A Simple and Universal Swarm Intelligence Engine, Predicting Anything. 简洁通用的群体智能引擎,预测万物
Python agent-memoryfinancial-forecastingfuture-predictionknowledge-graphllmsmulti-agent-simulationpublic-opinion-analysispython3social-predictionswarm-intelligence
feat(report_agent): enhance interview text processing and response handling; improve quote extraction and formatting for better clarity
666ghj7601d78fd4d9
diff --git a/backend/app/services/zep_tools.py b/backend/app/services/zep_tools.pyindex 4bd28966..1694632a 100644--- a/backend/app/services/zep_tools.py+++ b/backend/app/services/zep_tools.py@@ -308,7 +308,30 @@ class AgentInterview:if self.key_quotes:text += "\n**关键引言:**\n"for quote in self.key_quotes:- text += f"> \"{quote}\"\n"+ # 清理各种引号+ clean_quote = quote.replace('\u201c', '').replace('\u201d', '').replace('"', '')+ clean_quote = clean_quote.replace('\u300c', '').replace('\u300d', '')+ clean_quote = clean_quote.strip()+ # 去掉开头的标点+ while clean_quote and clean_quote[0] in ',,;;::、。!?\n\r\t ':+ clean_quote = clean_quote[1:]+ # 过滤包含问题编号的垃圾内容(问题1-9)+ skip = False+ for d in '123456789':+ if f'\u95ee\u9898{d}' in clean_quote:+ skip = True+ break+ if skip:+ continue+ # 截断过长内容(按句号截断,而非硬截断)+ if len(clean_quote) > 150:+ dot_pos = clean_quote.find('\u3002', 80)+ if dot_pos > 0:+ clean_quote = clean_quote[:dot_pos + 1]+ else:+ clean_quote = clean_quote[:147] + "..."+ if clean_quote and len(clean_quote) >= 10:+ text += f'> "{clean_quote}"\n'return text@@ -350,27 +373,26 @@ class InterviewResult:def to_text(self) -> str:"""转换为详细的文本格式,供LLM理解和报告引用"""text_parts = [- f"## 🎤 深度采访报告",+ "## 深度采访报告",f"**采访主题:** {self.interview_topic}",f"**采访人数:** {self.interviewed_count} / {self.total_agents} 位模拟Agent",- f"\n### 采访对象选择理由",- f"{self.selection_reasoning}",- f"\n---"+ "\n### 采访对象选择理由",+ self.selection_reasoning or "(自动选择)",+ "\n---",+ "\n### 采访实录",]-- # 各Agent的采访内容+if self.interviews:- text_parts.append(f"\n### 采访实录")for i, interview in enumerate(self.interviews, 1):text_parts.append(f"\n#### 采访 #{i}: {interview.agent_name}")text_parts.append(interview.to_text())text_parts.append("\n---")-- # 采访摘要- if self.summary:- text_parts.append(f"\n### 采访摘要与核心观点")- text_parts.append(self.summary)-+ else:+ text_parts.append("(无采访记录)\n\n---")++ text_parts.append("\n### 采访摘要与核心观点")+ text_parts.append(self.summary or "(无摘要)")+return "\n".join(text_parts)@@ -1329,8 +1351,18 @@ class ZepToolsService:# 将问题合并为一个采访promptcombined_prompt = "\n".join([f"{i+1}. {q}" for i, q in enumerate(result.interview_questions)])- # 添加优化前缀,避免Agent调用工具而直接回复文本- INTERVIEW_PROMPT_PREFIX = "结合你的人设、所有的过往记忆与行动,不调用任何工具直接用文本回复我:"+ # 添加优化前缀,约束Agent回复格式+ INTERVIEW_PROMPT_PREFIX = (+ "你正在接受一次采访。请结合你的人设、所有的过往记忆与行动,"+ "以纯文本方式直接回答以下问题。\n"+ "回复要求:\n"+ "1. 直接用自然语言回答,不要调用任何工具\n"+ "2. 不要返回JSON格式或工具调用格式\n"+ "3. 不要使用Markdown标题(如#、##、###)\n"+ "4. 按问题编号逐一回答,每个回答以「问题X:」开头(X为问题编号)\n"+ "5. 每个问题的回答之间用空行分隔\n"+ "6. 回答要有实质内容,每个问题至少回答2-3句话\n\n"+ )optimized_prompt = f"{INTERVIEW_PROMPT_PREFIX}{combined_prompt}"# Step 4: 调用真实的采访API(不指定platform,默认双平台同时采访)@@ -1380,26 +1412,43 @@ class ZepToolsService:twitter_response = twitter_result.get("response", "")reddit_response = reddit_result.get("response", "")-- # 合并两个平台的回答- response_parts = []- if twitter_response:- response_parts.append(f"【Twitter平台回答】\n{twitter_response}")- if reddit_response:- response_parts.append(f"【Reddit平台回答】\n{reddit_response}")-- if response_parts:- response_text = "\n\n".join(response_parts)- else:- response_text = "[无回复]"-++ # 清理可能的工具调用 JSON 包裹+ twitter_response = self._clean_tool_call_response(twitter_response)+ reddit_response = self._clean_tool_call_response(reddit_response)++ # 始终输出双平台标记+ twitter_text = twitter_response if twitter_response else "(该平台未获得回复)"+ reddit_text = reddit_response if reddit_response else "(该平台未获得回复)"+ response_text = f"【Twitter平台回答】\n{twitter_text}\n\n【Reddit平台回答】\n{reddit_text}"+# 提取关键引言(从两个平台的回答中)import recombined_responses = f"{twitter_response} {reddit_response}"- key_quotes = re.findall(r'[""「」『』]([^""「」『』]{10,100})[""「」『』]', combined_responses)++ # 清理响应文本:去掉标记、编号、Markdown 等干扰+ clean_text = re.sub(r'#{1,6}\s+', '', combined_responses)+ clean_text = re.sub(r'\{[^}]*tool_name[^}]*\}', '', clean_text)+ clean_text = re.sub(r'[*_`|>~\-]{2,}', '', clean_text)+ clean_text = re.sub(r'问题\d+[::]\s*', '', clean_text)+ clean_text = re.sub(r'【[^】]+】', '', clean_text)++ # 策略1(主): 提取完整的有实质内容的句子+ sentences = re.split(r'[。!?]', clean_text)+ meaningful = [+ s.strip() for s in sentences+ if 20 <= len(s.strip()) <= 150+ and not re.match(r'^[\s\W,,;;::、]+', s.strip())+ and not s.strip().startswith(('{', '问题'))+ ]+ meaningful.sort(key=len, reverse=True)+ key_quotes = [s + "。" for s in meaningful[:3]]++ # 策略2(补充): 正确配对的中文引号「」内长文本if not key_quotes:- sentences = combined_responses.split('。')- key_quotes = [s.strip() + '。' for s in sentences if len(s.strip()) > 20][:3]+ paired = re.findall(r'\u201c([^\u201c\u201d]{15,100})\u201d', clean_text)+ paired += re.findall(r'\u300c([^\u300c\u300d]{15,100})\u300d', clean_text)+ key_quotes = [q for q in paired if not re.match(r'^[,,;;::、]', q)][:3]interview = AgentInterview(agent_name=agent_name,@@ -1435,6 +1484,27 @@ class ZepToolsService:logger.info(f"InterviewAgents完成: 采访了 {result.interviewed_count} 个Agent(双平台)")return result+ @staticmethod+ def _clean_tool_call_response(response: str) -> str:+ """清理 Agent 回复中的 JSON 工具调用包裹,提取实际内容"""+ if not response or not response.strip().startswith('{'):+ return response+ text = response.strip()+ if 'tool_name' not in text[:80]:+ return response+ import re as _re+ try:+ data = json.loads(text)+ if isinstance(data, dict) and 'arguments' in data:+ for key in ('content', 'text', 'body', 'message', 'reply'):+ if key in data['arguments']:+ return str(data['arguments'][key])+ except (json.JSONDecodeError, KeyError, TypeError):+ match = _re.search(r'"content"\s*:\s*"((?:[^"\\]|\\.)*)"', text)+ if match:+ return match.group(1).replace('\\n', '\n').replace('\\"', '"')+ return response+def _load_agent_profiles(self, simulation_id: str) -> List[Dict[str, Any]]:"""加载模拟的Agent人设文件"""import os@@ -1581,6 +1651,8 @@ class ZepToolsService:2. 针对不同角色可能有不同答案3. 涵盖事实、观点、感受等多个维度4. 语言自然,像真实采访一样+5. 每个问题控制在50字以内,简洁明了+6. 直接提问,不要包含背景说明或前缀返回JSON格式:{"questions": ["问题1", "问题2", ...]}"""@@ -1633,7 +1705,14 @@ class ZepToolsService:2. 指出观点的共识和分歧3. 突出有价值的引言4. 客观中立,不偏袒任何一方-5. 控制在1000字内"""+5. 控制在1000字内++格式约束(必须遵守):+- 使用纯文本段落,用空行分隔不同部分+- 不要使用Markdown标题(如#、##、###)+- 不要使用分割线(如---、***)+- 引用受访者原话时使用中文引号「」+- 可以使用**加粗**标记关键词,但不要使用其他Markdown语法"""user_prompt = f"""采访主题:{interview_requirement}diff --git a/frontend/src/components/Step4Report.vue b/frontend/src/components/Step4Report.vueindex 28d3cf11..f44aedc4 100644--- a/frontend/src/components/Step4Report.vue+++ b/frontend/src/components/Step4Report.vue@@ -849,27 +849,36 @@ const parseInterview = (text) => {interview.redditAnswer = redditMatch[1].trim()}- // 如果只有一个平台的回答,将其作为主回答- // 这样无论显示哪个平台都能有内容+ // 平台回退逻辑(兼容旧格式:只有一个平台标记的情况)if (!twitterMatch && redditMatch) {- // 只有 Reddit 回答,将其也设为 twitterAnswer 作为默认显示- interview.twitterAnswer = interview.redditAnswer+ // 只有 Reddit 回答,仅在非占位文本时复制为默认显示+ if (interview.redditAnswer && interview.redditAnswer !