ttrss-mcp

Instructions

Implementation Reference

• src/tools/report.ts:19-51 (handler)

  The handler function for the `generate_daily_report` tool, which fetches headlines, filters them, and formats a report.

  async (params) => {
    try {
      const cutoffTime = Math.floor(Date.now() / 1000) - params.hours * 3600;
  
      const headlines = await client.getHeadlines({
        feed_id: -4,
        limit: params.limit,
        view_mode: params.view_mode,
        show_excerpt: params.include_content,
        order_by: "feed_dates",
      });
  
      const filtered = headlines.filter((h) => h.updated >= cutoffTime);
  
      if (filtered.length === 0) {
        return {
          content: [{
            type: "text" as const,
            text: `过去 ${params.hours} 小时内没有${params.view_mode === "unread" ? "未读" : ""}文章。`,
          }],
        };
      }
  
      const grouped = groupByFeed(filtered);
      const report = formatReport(grouped, params.hours, params.include_content);
      return { content: [{ type: "text" as const, text: report }] };
    } catch (e: unknown) {
      return {
        content: [{ type: "text" as const, text: `生成日报失败: ${(e as Error).message}` }],
        isError: true,
      };
    }
  },

• src/tools/report.ts:7-52 (registration)

  Registration of the `generate_daily_report` tool using `server.tool`.

  server.tool(
    "generate_daily_report",
    "生成日报：汇总近期文章，按分类和订阅源分组输出。默认获取未读文章，可指定时间范围。",
    {
      hours: z.number().default(24).describe("获取最近 N 小时内的文章，默认 24"),
      limit: z.number().default(200).describe("最多获取的文章数量"),
      view_mode: z
        .enum(["all_articles", "unread", "adaptive", "marked", "updated"])
        .default("all_articles")
        .describe("过滤模式"),
      include_content: z.boolean().default(false).describe("是否包含文章摘要"),
    },
    async (params) => {
      try {
        const cutoffTime = Math.floor(Date.now() / 1000) - params.hours * 3600;
  
        const headlines = await client.getHeadlines({
          feed_id: -4,
          limit: params.limit,
          view_mode: params.view_mode,
          show_excerpt: params.include_content,
          order_by: "feed_dates",
        });
  
        const filtered = headlines.filter((h) => h.updated >= cutoffTime);
  
        if (filtered.length === 0) {
          return {
            content: [{
              type: "text" as const,
              text: `过去 ${params.hours} 小时内没有${params.view_mode === "unread" ? "未读" : ""}文章。`,
            }],
          };
        }
  
        const grouped = groupByFeed(filtered);
        const report = formatReport(grouped, params.hours, params.include_content);
        return { content: [{ type: "text" as const, text: report }] };
      } catch (e: unknown) {
        return {
          content: [{ type: "text" as const, text: `生成日报失败: ${(e as Error).message}` }],
          isError: true,
        };
      }
    },
  );

• src/tools/report.ts:10-18 (schema)

  Zod schema definition for the `generate_daily_report` tool inputs.

  {
    hours: z.number().default(24).describe("获取最近 N 小时内的文章，默认 24"),
    limit: z.number().default(200).describe("最多获取的文章数量"),
    view_mode: z
      .enum(["all_articles", "unread", "adaptive", "marked", "updated"])
      .default("all_articles")
      .describe("过滤模式"),
    include_content: z.boolean().default(false).describe("是否包含文章摘要"),
  },

• src/tools/report.ts:55-124 (helper)

  Helper functions for grouping articles and formatting the final report string.

  interface GroupedArticles {
    [feedTitle: string]: TtrssHeadline[];
  }
  
  function groupByFeed(headlines: TtrssHeadline[]): GroupedArticles {
    const groups: GroupedArticles = {};
    for (const h of headlines) {
      const key = h.feed_title || "未知来源";
      if (!groups[key]) groups[key] = [];
      groups[key].push(h);
    }
    return groups;
  }
  
  function formatReport(
    grouped: GroupedArticles,
    hours: number,
    includeContent: boolean,
  ): string {
    const totalArticles = Object.values(grouped).reduce(
      (sum, arr) => sum + arr.length,
      0,
    );
    const feedCount = Object.keys(grouped).length;
  
    const lines: string[] = [
      `# 日报 (过去 ${hours} 小时)`,
      "",
      `共 ${totalArticles} 篇文章，来自 ${feedCount} 个订阅源。`,
      "",
    ];
  
    const sortedFeeds = Object.entries(grouped).sort(
      ([, a], [, b]) => b.length - a.length,
    );
  
    for (const [feedTitle, articles] of sortedFeeds) {
      lines.push(`## ${feedTitle} (${articles.length} 篇)`);
      lines.push("");
  
      const sorted = articles.sort((a, b) => b.updated - a.updated);
  
      for (const article of sorted) {
        const time = new Date(article.updated * 1000).toLocaleString("zh-CN", {
          timeZone: "Asia/Shanghai",
          month: "2-digit",
          day: "2-digit",
          hour: "2-digit",
          minute: "2-digit",
        });
        const flags = [
          article.unread ? "未读" : "",
          article.marked ? "⭐" : "",
        ]
          .filter(Boolean)
          .join(" ");
  
        lines.push(
          `- [${article.title}](${article.link}) | ${time}${flags ? " | " + flags : ""}`,
        );
  
        if (includeContent && article.excerpt) {
          lines.push(`  > ${article.excerpt.slice(0, 150).replace(/\n/g, " ")}`);
        }
      }
      lines.push("");
    }
  
    return lines.join("\n");
  }