }) - // console.log('[MCP Prompt Stream] Found valid tool uses:', validToolUses.length) - // 发送 step-start 事件（工具调用步骤开始） - controller.enqueue({ - type: 'start-step', - request: {}, - warnings: [] - }) + // 发送步骤开始事件 + streamEventManager.sendStepStartEvent(controller) // 执行工具调用 - executedResults = [] - for (const toolUse of validToolUses) { - try { - const tool = tools[toolUse.toolName] - if (!tool || typeof tool.execute !== 'function') { - throw new Error(`Tool "${toolUse.toolName}" has no execute method`) - } - // 发送 tool-input-start 事件 - controller.enqueue({ - type: 'tool-input-start', - id: toolUse.id, - toolName: toolUse.toolName - }) + const executedResults = await toolExecutor.executeTools(validToolUses, tools, controller) - console.log(`[MCP Prompt Stream] Executing tool: ${toolUse.toolName}`, toolUse.arguments) - console.log('toolUse,toolUse', toolUse) - // 发送 tool-call 事件 - controller.enqueue({ - type: 'tool-call', - toolCallId: toolUse.id, - toolName: toolUse.toolName, - input: tool.inputSchema - }) + // 发送步骤完成事件 + streamEventManager.sendStepFinishEvent(controller, chunk) - const result = await tool.execute(toolUse.arguments, { - toolCallId: toolUse.id, - messages: [], - abortSignal: new AbortController().signal - }) - - // 发送 tool-result 事件 - controller.enqueue({ - type: 'tool-result', - toolCallId: toolUse.id, - toolName: toolUse.toolName, - input: toolUse.arguments, - output: result - }) - - executedResults.push({ - toolCallId: toolUse.id, - toolName: toolUse.toolName, - result, - isError: false - }) - } catch (error) { - console.error(`[MCP Prompt Stream] Tool execution failed: ${toolUse.toolName}`, error) - - // 使用 AI SDK 标准错误格式 - const toolError: TypedToolError = { - type: 'tool-error', - toolCallId: toolUse.id, - toolName: toolUse.toolName, - input: toolUse.arguments, - error: error instanceof Error ? error.message : String(error) - } - - controller.enqueue(toolError) - - // 发送标准错误事件 - controller.enqueue({ - type: 'error', - error: toolError.error - }) - - // // 发送 tool-result 错误事件 - // controller.enqueue({ - // type: 'tool-result', - // toolCallId: toolUse.id, - // toolName: toolUse.toolName, - // args: toolUse.arguments, - // isError: true, - // result: toolError.message - // }) - - executedResults.push({ - toolCallId: toolUse.id, - toolName: toolUse.toolName, - result: toolError.error, - isError: true - }) - } - } - - // 发送最终的 step-finish 事件 - controller.enqueue({ - type: 'finish-step', - finishReason: 'stop', - response: chunk.response, - usage: chunk.usage, - providerMetadata: chunk.providerMetadata - }) - - // 递归调用逻辑 + // 处理递归调用 if (validToolUses.length > 0) { - // console.log('[MCP Prompt] Starting recursive call after tool execution...') + const toolResultsText = toolExecutor.formatToolResults(executedResults) + const recursiveParams = streamEventManager.buildRecursiveParams( + context, + textBuffer, + toolResultsText, + tools + ) - // 构建工具结果的文本表示，使用Cherry Studio标准格式 - const toolResultsText = executedResults - .map((tr) => { - if (!tr.isError) { - return `\n ${tr.toolName}\n ${JSON.stringify(tr.result)}\n` - } else { - const error = tr.result || 'Unknown error' - return `\n ${tr.toolName}\n ${error}\n` - } - }) - .join('\n\n') - // console.log('context.originalParams.messages', context.originalParams.messages) - // 构建新的对话消息 - const newMessages: ModelMessage[] = [ - ...(context.originalParams.messages || []), - { - role: 'assistant', - content: textBuffer - }, - { - role: 'user', - content: toolResultsText - } - ] - - // 递归调用，继续对话，重新传递 tools - const recursiveParams = { - ...context.originalParams, - messages: newMessages, - tools: tools - } - context.originalParams.messages = newMessages - - try { - const recursiveResult = await context.recursiveCall(recursiveParams) - - // 将递归调用的结果流接入当前流 - if (recursiveResult && recursiveResult.fullStream) { - const reader = recursiveResult.fullStream.getReader() - try { - while (true) { - const { done, value } = await reader.read() - if (done) { - break - } - if (value.type === 'finish') { - // 迭代的流不发finish - break - } - // 将递归流的数据传递到当前流 - controller.enqueue(value) - } - } finally { - reader.releaseLock() - } - } else { - console.warn('[MCP Prompt] No fullstream found in recursive result:', recursiveResult) - } - } catch (error) { - console.error('[MCP Prompt] Recursive call failed:', error) - - // 使用 AI SDK 标准错误格式，但不中断流 - controller.enqueue({ - type: 'error', - error: { - message: error instanceof Error ? error.message : String(error), - name: error instanceof Error ? error.name : 'RecursiveCallError' - } - }) - - // 继续发送文本增量，保持流的连续性 - controller.enqueue({ - type: 'text-delta', - id: stepId, - text: '\n\n[工具执行后递归调用失败，继续对话...]' - }) - } + await streamEventManager.handleRecursiveCall(controller, recursiveParams, context, stepId) } // 清理状态 textBuffer = '' - executedResults = [] return } diff --git a/src/renderer/src/aiCore/chunk/AiSdkToChunkAdapter.ts b/src/renderer/src/aiCore/chunk/AiSdkToChunkAdapter.ts index d30d6a9191..c34a95851d 100644 --- a/src/renderer/src/aiCore/chunk/AiSdkToChunkAdapter.ts +++ b/src/renderer/src/aiCore/chunk/AiSdkToChunkAdapter.ts @@ -83,7 +83,6 @@ export class AiSdkToChunkAdapter { chunk: TextStreamPart, final: { text: string; reasoningContent: string; webSearchResults: any[]; reasoningId: string } ) { - console.log('AI SDK chunk type:', chunk.type, chunk) switch (chunk.type) { // === 文本相关事件 === case 'text-start': @@ -101,7 +100,7 @@ export class AiSdkToChunkAdapter { case 'text-end': this.onChunk({ type: ChunkType.TEXT_COMPLETE, - text: (chunk.providerMetadata?.text?.value as string) || final.text || '' + text: (chunk.providerMetadata?.text?.value as string) ?? final.text ?? '' }) final.text = '' break diff --git a/src/renderer/src/pages/home/Messages/MessageContent.tsx b/src/renderer/src/pages/home/Messages/MessageContent.tsx index 420cd7d466..61b113a011 100644 --- a/src/renderer/src/pages/home/Messages/MessageContent.tsx +++ b/src/renderer/src/pages/home/Messages/MessageContent.tsx @@ -15,9 +15,7 @@ const MessageContent: React.FC = ({ message }) => { <> {!isEmpty(message.mentions) && ( - {message.mentions?.map((model) => ( - {'@' + model.name} - ))} + {message.mentions?.map((model) => {'@' + model.name})} )}