== '(该平台未获得回复)') {+ interview.twitterAnswer = interview.redditAnswer+ }} else if (twitterMatch && !redditMatch) {- // 只有 Twitter 回答,将其也设为 redditAnswer- interview.redditAnswer = interview.twitterAnswer+ if (interview.twitterAnswer && interview.twitterAnswer !== '(该平台未获得回复)') {+ interview.redditAnswer = interview.twitterAnswer+ }} else if (!twitterMatch && !redditMatch) {- // 如果没有明确分平台,整体作为回答+ // 没有分平台标记(极旧格式),整体作为回答interview.twitterAnswer = answerText}}- // 提取关键引言+ // 提取关键引言(兼容多种引号格式)const quotesMatch = block.match(/\*\*关键引言:\*\*\n([\s\S]*?)(?=\n---|\n####|$)/)if (quotesMatch) {const quotesText = quotesMatch[1]- const quoteMatches = quotesText.match(/> "([^"]+)"/g)+ // 优先匹配 > "text" 格式+ let quoteMatches = quotesText.match(/> "([^"]+)"/g)+ // 回退:匹配 > "text" 或 > \u201Ctext\u201D(中文引号)+ if (!quoteMatches) {+ quoteMatches = quotesText.match(/> [\u201C""]([^\u201D""]+)[\u201D""]/g)+ }if (quoteMatches) {- interview.quotes = quoteMatches.map(q => q.replace(/^> "|"$/g, '').trim())+ interview.quotes = quoteMatches+ .map(q => q.replace(/^> [\u201C""]|[\u201D""]$/g, '').trim())+ .filter(q => q)}}@@ -1314,79 +1323,100 @@ const InterviewDisplay = {return text.substring(0, 400) + '...'}+ // 检查是否为平台占位文本+ const isPlaceholderText = (text) => {+ if (!text) return true+ const t = text.trim()+ return t === '(该平台未获得回复)' || t === '(该平台未获得回复)' || t === '[无回复]'+ }+// 尝试按问题编号分割回答const splitAnswerByQuestions = (answerText, questionCount) => {if (!answerText || questionCount <= 0) return [answerText]-- // 更健壮的分割逻辑:查找所有 "数字." 格式的编号位置- // 支持格式:- // - "1. \n内容" (数字+点+空格+换行+内容)- // - "\n\n2. \n内容" (换行+数字+点+空格+换行+内容)- // 使用更宽松的匹配:开头或换行后的数字+点+空白- const numberPattern = /(?:^|[\r\n]+)(\d+)\.\s+/g- const matches = []+ if (isPlaceholderText(answerText)) return ['']++ // 支持两种编号格式:+ // 1. "问题X:" 或 "问题X:" (中文格式,后端新格式)+ // 2. "1. " 或 "\n1. " (数字+点,旧格式兼容)+ let matches = []let match-- while ((match = numberPattern.exec(answerText)) !== null) {++ // 优先尝试 "问题X:" 格式+ const cnPattern = /(?:^|[\r\n]+)问题(\d+)[::]\s*/g+ while ((match = cnPattern.exec(answerText)) !== null) {matches.push({num: parseInt(match[1]),index: match.index,fullMatch: match[0]})}-++ // 如果没匹配到,回退到 "数字." 格式+ if (matches.length === 0) {+ const numPattern = /(?:^|[\r\n]+)(\d+)\.\s+/g+ while ((match = numPattern.exec(answerText)) !== null) {+ matches.push({+ num: parseInt(match[1]),+ index: match.index,+ fullMatch: match[0]+ })+ }+ }+// 如果没有找到编号或只找到一个,返回整体if (matches.length <= 1) {- // 尝试移除开头的编号(格式:1. \n 或 1. )- const cleaned = answerText.replace(/^\d+\.\s+/, '').trim()+ const cleaned = answerText+ .replace(/^问题\d+[::]\s*/, '')+ .replace(/^\d+\.