LLM Channel summary Slackbot

slack.ts

import { App, } from '@slack/bolt';
import { Anthropic } from '@anthropic-ai/sdk';

const app = new App({
  signingSecret: process.env.SLACK_SIGNING_SECRET,
  token: process.env.SLACK_BOT_TOKEN,
});

const anthropic = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY });

(async () => {
  app.command('/summary', async ({ ack, respond, command, client }) => {
    await ack();

    const messages = await client.conversations.history({
      channel: command.channel_id,
      limit: 999,
      latest: Math.floor(Date.now() / 1000) - 24 * 60 * 60;
    }).map((message) => `${message.username}: ${message.text}`).join('\n\n');

    const llmResponse = await anthropic.messages.create({
      model: 'claude-3-sonnet-20240229',
      max_tokens: 4096,
      system: `
        You are a personal assistant whose job it is to summarize activity Slack channels.
        You will read a transcript provided and then identify the most important conversations that occurred.`,
      messages: [{ role: 'user', content: 'Here is the transcript for your task:\n\n' + messages }],
    });
    
    await respond({
      blocks: [{
        type: 'section',
        text: {
          type: 'mrkdwn',
          text: 'Channel Summary for the Last Day',
        },
      }, {
        type: 'section',
        text: {
          type: 'mrkdwn',
          text: llmResponse.messages[0].content,
        },
      }],
    });
  });

  await app.start(3000);
  console.log('⚡️ Slackbot is running!');
})();