@yeaft/webchat-agent 0.1.84 → 0.1.86

package/conversation.js

@@ -607,6 +607,53 @@ export async function handleUserInput(msg) {
     console.log(`[RolePlay] Human responded, resuming from ${fromRole}'s request`);
   }
 
+  // ★ RolePlay: handle @mention targetRole routing
+  const targetRole = msg.targetRole;
+  if (targetRole && rpSession && rpSession._routeInitialized) {
+    const roleNames = new Set(rpSession.roles.map(r => r.name));
+    if (roleNames.has(targetRole)) {
+      const targetRoleConfig = rpSession.roles.find(r => r.name === targetRole);
+      const targetLabel = targetRoleConfig
+        ? (targetRoleConfig.icon ? `${targetRoleConfig.icon} ${targetRoleConfig.displayName}` : targetRoleConfig.displayName)
+        : targetRole;
+      const targetClaudeMd = targetRoleConfig?.claudeMd || '';
+
+      // Prepend ROLE signal so Claude responds as the target role
+      let rolePrefix = `---ROLE: ${targetRole}---\n\n`;
+      rolePrefix += `用户指定由 ${targetLabel} 来回复。\n`;
+      if (targetClaudeMd) {
+        rolePrefix += `<role-context>\n${targetClaudeMd}\n</role-context>\n\n`;
+      }
+      rolePrefix += `请以 ${targetLabel} 的身份回复以下消息：\n\n`;
+      effectivePrompt = rolePrefix + effectivePrompt;
+
+      // Update rpSession state
+      const prevRole = rpSession.currentRole;
+      rpSession.currentRole = targetRole;
+      if (rpSession.roleStates[targetRole]) {
+        rpSession.roleStates[targetRole].status = 'active';
+      }
+      if (prevRole && prevRole !== targetRole && rpSession.roleStates[prevRole]) {
+        rpSession.roleStates[prevRole].status = 'idle';
+      }
+
+      // Send roleplay_status update to frontend
+      ctx.sendToServer({
+        type: 'roleplay_status',
+        conversationId,
+        currentRole: rpSession.currentRole,
+        round: rpSession.round,
+        features: rpSession.features ? Array.from(rpSession.features.values()) : [],
+        roleStates: rpSession.roleStates || {},
+        waitingHuman: false
+      });
+
+      console.log(`[RolePlay] @mention routing: ${prevRole || 'none'} -> ${targetRole}`);
+    } else {
+      console.warn(`[RolePlay] @mention target role not found: ${targetRole}`);
+    }
+  }
+
   const userMessage = {
     type: 'user',
     message: { role: 'user', content: effectivePrompt }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yeaft/webchat-agent",
-  "version": "0.1.84",
+  "version": "0.1.86",
   "description": "Remote agent for Yeaft WebChat — connects worker machines to the central server",
   "main": "index.js",
   "type": "module",

package/roleplay-dir.js

@@ -24,9 +24,9 @@ import { getRolePlayMessages } from './roleplay-i18n.js';
 // Each entry is a list of role names (keys into roleTemplates in i18n).
 const TEAM_ROLES = {
   dev:     ['pm', 'dev', 'reviewer', 'tester'],
-  writing: ['planner', 'writer', 'editor'],
-  trading: ['strategist', 'analyst', 'macro', 'risk', 'trader'],
-  video:   ['director', 'scriptwriter', 'storyboard'],
+  writing: ['editor', 'writer', 'proofreader'],
+  trading: ['analyst', 'strategist', 'risk-manager'],
+  video:   ['director', 'writer', 'producer'],
   custom:  ['pm', 'dev', 'reviewer', 'tester'], // default same as dev
 };
 
@@ -193,9 +193,13 @@ export async function writeSessionClaudeMd(projectDir, sessionName, config) {
   const roleSection = buildRoleSection(teamType, language, customRoles);
 