\s+/, '')+ .trim()return [cleaned || answerText]}-+// 按编号提取各部分const parts = []for (let i = 0; i < matches.length; i++) {const current = matches[i]const next = matches[i + 1]-+const startIdx = current.index + current.fullMatch.lengthconst endIdx = next ? next.index : answerText.length-+let part = answerText.substring(startIdx, endIdx).trim()- // 移除末尾可能的多余换行part = part.replace(/[\r\n]+$/, '').trim()parts.push(part)}-- // 如果分割成功且数量合理,返回分割结果+if (parts.length > 0 && parts.some(p => p)) {return parts}-+return [answerText]}// 获取某个问题对应的回答const getAnswerForQuestion = (interview, qIdx, platform) => {const answer = platform === 'twitter' ? interview.twitterAnswer : (interview.redditAnswer || interview.twitterAnswer)- if (!answer) return ''-+ if (!answer || isPlaceholderText(answer)) return answer || ''+const questionCount = interview.questions?.length || 1const answers = splitAnswerByQuestions(answer, questionCount)-- // 如果只有一个回答部分,或者索引超出,返回完整回答- if (answers.length === 1 || qIdx >= answers.length) {- return qIdx === 0 ? answer : ''++ // 分割成功且索引有效+ if (answers.length > 1 && qIdx < answers.length) {+ return answers[qIdx] || ''}-- return answers[qIdx] || ''++ // 分割失败:第一个问题返回完整回答,其余返回空+ return qIdx === 0 ? answer : ''}- // 检查某个问题是否有双平台回答+ // 检查某个问题是否有双平台回答(过滤占位文本)const hasMultiplePlatforms = (interview, qIdx) => {if (!interview.twitterAnswer || !interview.redditAnswer) return falseconst twitterAnswer = getAnswerForQuestion(interview, qIdx, 'twitter')const redditAnswer = getAnswerForQuestion(interview, qIdx, 'reddit')- return twitterAnswer && redditAnswer && twitterAnswer !== redditAnswer+ // 两个平台都有真实回答(非占位文本)且内容不同+ return !isPlaceholderText(twitterAnswer) && !isPlaceholderText(redditAnswer) && twitterAnswer !== redditAnswer}return () => h('div', { class: 'interview-display' }, [@@ -1453,7 +1483,8 @@ const InterviewDisplay = {const hasDualPlatform = hasMultiplePlatforms(interview, qIdx)const expandKey = `${activeIndex.value}-${qIdx}`const isExpanded = expandedAnswers.value.has(expandKey)-+ const isPlaceholder = isPlaceholderText(answerText)+return h('div', { class: 'qa-pair', key: qIdx }, [// Question Blockh('div', { class: 'qa-question' }, [@@ -1463,14 +1494,14 @@ const InterviewDisplay = {h('div', { class: 'qa-text' }, question)])]),-+// Answer Block- answerText && h('div', { class: 'qa-answer' }, [+ answerText && h('div', { class: ['qa-answer', { 'answer-placeholder': isPlaceholder }] }, [h('div', { class: 'qa-badge a-badge' }, `A${qIdx + 1}`),h('div', { class: 'qa-content' }, [h('div', { class: 'qa-answer-header' }, [h('div', { class: 'qa-sender' }, interview?.name || 'Agent'),- // 双平台切换按钮+ // 双平台切换按钮(仅在有真实双平台回答时显示)hasDualPlatform && h('div', { class: 'platform-switch' }, [h('button', {class: ['platform-btn', { active: currentPlatform === 'twitter' }],@@ -1494,14 +1525,16 @@ const InterviewDisplay = {])])]),- h('div', {- class: 'qa-text answer-text',- innerHTML: formatAnswer(answerText, isExpanded)- .replace(/\*\*(.+?)\*\*/g, '<strong>$1</strong>')- .replace(/\n/g, '<br>')+ h('div', {+ class: ['qa-text', 'answer-text', { 'placeholder-text': isPlaceholder }],+ innerHTML: isPlaceholder+ ? answerText+ : formatAnswer(answerText, isExpanded)+ .replace(/\*\*(.+?)\*\*/g, '<strong>$1</strong>')+ .replace(/\n/g, '<br>')}),- // Expand/Collapse Button- answerText.length > 400 && h('button', {+ // Expand/Collapse Button(占位文本不显示)+ !isPlaceholder && answerText.length > 400 && h('button', {class: 'expand-answer-btn',onClick: () => toggleAnswer(expandKey)}, isExpanded ? 'Show Less' : 'Show More')@@ -3913,6 +3946,15 @@ watch(() => props.reportId, (newId) => {margin-top: 0;}+:deep(.interview-display .answer-placeholder) {+ opacity: 0.6;+}++:deep(.interview-display .placeholder-text) {+ font-style: italic;+ color: #9CA3AF;+}+:deep(.interview-display .qa-answer-header) {display: flex;justify-content: space-between;