diff --git a/backend/app/services/oasis_profile_generator.py b/backend/app/services/oasis_profile_generator.py index d7b1122..3009e10 100644 --- a/backend/app/services/oasis_profile_generator.py +++ b/backend/app/services/oasis_profile_generator.py @@ -684,43 +684,43 @@ class OasisProfileGenerator: ) -> str: """构建个人实体的详细人设提示词""" - attrs_str = json.dumps(entity_attributes, ensure_ascii=False) if entity_attributes else "无" - context_str = context[:3000] if context else "无额外上下文" + attrs_str = json.dumps(entity_attributes, ensure_ascii=False) if entity_attributes else "None" + context_str = context[:3000] if context else "No additional context" return f"""Generate detailed social media user persona for the entity, maximizing fidelity to real-world situations. -实体名称: {entity_name} -实体类型: {entity_type} -实体摘要: {entity_summary} -实体属性: {attrs_str} +Entity name: {entity_name} +Entity type: {entity_type} +Entity summary: {entity_summary} +Entity attributes: {attrs_str} -上下文信息: +Context information: {context_str} -请生成JSON，包含以下字段: +Please generate JSON with the following fields: -1. bio: 社交媒体简介，200字 -2. persona: 详细人设描述（2000字的纯文本），需包含: - - 基本信息（年龄、职业、教育背景、所在地） - - 人物背景（重要经历、与事件的关联、社会关系） - - 性格特征（MBTI类型、核心性格、情绪表达方式） - - 社交媒体行为（发帖频率、内容偏好、互动风格、语言特点） - - 立场观点（对话题的态度、可能被激怒/感动的内容） - - 独特特征（口头禅、特殊经历、个人爱好） - - 个人记忆（人设的重要部分，要介绍这个个体与事件的关联，以及这个个体在事件中的已有动作与反应） -3. age: 年龄数字（必须是整数） -4. gender: 性别，必须是英文: "male" 或 "female" -5. mbti: MBTI类型（如INTJ、ENFP等） -6. country: 国家（使用中文，如"中国"） -7. profession: 职业 -8. interested_topics: 感兴趣话题数组 +1. bio: Social media bio, 200 characters +2. persona: Detailed persona description (2000 chars plain text), must include: + - Basic info (age, occupation, education, location) + - Background (important experiences, connection to event, social relations) + - Personality (MBTI type, core personality, emotional expression style) + - Social media behavior (posting frequency, content preferences, interaction style, language traits) + - Stance (attitude toward topics, content that may trigger/move them) + - Unique traits (catchphrases, special experiences, personal hobbies) + - Personal memory (important part of persona: individual's connection to the event, and their existing actions and reactions) +3. age: Age number (must be integer) +4. gender: Gender, must be English: "male" or "female" +5. mbti: MBTI type (e.g. INTJ, ENFP) +6. country: Country (use English, e.g. "Thailand", "China") +7. profession: Occupation +8. interested_topics: Array of topics of interest -重要: -- 所有字段值必须是字符串或数字，不要使用换行符 -- persona必须是一段连贯的文字描述 +Important: +- All field values must be strings or numbers, do not use newlines +- persona must be a coherent text description - {get_language_instruction()} (gender字段必须用英文male/female) -- 内容要与实体信息保持一致 -- age必须是有效的整数，gender必须是"male"或"female" +- Content must be consistent with entity information +- age must be a valid integer, gender must be "male" or "female" """ def _build_group_persona_prompt( @@ -733,42 +733,42 @@ class OasisProfileGenerator: ) -> str: """构建群体/机构实体的详细人设提示词""" - attrs_str = json.dumps(entity_attributes, ensure_ascii=False) if entity_attributes else "无" - context_str = context[:3000] if context else "无额外上下文" + attrs_str = json.dumps(entity_attributes, ensure_ascii=False) if entity_attributes else "None" + context_str = context[:3000] if context else "No additional context" return f"""Generate detailed social media account settings for organization/group entity, maximizing fidelity to real-world situations. -实体名称: {entity_name} -实体类型: {entity_type} -实体摘要: {entity_summary} -实体属性: {attrs_str} +Entity name: {entity_name} +Entity type: {entity_type} +Entity summary: {entity_summary} +Entity attributes: {attrs_str} -上下文信息: +Context information: {context_str} -请生成JSON，包含以下字段: +Please generate JSON with the following fields: -1. bio: 官方账号简介，200字，专业得体 -2. persona: 详细账号设定描述（2000字的纯文本），需包含: - - 机构基本信息（正式名称、机构性质、成立背景、主要职能） - - 账号定位（账号类型、目标受众、核心功能） - - 发言风格（语言特点、常用表达、禁忌话题） - - 发布内容特点（内容类型、发布频率、活跃时间段） - - 立场态度（对核心话题的官方立场、面对争议的处理方式） - - 特殊说明（代表的群体画像、运营习惯） - - 机构记忆（机构人设的重要部分，要介绍这个机构与事件的关联，以及这个机构在事件中的已有动作与反应） -3. age: 固定填30（机构账号的虚拟年龄） -4. gender: 