   // Build workflow section
-  const workflow = (teamType === 'dev')
-    ? m.devWorkflow
-    : m.genericWorkflow;
+  const workflowMap = {
+    dev: m.devWorkflow,
+    writing: m.writingWorkflow,
+    trading: m.tradingWorkflow,
+    video: m.videoWorkflow,
+  };
+  const workflow = workflowMap[teamType] || m.genericWorkflow;
 
   const content = `${m.sessionTitle(sessionName)}
 
@@ -278,7 +282,7 @@ export function getDefaultRoles(teamType, language = 'zh-CN') {
     if (!tmpl) return { name, displayName: name, icon: '' };
 
     // Extract icon and displayName from heading: "## 📋 PM-乔布斯 (pm)"
-    const headingMatch = tmpl.heading.match(/^##\s*(\S+)\s+(.+?)\s*\((\w+)\)\s*$/);
+    const headingMatch = tmpl.heading.match(/^##\s*(\S+)\s+(.+?)\s*\(([\w-]+)\)\s*$/);
     if (headingMatch) {
       return {
         name: headingMatch[3],

package/roleplay-i18n.js

@@ -189,6 +189,37 @@ const messages = {
 
 风格：视觉思维，善于用画面讲故事。`,
       },
+      // RolePlay-specific roles (not in Crew templates)
+      proofreader: {
+        heading: '## 🔎 审校-马伯庸 (proofreader)',
+        content: `你是审校-马伯庸。你的职责：
+- 检查内容的逻辑一致性和事实准确性
+- 审核文字质量、错别字和表达规范
+- 核实引用和数据的准确性
+- 提出具体的修改建议
+
+风格：考据成瘾，逻辑洁癖，毒舌但建设性，指出问题必给修改方案。`,
+      },
+      'risk-manager': {
+        heading: '## 🛡️ 风控官-塔勒布 (risk-manager)',
+        content: `你是风控官-塔勒布。你的职责：
+- 对策略进行压力测试和尾部风险评估
+- 检查仓位是否符合风控原则
+- 审核止损设置和对冲方案
+- 如果策略风险不可接受，直接打回并说明原因
+
+风格：尾部风险偏执狂，反脆弱思维，杠铃策略信徒。`,
+      },
+      producer: {
+        heading: '## 🎬 制片-徐克 (producer)',
+        content: `你是制片-徐克。你的职责：
+- 审核脚本和分镜的可执行性
+- 评估制作资源需求和技术可行性
+- 把控制作进度和质量标准
+- 生成最终的 AI 视频 prompt 序列
+
+风格：视觉想象力爆棚，技术与艺术兼备。`,
+      },
     },
 
     routeProtocol: `# ROUTE 协议
@@ -214,6 +245,18 @@ taskTitle: {任务标题}（可选）
 3. **审查者** Code Review（不通过 → 返回开发者修复）
 4. **测试者** 运行测试 & 验证（有 bug → 返回开发者修复）
 5. **PM** 验收总结`,
+    writingWorkflow: `1. **编辑** 分析需求，确定内容方向和框架
+2. **作者** 根据大纲撰写内容
+3. **审校** 检查逻辑一致性、事实准确性和文字质量（不通过 → 返回作者修改）
+4. **编辑** 验收最终成果`,
+    tradingWorkflow: `1. **分析师** 研究市场，输出技术分析和关键价位
+2. **策略师** 综合分析，制定投资策略和仓位方案
+3. **风控官** 压力测试策略，评估尾部风险（不通过 → 返回策略师调整）
+4. **策略师** 确认最终方案并总结`,
+    videoWorkflow: `1. **导演** 确定主题、情绪基调和视觉风格
+2. **编剧** 构思故事线，撰写分段脚本
+3. **制片** 审核可行性，生成最终 prompt 序列（不通过 → 返回编剧调整）
+4. **导演** 最终审核并验收`,
     genericWorkflow: '按角色顺序依次完成任务。',
 
     projectPathTitle: '# 项目路径',
@@ -399,6 +442,37 @@ Style: Good storytelling, focused on emotional resonance.`,
 
