feihong-code 8.0.0 → 8.0.2

package/agent-team/agents/l4_analysis.py

@@ -0,0 +1,118 @@
+"""L4 分析决策层：10 GEO商机诊断官 / 16 行业深度分析师 / 17 投资研究咨询官。"""
+import json
+from datetime import datetime
+
+from core.agent import BaseAgent
+
+
+def _ts() -> str:
+    return datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+
+# ---------------------------------------------------------------- 10
+def run_logic_10(data, llm=None):
+    """GEO 商机诊断：基于地理与经营数据诊断潜在商机。"""
+    region = data.get("region", "泉州")
+    business = data.get("business_type", "企业服务")
+    diagnosis = {
+        "region": region,
+        "business_type": business,
+        "zones": [
+            {"zone": f"{region}·核心商圈", "score": 92, "reason": "人口密度高、企业密集"},
+            {"zone": f"{region}·产业园区", "score": 85, "reason": "目标客群集中"},
+        ],
+        "priority": "建议优先覆盖核心商圈",
+        "diagnosed_at": _ts(),
+    }
+    ai_note = ""
+    if llm is not None and getattr(llm, "provider", "mock") != "mock":
+        try:
+            ai_note = llm.complete(
+                f"区域 {region}、业态 {business}，给出获客优先级判断，不超过 60 字。",
+                system="你是 GEO 商业诊断专家。")
+        except Exception:
+            ai_note = ""
+    return {"diagnosis": diagnosis, "ai_note": ai_note}
+
+
+class GeoDiagnosisAgent(BaseAgent):
+    module_id = "10"
+    name = "GEO商机诊断官"
+    layer = "L4"
+    priority = "P3"
+    description = "基于地理与经营数据诊断潜在商机"
+    input_schema = {"region": "目标区域", "business_type": "业态"}
+    output_keys = {"diagnosis": "商机诊断报告"}
+    run_logic = staticmethod(run_logic_10)
+
+
+# ---------------------------------------------------------------- 16
+def run_logic_16(data, llm=None):
+    """行业深度分析：输出行业趋势与竞争格局。"""
+    industry = data.get("industry", "企业服务")
+    scope = data.get("scope", "国内")
+    report = {
+        "industry": industry,
+        "scope": scope,
+        "market_size": "示例口径：120 亿元（需以权威数据源校准）",
+        "growth": "年增速约 18%",
+        "competition": "格局分散，头部集中度提升中",
+        "conclusion": "进入窗口期，建议聚焦细分场景建立壁垒",
+        "data_note": "本报告为分析框架示意，关键数据需接入权威数据源后校准",
+        "generated_at": _ts(),
+    }
+    ai_note = ""
+    if llm is not None and getattr(llm, "provider", "mock") != "mock":
+        try:
+            ai_note = llm.complete(
+                f"行业 {industry}、范围 {scope}，给 1 条最关键的竞争判断，不超过 60 字。",
+                system="你是行业研究首席分析师。")
+        except Exception:
+            ai_note = ""
+    return {"report": report, "ai_note": ai_note}
+
+
+class IndustryAnalystAgent(BaseAgent):
+    module_id = "16"
+    name = "行业深度分析师"
+    layer = "L4"
+    priority = "P3"
+    description = "输出行业趋势与竞争格局分析"
+    input_schema = {"industry": "行业", "scope": "范围"}
+    output_keys = {"report": "行业分析报告"}
+    run_logic = staticmethod(run_logic_16)
+
+
+# ---------------------------------------------------------------- 17
+def run_logic_17(data, llm=None):
+    """投资研究：项目 / 投资研究与尽调辅助。"""
+    company = data.get("company", "示例项目")
+    materials = data.get("materials", "BP 摘要")
+    report = {
+        "company": company,
+        "business_model": "B2B 订阅 + 增值服务",
+        "risks": ["获客成本偏高", "复购率待验证"],
+        "valuation_notes": "建议以 DCF + 可比公司交叉验证（需财务数据支撑）",
+        "conclusion": "方向成立，需补充现金流与复购数据后再决策",
+        "generated_at": _ts(),
+    }
+    ai_note = ""
+    if llm is not None and getattr(llm, "provider", "mock") != "mock":
+        try:
+            ai_note = llm.complete(
+                f"项目 {company}、材料 {materials[:100]}，提示最需要核实的一个风险点，不超过 50 字。",
+                system="你是投资研究顾问。")
+        except Exception:
+            ai_note = ""
+    return {"report": report, "ai_note": ai_note}
+
+
+class InvestResearchAgent(BaseAgent):
+    module_id = "17"
+    name = "投资研究咨询官"
+    layer = "L4"
+    priority = "P3"
+    description = "项目 / 投资研究与尽调辅助"
+    input_schema = {"company": "项目/公司", "materials": "材料摘要"}
+    output_keys = {"report": "投研报告"}
+    run_logic = staticmethod(run_logic_17)