固定填"other"（机构账号使用other表示非个人） -5. mbti: MBTI类型，用于描述账号风格，如ISTJ代表严谨保守 -6. country: 国家（使用中文，如"中国"） -7. profession: 机构职能描述 -8. interested_topics: 关注领域数组 +1. bio: Official account bio, 200 characters, professional and appropriate +2. persona: Detailed account settings description (2000 chars plain text), must include: + - Organization basics (official name, nature, founding background, main functions) + - Account positioning (account type, target audience, core function) + - Speaking style (language traits, common expressions, taboo topics) + - Content characteristics (content type, posting frequency, active hours) + - Stance (official position on core topics, handling of controversies) + - Special notes (represented group profile, operational habits) + - Organization memory (important part: organization's connection to the event, and their existing actions and reactions) +3. age: Always 30 (virtual age for organization accounts) +4. gender: Always "other" (organization accounts use "other" to indicate non-personal) +5. mbti: MBTI type to describe account style, e.g. ISTJ for rigorous and conservative +6. country: Country (use English, e.g. "Thailand", "China") +7. profession: Organization function description +8. interested_topics: Array of areas of interest -重要: -- 所有字段值必须是字符串或数字，不允许null值 -- persona必须是一段连贯的文字描述，不要使用换行符 +Important: +- All field values must be strings or numbers, null values not allowed +- persona must be a coherent text description，不要使用换行符 - {get_language_instruction()} (gender字段必须用英文"other") -- age必须是整数30，gender必须是字符串"other" +- age must be integer 30, gender must be string "other" - Organization account posts must match its identity positioning""" def _generate_profile_rule_based( diff --git a/backend/app/services/simulation_config_generator.py b/backend/app/services/simulation_config_generator.py index 3729e2a..7ad919b 100644 --- a/backend/app/services/simulation_config_generator.py +++ b/backend/app/services/simulation_config_generator.py @@ -544,24 +544,24 @@ class SimulationConfigGenerator: {context_truncated} -## 任务 +## Task Please generate time config JSON. -### 基本原则（仅供参考，需根据具体事件和参与群体灵活调整）： +### Basic Principles (for reference only, adjust flexibly based on specific event and participant group): - Please infer the target user group's timezone and daily routine based on the simulation scenario. Below is a reference example for UTC+8 -- 凌晨0-5点几乎无人活动（活跃度系数0.05） -- 早上6-8点逐渐活跃（活跃度系数0.4） -- 工作时间9-18点中等活跃（活跃度系数0.7） -- 晚间19-22点是高峰期（活跃度系数1.5） -- 23点后活跃度下降（活跃度系数0.5） -- 一般规律：凌晨低活跃、早间渐增、工作时段中等、晚间高峰 -- **重要**：以下示例值仅供参考，你需要根据事件性质、参与群体特点来调整具体时段 - - 例如：学生群体高峰可能是21-23点；媒体全天活跃；官方机构只在工作时间 - - 例如：突发热点可能导致深夜也有讨论，off_peak_hours 可适当缩短 +- Midnight 0-5am: almost no activity (activity multiplier 0.05) +- Morning 6-8am: gradually active (activity multiplier 0.4) +- Work hours 9-18pm: moderately active (activity multiplier 0.7) +- Evening 19-22pm: peak hours (activity multiplier 1.5) +- After 23pm: activity decreases (activity multiplier 0.5) +- General pattern: low activity late night, increasing morning, moderate during work hours, peak in evening +- **Important**: The example values above are for reference only. You must adjust specific time periods based on event nature and participant characteristics + - Example: Student groups may peak at 21-23pm; media is active all day; government agencies only during work hours + - Example: Breaking hot topics may cause late-night discussions, off_peak_hours can be shortened accordingly -### 返回JSON格式（不要markdown） +### Return in JSON format (no markdown) -示例： +Example: {{ "total_simulation_hours": 72, "minutes_per_round": 60, @@ -574,15 +574,15 @@ Please generate time config JSON. "reasoning": "Explanation of time config for this event" }} -字段说明： -- total_simulation_hours (int): 模拟总时长，24-168小时，突发事件短、持续话题长 -- minutes_per_round (int): 每轮时长，30-120分钟，建议60分钟 -- agents_per_hour_min (int): 每小时最少激活Agent数（取值范围: 1-{max_agents_allowed}） -- agents_per_hour_max (int): 每小时最多激活Agent数（取值范围: 1-{max_agents_allowed}） -- peak_hours (int数组): 高峰时段，根据事件参与群体调整 -- off_peak_hours (int数组): 低谷时段，通常深夜凌晨 -- morning_hours (int数组): 早间时段 -- work_hours (int数组): 