 Style: Visual thinking, good at telling stories with images.`,
       },
+      // RolePlay-specific roles (not in Crew templates)
+      proofreader: {
+        heading: '## 🔎 Proofreader-Tolkien (proofreader)',
+        content: `You are Proofreader-Tolkien. Your responsibilities:
+- Check content for logical consistency and factual accuracy
+- Review writing quality, typos, and expression standards
+- Verify accuracy of citations and data
+- Provide specific revision suggestions
+
+Style: Research addict, logic purist, sharp but constructive — every critique comes with a fix.`,
+      },
+      'risk-manager': {
+        heading: '## 🛡️ Risk-Officer-Taleb (risk-manager)',
+        content: `You are Risk-Officer-Taleb. Your responsibilities:
+- Stress-test strategies and assess tail risks
+- Verify positions comply with risk principles
+- Review stop-loss settings and hedging plans
+- Reject strategies with unacceptable risk — explain which principle is violated
+
+Style: Tail risk obsessive, antifragile thinking, barbell strategy devotee.`,
+      },
+      producer: {
+        heading: '## 🎬 Producer-Spielberg (producer)',
+        content: `You are Producer-Spielberg. Your responsibilities:
+- Review script and storyboard feasibility
+- Assess production resource needs and technical viability
+- Control production schedule and quality standards
+- Generate final AI video prompt sequences
+
+Style: Visual imagination overflows, art and craft in equal measure.`,
+      },
     },
 
     routeProtocol: `# ROUTE Protocol
@@ -424,6 +498,18 @@ Rules:
 3. **Reviewer** code review (if fails → back to Dev)
 4. **Tester** runs tests & verifies (if bugs → back to Dev)
 5. **PM** acceptance & summary`,
+    writingWorkflow: `1. **Editor** analyzes requirements, determines content direction and framework
+2. **Writer** writes content based on outline
+3. **Proofreader** checks logical consistency, factual accuracy, and writing quality (if fails → back to Writer)
+4. **Editor** final acceptance of deliverables`,
+    tradingWorkflow: `1. **Analyst** researches market, outputs technical analysis and key levels
+2. **Strategist** synthesizes analysis, formulates investment strategy and position plan
+3. **Risk Manager** stress-tests strategy, assesses tail risks (if fails → back to Strategist)
+4. **Strategist** confirms final plan and summarizes`,
+    videoWorkflow: `1. **Director** establishes theme, emotional tone, and visual style
+2. **Screenwriter** conceives storyline, writes segmented script
+3. **Producer** reviews feasibility, generates final prompt sequence (if fails → back to Screenwriter)
+4. **Director** final review and acceptance`,
     genericWorkflow: 'Complete tasks by following the role sequence.',
 
     projectPathTitle: '# Project Path',

package/roleplay.js

@@ -483,8 +483,17 @@ function getWorkflow(teamType, roles, isZh) {
   if (teamType === 'dev') {
     return buildDevWorkflow(roleNames, isZh);
   }
+  if (teamType === 'writing') {
+    return buildWritingWorkflow(roleNames, isZh);
+  }
+  if (teamType === 'trading') {
+    return buildTradingWorkflow(roleNames, isZh);
+  }
+  if (teamType === 'video') {
+    return buildVideoWorkflow(roleNames, isZh);
+  }
 
-  // Generic fallback for other team types
+  // Generic fallback for custom / unknown team types
   return isZh ? '按角色顺序依次完成任务。' : 'Complete tasks by following the role sequence.';
 }
 
@@ -516,9 +525,71 @@ function buildDevWorkflow(roleNames, isZh) {
   return steps.join('\n');
 }
 
