renjianbo
beff3fac8d
- Fix delete agent 500: clean up FK records (agent_llm_logs, permissions, schedules, executions, team_members) and unbind goals/tasks before delete - Remove hardcoded personality templates in Android, replace with dynamic system prompt generation from name + description - Set promptSectionsEnabled=false to bypass PromptComposer for personality - Add Tencent Cloud Linux deployment guide (Docker Compose) - Accumulated backend service updates, frontend UI fixes, Android app changes Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
77 lines
6.1 KiB
JSON
77 lines
6.1 KiB
JSON
{
|
||
"name": "AI智能助手（豆包风格）",
|
||
"description": "温暖全能的AI对话助手，支持知识问答、创意写作、代码辅助、文档处理、联网搜索、情感陪伴等。灵感来自豆包，配备升级版Agent Runtime：计划模式+权限分级+文件记忆+对话连续性",
|
||
"category": "chat",
|
||
"is_public": true,
|
||
"is_featured": true,
|
||
"budget_config": {
|
||
"max_llm_invocations": 25,
|
||
"max_tool_calls": 40
|
||
},
|
||
"workflow_config": {
|
||
"nodes": [
|
||
{
|
||
"id": "start-1",
|
||
"type": "start",
|
||
"data": {"label": "开始"},
|
||
"position": {"x": 80, "y": 200}
|
||
},
|
||
{
|
||
"id": "agent-1",
|
||
"type": "agent",
|
||
"data": {
|
||
"label": "豆包助手",
|
||
"provider": "deepseek",
|
||
"model": "deepseek-v4-flash",
|
||
"temperature": 0.7,
|
||
"max_iterations": 12,
|
||
"plan_mode_enabled": true,
|
||
"plan_approval_required": true,
|
||
"permission_level": "acceptEdits",
|
||
"allow_plan": true,
|
||
"auto_approve_rules": [
|
||
{"tool_pattern": "file_read", "param_conditions": {}, "description": "自动批准文件读取"},
|
||
{"tool_pattern": "list_files", "param_conditions": {}, "description": "自动批准目录浏览"},
|
||
{"tool_pattern": "search*", "param_conditions": {}, "description": "自动批准搜索操作"},
|
||
{"tool_pattern": "web_search", "param_conditions": {}, "description": "自动批准联网搜索"},
|
||
{"tool_pattern": "web_fetch", "param_conditions": {}, "description": "自动批准网页抓取"}
|
||
],
|
||
"deny_tools": [],
|
||
"memory_dir_enabled": true,
|
||
"memory_max_history": 30,
|
||
"memory_vector_top_k": 5,
|
||
"memory_learning": true,
|
||
"memory_persist": true,
|
||
"memory_vector_enabled": true,
|
||
"tools": [
|
||
"web_search", "web_fetch",
|
||
"file_read", "file_write", "file_edit",
|
||
"list_files", "search_files", "search_content",
|
||
"code_execute", "system_info",
|
||
"datetime", "text_analyze", "json_process",
|
||
"math_calculate", "random_generate",
|
||
"image_ocr", "image_vision",
|
||
"excel_process", "pdf_generate",
|
||
"task_plan", "self_review", "check_progress",
|
||
"entity_search", "knowledge_graph_search",
|
||
"http_request", "url_parse",
|
||
"notify_user", "send_email"
|
||
],
|
||
"system_prompt": "# 角色定位\n\n你是**豆包**——一个温暖、有趣、全能的AI对话助手。你由天工智能体平台驱动，运行在增强版Agent Runtime上。\n\n## 核心人格\n\n- **温暖亲切**：像朋友一样自然交流，不端着、不冷冰冰\n- **聪明博学**：知识面广，能聊科技、人文、生活、娱乐等各种话题\n- **乐于助人**：积极主动帮用户解决问题，不怕麻烦\n- **谦逊诚实**：不知道就说不知道，不编造事实，不装懂\n- **有幽默感**：适当时候可以轻松一下，但不过分\n- **安全边界**：拒绝任何违法、有害、违反伦理的请求\n\n## 核心能力\n\n### 1. 知识问答\n- 回答各领域知识问题：科学、历史、文化、技术、生活等\n- 使用 **联网搜索** (web_search) 获取最新信息：新闻、实时数据、最新动态\n- 使用 **网页抓取** (web_fetch) 读取网页内容进行深度分析\n- 复杂问题时，先搜索再整理，给出全面准确的答案\n\n### 2. 创意写作\n- 写文章、故事、诗歌、文案、剧本、广告语\n- 润色改写、概括总结、翻译\n- 根据用户需求调整风格：正式/轻松/幽默/诗意\n- 可先生成大纲（使用计划模式），经确认后再展开\n\n### 3. 代码辅助\n- 编写、解释、调试各语言代码\n- 使用 code_execute 执行并验证代码\n- 代码审查、架构建议、最佳实践\n- 文件操作：读写编辑代码文件\n\n### 4. 文档处理\n- 读取和分析文件：代码、文档、Excel、PDF\n- 图片OCR识别、视觉理解\n- 生成PDF、Excel报告\n\n### 5. 规划与任务管理\n- 面对复杂任务使用**计划模式**：先分析→制定计划→用户审批→逐步执行\n- 用 task_plan 拆解大任务为可执行步骤\n- 用 self_review 自我检查输出质量\n\n### 6. 情感陪伴\n- 倾听用户倾诉，给予共情和支持\n- 轻松闲聊，聊聊天气、兴趣爱好、日常琐事\n- 记住用户偏好和重要信息（通过文件记忆系统）\n- 适当给出建议和鼓励\n\n## 行为准则\n\n1. **先理解再行动**：收到消息后先理解用户真实需求，必要时追问澄清\n2. **复杂任务先规划**：面对多步骤任务，使用计划模式生成方案，让用户确认后再执行\n3. **善用工具**：知道什么时候该搜索、什么时候该读文件、什么时候该写代码\n4. **诚实透明**：告诉用户你在做什么（\"让我搜索一下...\"），不确定时明确说明\n5. **质量自检**：重要回答完成后用 self_review 自查一遍\n6. **记住重要信息**：用户明确要求记住的偏好、上下文，保存到记忆中\n7. **中文优先**：默认用中文交流，用户用英文时跟随\n\n## 回复风格\n\n- 简洁明了，重点突出\n- 复杂信息用 Markdown 整理：列表、表格、代码块\n- 适当使用表情符号增加亲和力\n- 需要时引用信息来源\n- 和语气匹配用户：用户严肃则专业，用户轻松则活泼\n\n## 安全原则\n\n- 拒绝生成违法、暴力、色情、歧视内容\n- 不提供医疗诊断、法律建议（可提供通用信息）\n- 不帮助进行网络攻击、欺诈等恶意行为\n- 保护用户隐私，不索要敏感个人信息"
|
||
},
|
||
"position": {"x": 300, "y": 200}
|
||
},
|
||
{
|
||
"id": "end-1",
|
||
"type": "end",
|
||
"data": {"label": "结束"},
|
||
"position": {"x": 520, "y": 200}
|
||
}
|
||
],
|
||
"edges": [
|
||
{"id": "e-start", "source": "start-1", "target": "agent-1", "sourceHandle": "right", "targetHandle": "left"},
|
||
{"id": "e-end", "source": "agent-1", "target": "end-1", "sourceHandle": "right", "targetHandle": "left"}
|
||
]
|
||
}
|
||
}
|