Topic Radar｜有據選題
Choose your next topic with an explainable ranking
用看得懂的評分，決定下一個選題
For creators and editors with a shortlist: rank supplied topics using audience relevance, usefulness and novelty, keeping each reason and source visible. Local Python/Skill download, with no real-time news feed or popularity prediction.
提供給已有候選題目的創作者與編輯：依受眾相關性、實用性與新穎性排序，保留每項理由與來源。本機 Python／Skill 下載包，不含即時新聞串流或人氣預測。
Watch the walkthrough / 觀看操作影片
Actual example — synthetic data
實際範例 — 合成資料
The included demo ranks the shipping update 42/50, ahead of the supplier relationship at 34/50. Every topic retains its three input scores, explanations and quoted source references. These are analyst judgments, not measured audience interest or probabilities.
套件實際範例將出貨更新排在 42/50 分，供應關係為 34/50 分。每個題目保留三項輸入評分、理由與來源引用。評分屬分析者判斷，不是實測受眾興趣或機率。
| Input | Output |
|---|---|
| Topics + three 1–5 scores + reasons + excerpts | Ranked shortlist with scores, reasons and sources |
| 題目＋三項 1～5 分評分＋理由＋摘錄 | 附分數、理由與來源的排序清單 |
Formula: relevance × 4 + usefulness × 4 + novelty × 2. Maximum 50; equal scores sort by topic ID.
公式：相關性 × 4＋實用性 × 4＋新穎性 × 2，滿分 50；同分依題目 ID 排序。
Three steps
三步開始
- Unzip and open a terminal in the SKILL.md folder. Specify your audience and prepare topics, scores, reasons and sources using references/input.md.
解壓後在 SKILL.md 資料夾開啟終端機；確定受眾，依 references/input.md 準備題目、評分、理由與來源。 - Run the included demo, then replace its JSON with your own inputs.
執行套件範例，再換成自己的 JSON 資料。
python3 scripts/rank.py --input examples/demo.json
- Receive the ranked Markdown report in the terminal. Your own AI client may help propose scores using SKILL.md; those remain labeled judgments.
在終端機取得 Markdown 排序報告。自己的 AI 客戶端可依 SKILL.md 協助提出評分，仍屬已標示的分析判斷。
The script calculates from supplied scores; it does not measure the whole internet or guarantee traffic, engagement or sales.
腳本依輸入分數計算，不衡量全網熱度，也不保證流量、互動或成交。
Requirements and limits
環境與限制
Python 3.10+ recommended. Command-line tool, no GUI; standard library only and no model API calls. Tested on macOS; clean Windows/Linux installation is unverified. On Windows, replace python3 with py -3. Input schema and limits are in references/input.md.
建議 Python 3.10 以上，需要終端機操作，沒有圖形介面；只使用標準函式庫，不呼叫模型 API。已於 macOS 實測，Windows／Linux 全新安裝尚未驗證。Windows 可用 py -3 取代 python3；資料格式與限制見 references/input.md。
Invalid references and nonmatching quotations are rejected. Reference consistency does not establish source truth, completeness or whether a claim is supported in meaning. No automatic web search or news feed is included. Any AI-assisted preparation uses your own AI client and quota; integration varies by client.
錯誤来源編號或不吻合引文會被拒絕。引用吻合不代表來源真實、完整，也不保證主張的語意獲得支持。不含自動網路搜尋或新聞串流；AI 輔助整理使用買家自己的客戶端與額度，各客戶端整合方式不同。
Price and delivery
費用與交付
US$9 one-time purchase: version 0.1.0 Beta, Skill instructions, readable Python source and synthetic examples. No subscription, hosted service, model credits or manual research service. You provide the computer and pay for optional AI/data services.
US$9 一次買斷：交付 0.1.0 Beta、技能指引、可讀 Python 原碼與合成範例。不含訂閱、託管服務、模型額度或人工研究。買家自備電腦並負擔選用 AI／資料服務費用。
Support: [email protected] — installation questions, suggestions and reproducible bugs; no fixed response time or custom integration. Send synthetic examples and redacted errors, never keys or private documents.
聯絡信箱：[email protected]，受理安裝問題、修改建議與可重現錯誤；不承諾固定回覆時間或客製整合。請提供合成範例與去識別錯誤，不要寄送金鑰或私密文件。