-// ---------------------------------------------------------------------------
-// RolePlay ROUTE protocol support
-// ---------------------------------------------------------------------------
+function buildWritingWorkflow(roleNames, isZh) {
+  const hasEditor = roleNames.includes('editor');
+  const hasWriter = roleNames.includes('writer');
+  const hasProofreader = roleNames.includes('proofreader');
+
+  const steps = [];
+
+  if (isZh) {
+    if (hasEditor) steps.push(`${steps.length + 1}. **编辑** 分析需求，确定内容方向和框架`);
+    if (hasWriter) steps.push(`${steps.length + 1}. **作者** 根据大纲撰写内容`);
+    if (hasProofreader) steps.push(`${steps.length + 1}. **审校** 检查逻辑一致性、事实准确性和文字质量（不通过 → 返回作者修改）`);
+    if (hasEditor) steps.push(`${steps.length + 1}. **编辑** 验收最终成果`);
+  } else {
+    if (hasEditor) steps.push(`${steps.length + 1}. **Editor** analyzes requirements, determines content direction and framework`);
+    if (hasWriter) steps.push(`${steps.length + 1}. **Writer** writes content based on outline`);
+    if (hasProofreader) steps.push(`${steps.length + 1}. **Proofreader** checks logical consistency, factual accuracy, and writing quality (if fails → back to Writer)`);
+    if (hasEditor) steps.push(`${steps.length + 1}. **Editor** final acceptance of deliverables`);
+  }
+
+  return steps.join('\n');
+}
+
+function buildTradingWorkflow(roleNames, isZh) {
+  const hasAnalyst = roleNames.includes('analyst');
+  const hasStrategist = roleNames.includes('strategist');
+  const hasRiskManager = roleNames.includes('risk-manager');
+
+  const steps = [];
+
+  if (isZh) {
+    if (hasAnalyst) steps.push(`${steps.length + 1}. **分析师** 研究市场，输出技术分析和关键价位`);
+    if (hasStrategist) steps.push(`${steps.length + 1}. **策略师** 综合分析，制定投资策略和仓位方案`);
+    if (hasRiskManager) steps.push(`${steps.length + 1}. **风控官** 压力测试策略，评估尾部风险（不通过 → 返回策略师调整）`);
+    if (hasStrategist) steps.push(`${steps.length + 1}. **策略师** 确认最终方案并总结`);
+  } else {
+    if (hasAnalyst) steps.push(`${steps.length + 1}. **Analyst** researches market, outputs technical analysis and key levels`);
+    if (hasStrategist) steps.push(`${steps.length + 1}. **Strategist** synthesizes analysis, formulates investment strategy and position plan`);
+    if (hasRiskManager) steps.push(`${steps.length + 1}. **Risk Manager** stress-tests strategy, assesses tail risks (if fails → back to Strategist)`);
+    if (hasStrategist) steps.push(`${steps.length + 1}. **Strategist** confirms final plan and summarizes`);
+  }
+
+  return steps.join('\n');
+}
+
+function buildVideoWorkflow(roleNames, isZh) {
+  const hasDirector = roleNames.includes('director');
+  const hasWriter = roleNames.includes('writer');
+  const hasProducer = roleNames.includes('producer');
+
+  const steps = [];
+
+  if (isZh) {
+    if (hasDirector) steps.push(`${steps.length + 1}. **导演** 确定主题、情绪基调和视觉风格`);
+    if (hasWriter) steps.push(`${steps.length + 1}. **编剧** 构思故事线，撰写分段脚本`);
+    if (hasProducer) steps.push(`${steps.length + 1}. **制片** 审核可行性，生成最终 prompt 序列（不通过 → 返回编剧调整）`);
+    if (hasDirector) steps.push(`${steps.length + 1}. **导演** 最终审核并验收`);
+  } else {
+    if (hasDirector) steps.push(`${steps.length + 1}. **Director** establishes theme, emotional tone, and visual style`);
+    if (hasWriter) steps.push(`${steps.length + 1}. **Screenwriter** conceives storyline, writes segmented script`);
+    if (hasProducer) steps.push(`${steps.length + 1}. **Producer** reviews feasibility, generates final prompt sequence (if fails → back to Screenwriter)`);
+    if (hasDirector) steps.push(`${steps.length + 1}. **Director** final review and acceptance`);
+  }
+
+  return steps.join('\n');
+}
 
 // Re-export parseRoutes for use by claude.js and tests
 export { parseRoutes } from './crew/routing.js';