用 Codex 把业务运营文档从"手写苦力"变成"自动流水线"

Task: Read the files in /workspace/inputs/ (initiative.csv, confluence_notes.md).
Generate a concise initiative brief in Markdown that includes:
- Objective (1-2 sentences)
- Key deliverables (bulleted list)
- Timeline summary
- Risks and mitigations
- Stakeholders

Write the output to /workspace/output/initiative_brief.md.

Task: Read okr_q2.md, market_analysis.md, and retro_q2.md from /workspace/inputs/.
Write a strategy update for the executive team in Chinese. Structure it as:
- 本季度核心成果（3-5 条）
- 市场环境变化及影响
- 下季度战略调整建议
- 需领导决策的关键问题

Keep each section under 200 words. Save to /workspace/output/strategy_update.md.

Task: Read option_a_analysis.md, option_b_analysis.md, and risk_register.csv from /workspace/inputs/.
Create a leadership decision packet in Markdown with:
1. Executive summary (1 paragraph)
2. Options comparison table (columns: Option, Cost, Timeline, Risk Level, Expected Outcome)
3. Risk assessment for each option
4. Recommendation with reasoning

Save to /workspace/output/decision_packet.md.

Task: Read git_log.txt, slack_summary.json, and board_status.csv from /workspace/inputs/.
Generate a weekly progress update email in Chinese that includes:
- 本周完成事项（按优先级排列）
- 进行中的工作及预计完成时间
- 阻塞项及需要的支持
- 下周计划

Format as an email body. Save to /workspace/output/weekly_update.md.

import os
import json
import time
from pathlib import Path
from openai import OpenAI

client = OpenAI()  # 确保已设置 OPENAI_API_KEY 环境变量

# ---------- 配置 ----------
PROJECTS = [
    {"name": "payment_v2", "dir": "payment_v2"},
    {"name": "user_growth", "dir": "user_growth"},
    {"name": "infra_migration", "dir": "infra_migration"},
]

BRIEF_PROMPT_TEMPLATE = """\
Read all files in the attached workspace for the project "{project_name}".
Generate a concise initiative brief in Chinese (Markdown format) that includes:
- 项目目标（1-2 句）
- 关键交付物（列表）
- 时间线摘要
- 风险与应对
- 相关方

Write the output to initiative_brief.md.
"""

OUTPUT_DIR = Path("./generated_briefs")
OUTPUT_DIR.mkdir(exist_ok=True)

# ---------- 读取项目素材 ----------
def load_project_files(project_dir: str) -> list[dict]:
    """扫描项目目录，把所有文件内容打包成 Codex 可用的输入"""
    base = Path("./project_inputs") / project_dir
    files = []
    for f in base.glob("*"):
        if f.is_file():
            content = f.read_text(encoding="utf-8", errors="replace")
            files.append({"file_name": f.name, "content": content})
    return files

# ---------- 调用 Codex ----------
def generate_brief(project: dict) -> str:
    """提交任务到 Codex，返回生成的简报内容"""
    prompt = BRIEF_PROMPT_TEMPLATE.format(project_name=project["name"])
    project_files = load_project_files(project["dir"])

    # 把素材文件作为上下文传入
    file_contexts = "\n\n".join(
        [f"--- {f['file_name']} ---\n{f['content']}" for f in project_files]
    )
    full_input = f"{prompt}\n\n以下是项目素材：\n\n{file_contexts}"

    response = client.responses.create(
        model="codex-mini-latest",  # Codex 代理模型
        input=full_input,
        tools=[{"type": "codex"}],   # 启用 Codex 工具模式
    )

    # 从响应中提取文本输出
    for item in response.output:
        if item.type == "message":
            for content in item.content:
                if content.type == "output_text":
                    return content.text

    return ""

# ---------- 批量执行 ----------
for project in PROJECTS:
    print(f"正在生成: {project['name']} ...")
    brief = generate_brief(project)
    if brief:
        out_path = OUTPUT_DIR / f"{project['name']}_brief.md"
        out_path.write_text(brief, encoding="utf-8")
        print(f"  ✓ 已保存到 {out_path}")
    else:
        print(f"  ✗ 生成失败，请检查素材目录")
    time.sleep(2)  # 避免速率限制

print(f"\n全部完成，输出目录: {OUTPUT_DIR}")

project_inputs/
  payment_v2/
    jira_export.csv
    confluence_notes.md
    risk_register.csv
  user_growth/
    okr_q3.md
    experiment_results.json
  infra_migration/
    migration_plan.md
    timeline.csv