package/agent-team/agents/l5_management.py

@@ -0,0 +1,52 @@
+"""L5 系统管理层：11 一人公司 AI 员工调度官。
+
+调度官的运行骨架在 core/scheduler.py；此处注册为可编排模块，
+提供「调度计划生成」能力（供其他模块/智能体调用），并声明系统级元信息。
+"""
+import json
+from datetime import datetime
+
+from core.agent import BaseAgent
+
+
+def _ts() -> str:
+    return datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+
+
+def run_logic_11(data, llm=None):
+    """调度计划：按优先级与依赖生成执行顺序。"""
+    tasks = data.get("tasks") or [
+        {"module_id": "01", "task_type": "lead_scan", "priority": "P0"},
+        {"module_id": "02", "task_type": "follow_up", "priority": "P0"},
+        {"module_id": "04", "task_type": "chat_analysis", "priority": "P0"},
+        {"module_id": "09", "task_type": "tiering", "priority": "P0"},
+    ]
+    order = sorted(tasks, key=lambda t: {"P0": 0, "P1": 1, "P2": 2, "P3": 3}.get(t.get("priority", "P1"), 1))
+    schedule = {
+        "order": [t["module_id"] for t in order],
+        "dependency_rule": "前置任务 success 后触发后续任务",
+        "retry_policy": "失败自动重试，上限 3 次，超限转 blocked 并告警",
+        "human_gate": "涉及对外触达的任务在人工确认节点暂停",
+        "estimated_minutes": len(order) * 3,
+        "planned_at": _ts(),
+    }
+    ai_note = ""
+    if llm is not None and getattr(llm, "provider", "mock") != "mock":
+        try:
+            ai_note = llm.complete(
+                f"任务列表 {json.dumps(tasks, ensure_ascii=False)}，指出编排上的一个风险，不超过 50 字。",
+                system="你是系统调度架构师。")
+        except Exception:
+            ai_note = ""
+    return {"schedule": schedule, "ai_note": ai_note}
+
+
+class DispatcherAgent(BaseAgent):
+    module_id = "11"
+    name = "一人公司AI员工调度官"
+    layer = "L5"
+    priority = "基建"
+    description = "编排、调度、监控全部 AI 员工，管理任务全生命周期"
+    input_schema = {"tasks": "待编排任务列表"}
+    output_keys = {"schedule": "执行顺序与策略"}
+    run_logic = staticmethod(run_logic_11)

package/agent-team/agents/registry.py

@@ -0,0 +1,60 @@
+"""模块注册表：集中注册 17 个 AI 员工，支持按 module_id 查询与实例化。"""
+from typing import Dict, List, Optional
+
+from core.agent import BaseAgent
+
+from agents.l1_acquisition import (
+    CommentOpportunityAgent, DailyTopicAgent, ForeignTradeAgent,
+    LeadRadarAgent, XiaohongshuAgent,
+)
+from agents.l2_content import (
+    AutoEditorAgent, BloggerTeardownAgent, GzhPublisherAgent,
+    VideoScriptAgent, ViralRadarAgent,
+)
+from agents.l3_conversion import ChatAnalysisAgent, FollowUpAgent, TieringAgent
+from agents.l4_analysis import GeoDiagnosisAgent, IndustryAnalystAgent, InvestResearchAgent
+from agents.l5_management import DispatcherAgent
+
+# 模块类注册表（顺序即展示顺序）
+AGENT_CLASSES: List[type] = [
+    LeadRadarAgent,        # 01
+    FollowUpAgent,         # 02
+    CommentOpportunityAgent,  # 03
+    ChatAnalysisAgent,     # 04
+    ViralRadarAgent,       # 05
+    DailyTopicAgent,       # 06
+    ForeignTradeAgent,     # 07
+    XiaohongshuAgent,      # 08
+    TieringAgent,          # 09
+    GeoDiagnosisAgent,     # 10
+    DispatcherAgent,       # 11
+    VideoScriptAgent,      # 12
+    BloggerTeardownAgent,  # 13
+    AutoEditorAgent,       # 14
+    GzhPublisherAgent,     # 15
+    IndustryAnalystAgent,  # 16
+    InvestResearchAgent,   # 17
+]
+
+
+def get_agent_classes() -> List[type]:
+    return list(AGENT_CLASSES)
+
+
+def find_agent_class(module_id: str) -> Optional[type]:
+    for cls in AGENT_CLASSES:
+        if cls.module_id == module_id:
+            return cls
+    return None
+
+
+def instantiate_all(llm) -> Dict[str, BaseAgent]:
+    """实例化全部模块（注入同一 LLM 客户端）。"""
+    agents: Dict[str, BaseAgent] = {}
+    for cls in AGENT_CLASSES:
+        agents[cls.module_id] = cls(llm=llm)
+    return agents
+
+
+def list_meta() -> List[Dict]:
+    return [cls(llm=None).meta() for cls in AGENT_CLASSES]