工作时段 +Field descriptions: +- total_simulation_hours (int): Total simulation duration, 24-168 hours, shorter for breaking events, longer for ongoing topics +- minutes_per_round (int): Minutes per round, 30-120 minutes, recommended 60 minutes +- agents_per_hour_min (int): Minimum agents activated per hour (range: 1-{max_agents_allowed}) +- agents_per_hour_max (int): Maximum agents activated per hour (range: 1-{max_agents_allowed}) +- peak_hours (int array): Peak hours, adjust based on event participant group +- off_peak_hours (int array): Off-peak hours, typically late night/early morning +- morning_hours (int array): Morning hours +- work_hours (int array): Work hours - reasoning (string): Brief explanation of why this config was chosen""" system_prompt = "You are a social media simulation expert. Return pure JSON format. Time config must match the target user group's daily routine in the simulation scenario." @@ -679,27 +679,27 @@ Simulation requirements: {simulation_requirement} {context_truncated} -## 可用实体类型及示例 +## Available Entity Types and Examples {type_info} -## 任务 +## Task Please generate event config JSON: -- 提取热点话题关键词 -- 描述舆论发展方向 -- 设计初始帖子内容，**每个帖子必须指定 poster_type（发布者类型）** +- Extract hot topic keywords +- Describe the narrative direction of public opinion +- Design initial post content, **each post must specify poster_type (publisher type)** -**重要**: poster_type 必须从上面的"可用实体类型"中选择，这样初始帖子才能分配给合适的 Agent 发布。 -例如：官方声明应由 Official/University 类型发布，新闻由 MediaOutlet 发布，学生观点由 Student 发布。 +**Important**: poster_type must be selected from the "Available Entity Types" above, so initial posts can be assigned to suitable Agents. +For example: Official statements should be posted by Official/University types, news by MediaOutlet, student opinions by Student. Return in JSON format (no markdown): {{ - "hot_topics": ["关键词1", "关键词2", ...], - "narrative_direction": "<舆论发展方向描述>", + "hot_topics": ["keyword1", "keyword2", ...], + "narrative_direction": "", "initial_posts": [ - {{"content": "帖子内容", "poster_type": "实体类型（必须从可用类型中选择）"}}, + {{"content": "post content", "poster_type": "entity type (must be from available types)"}}, ... ], - "reasoning": "<简要说明>" + "reasoning": "" }}""" system_prompt = "You are a public opinion analysis expert. Return pure JSON format. Note: poster_type must exactly match available entity types." @@ -834,33 +834,33 @@ Return in JSON format (no markdown): Simulation requirements: {simulation_requirement} -## 实体列表 +## Entity List ```json {json.dumps(entity_list, ensure_ascii=False, indent=2)} ``` -## 任务 +## Task Generate activity config for each entity, noting: - **Time must match target user group routine**: Below is a reference (UTC+8), please adjust based on simulation scenario -- **官方机构**（University/GovernmentAgency）：活跃度低(0.1-0.3)，工作时间(9-17)活动，响应慢(60-240分钟)，影响力高(2.5-3.0) -- **媒体**（MediaOutlet）：活跃度中(0.4-0.6)，全天活动(8-23)，响应快(5-30分钟)，影响力高(2.0-2.5) -- **个人**（Student/Person/Alumni）：活跃度高(0.6-0.9)，主要晚间活动(18-23)，响应快(1-15分钟)，影响力低(0.8-1.2) -- **公众人物/专家**：活跃度中(0.4-0.6)，影响力中高(1.5-2.0) +- **Government agencies** (University/GovernmentAgency): Low activity (0.1-0.3), active during work hours (9-17), slow response (60-240 min), high influence (2.5-3.0) +- **Media** (MediaOutlet): Medium activity (0.4-0.6), active all day (8-23), fast response (5-30 min), high influence (2.0-2.5) +- **Individuals** (Student/Person/Alumni): High activity (0.6-0.9), mainly evening (18-23), fast response (1-15 min), low influence (0.8-1.2) +- **Public figures/experts**: Medium activity (0.4-0.6), medium-high influence (1.5-2.0) Return in JSON format (no markdown): {{ "agent_configs": [ {{ - "agent_id": <必须与输入一致>, + "agent_id": , "activity_level": <0.0-1.0>, - "posts_per_hour": <发帖频率>, - "comments_per_hour": <评论频率>, - "active_hours": [<活跃小时列表，考虑中国人作息>], - "response_delay_min": <最小响应延迟分钟>, - "response_delay_max": <最大响应延迟分钟>, - "sentiment_bias": <-1.0到1.0>, + "posts_per_hour": , + "comments_per_hour": , + "active_hours": [], + "response_delay_min": , + "response_delay_max": , + "sentiment_bias": <-1.0 to 1.0>, "stance": "", - "influence_weight": <影响力权重> + "influence_weight": }}, ... ]