A code snippet to look up the affiliations of >100 HCI researchers with the help of Tavily and OpenAI Platform

search-affiliations.js

import OpenAI from "openai";
import { tavily } from "@tavily/core";
import { readFileSync, writeFileSync, statSync } from "fs";

const tavilyApiKey = process.env.TAVILY_API_KEY;
const openaiApiKey = process.env.OPENAI_API_KEY;
const model = "gpt-4o"; // "gpt-4o-mini"
const names = readFileSync("list.txt", { encoding: "utf8" }).split(/\r?\n/);
const title = "HCI researcher";
const wait = 1000;

const tavilyClient = tavily({ apiKey: tavilyApiKey });
const openai = new OpenAI({ apiKey: openaiApiKey });

let answers = [];
for (let name of names) {
  const res = await search(name, wait);
  answers.push(res);
}
writeFileSync(`result-${model}.txt`, answers.join("\n"), { encoding: "utf8" });

/**
 * Search for the current affiliations.
 *
 * @param {string} name
 * @param {number} wait
 * @returns
 */
async function search(name, wait) {
  // conduct a web search or load content from cache
  const stat = statSync(`search/${name}.txt`);
  const content = await (stat.isFile() ? loadFromCache(name) : buildFromWeb(name));

  // extract affiliation(s)
  const completion = await openai.chat.completions.create({
    model,
    messages: [
      {
        role: "system",
        content:
          "You are a human resource database that knows who belongs to which organization. When you return the affiliation information, rely on the supplemental information the user provides.",
      },
      {
        role: "user",
        content,
      },
    ],
  });

  // write results
  const affiliation = completion.choices[0].message.content;
  const answer = `${name}: ${affiliation}`;
  const text = `Q. ${content}\n\nA. ${answer}`;
  writeFileSync(`${model}/${name}.txt`, text, { encoding: "utf8" });

  // wait for specified time
  await new Promise((r) => setTimeout(r, wait));

  // return "name: affiliation(s)"
  return answer;
}

async function loadFromCache(name) {
  return readFileSync(`search/${name}.txt`, { encoding: "utf8" });
}

async function buildFromWeb(name) {
  const q = `What are the current affiliation(s) of the ${title} named ${name}?`;
  const response = await tavilyClient.search(q);
  const content = `${q} Return only the affiliations (organization names) as plain text. When there are multiple organizations, split them by commas. When the organization names have shorthands, postpend them to the full names in braces.\n\nFor your information, the latest web search returns the following information:\n\n${response.results
    .map(
      (res, no) =>
        `[Result ${no + 1} (confidence: ${res.score})]\n- url: ${
          res.url
        }\n- title: ${res.title}\n- content: ${res.content}`
    )
    .join("\n\n")}`;
  writeFileSync(`search/${name}.txt`, content, { encoding: "utf8" });
  return content;
}