package/agent-team/config.py

@@ -0,0 +1,39 @@
+"""全局配置：从环境变量与 .env 文件读取，禁止硬编码密钥。
+
+- LLM_PROVIDER=mock   无 API Key 时用本地确定性 mock，跑通全链路
+- LLM_PROVIDER=doubao 接入豆包大模型（需配置 DOUBAO_API_KEY / DOUBAO_MODEL）
+"""
+import os
+from pathlib import Path
+
+BASE_DIR = Path(__file__).resolve().parent
+
+
+def _load_dotenv(path: Path) -> None:
+    """极简 .env 加载器（KEY=VALUE，忽略 # 注释与空行）。"""
+    if not path.exists():
+        return
+    for line in path.read_text(encoding="utf-8").splitlines():
+        line = line.strip()
+        if not line or line.startswith("#") or "=" not in line:
+            continue
+        key, _, value = line.partition("=")
+        key, value = key.strip(), value.strip().strip('"').strip("'")
+        if key and key not in os.environ:
+            os.environ[key] = value
+
+
+_load_dotenv(BASE_DIR / ".env")
+
+# 数据存储
+DB_PATH = os.getenv("DB_PATH", str(BASE_DIR / "feihongzhi.db"))
+
+# LLM
+LLM_PROVIDER = os.getenv("LLM_PROVIDER", "mock")          # mock | doubao
+DOUBAO_API_KEY = os.getenv("DOUBAO_API_KEY", "")
+DOUBAO_MODEL = os.getenv("DOUBAO_MODEL", "doubao-seed-1-6-250615")
+DOUBAO_BASE_URL = os.getenv("DOUBAO_BASE_URL", "https://ark.cn-beijing.volces.com/api/v3")
+
+# 调度
+MAX_RETRY = int(os.getenv("MAX_RETRY", "3"))              # 失败重试上限
+ALERT_LEVEL = os.getenv("ALERT_LEVEL", "log")             # log | stdout

package/agent-team/core/agent.py

@@ -0,0 +1,73 @@
+"""BaseAgent：所有 AI 员工模块的基类。
+
+设计要点（支持 Skill 单独分发）：
+- 每个模块的核心业务逻辑写在模块级纯函数 run_logic(input_data, llm) 中，
+  不依赖类实例与数据库，可被 skill_packager 用 inspect.getsource 内嵌导出为独立 run.py。
+- BaseAgent 负责：契约校验（输入/输出 schema）、trace 记录、LLM 注入、错误上抛。
+"""
+from typing import Any, Callable, Dict, Optional
+
+from core.task import Task, TaskError
+
+# 简化 JSON Schema：字段名 -> 描述/约束
+EMPTY_SCHEMA: Dict[str, Any] = {}
+
+
+class BaseAgent:
+    module_id: str = ""
+    name: str = ""
+    layer: str = ""
+    priority: str = "P1"          # P0 | P1 | P2 | P3 | 基建
+    description: str = ""
+    input_schema: Dict[str, str] = EMPTY_SCHEMA      # 必填字段: 说明
+    output_keys: Dict[str, str] = EMPTY_SCHEMA       # 输出字段: 说明
+    need_human: bool = False                          # 是否需人工确认
+    run_logic: Callable[[Dict[str, Any], Any], Dict[str, Any]] = staticmethod(
+        lambda data, llm: {"result": data}
+    )
+
+    def __init__(self, llm: Optional[Any] = None):
+        self.llm = llm
+
+    # ------------------------------------------------------------------
+    def validate_input(self, data: Any) -> Dict[str, Any]:
+        if not isinstance(data, dict):
+            raise TaskError(f"[{self.module_id}] 输入必须是 JSON 对象，收到 {type(data).__name__}")
+        for field, desc in self.input_schema.items():
+            if field not in data:
+                raise TaskError(f"[{self.module_id}] 缺少必填输入字段 {field}（{desc}）")
+        return data
+
+    def validate_output(self, output: Any) -> Dict[str, Any]:
+        if not isinstance(output, dict):
+            raise TaskError(f"[{self.module_id}] 输出必须是 JSON 对象，收到 {type(output).__name__}")
+        return output
+
+    # ------------------------------------------------------------------
+    def execute(self, task: Task) -> Dict[str, Any]:
+        """调度官调用入口：契约校验 -> 纯函数执行 -> 输出校验 -> trace。"""
+        data = self.validate_input(task.input_json)
+        task.append_trace("validate_input", "ok")
+        output = self.run_logic(data, self.llm)
+        output = self.validate_output(output)
+        task.append_trace("run_logic", f"output_keys={sorted(output.keys())}")
+        return output
+
+    def persist(self, output: Dict[str, Any], task: Task) -> None:
+        """可选持久化钩子：模块结果落库（leads/customers/content_assets）。
+
+        仅主系统调度时调用；Skill 分发版 run.py 保持纯逻辑，由调用方决定落库。
+        """
+        return None
+
+    def meta(self) -> Dict[str, Any]:
+        return {
+            "module_id": self.module_id,
+            "name": self.name,
+            "layer": self.layer,
+            "priority": self.priority,
+            "description": self.description,
+            "input_schema": self.input_schema,
+            "output_keys": self.output_keys,
+            "need_human": self.need_human,
+        }