linkedin-studio/scripts/brain/tests/publish.test.ts

import { describe, test, beforeEach, afterEach } from "node:test";
import assert from "node:assert/strict";
import { mkdtempSync, rmSync, existsSync, readFileSync, writeFileSync, mkdirSync, readdirSync } from "node:fs";
import { join } from "node:path";
import { tmpdir } from "node:os";

import {
  writePublished,
  ingestText,
  scanInbox,
  listPublished,
  serializePublishedRecord,
  parsePublishedRecord,
  type PublishedRecord,
} from "../src/ingest.js";
import { mintContentId } from "../src/id.js";

const CAPTURED = "2026-06-23";

function publishedDir(root: string): string {
  return join(root, "ingest", "published");
}
function inboxDir(root: string): string {
  return join(root, "ingest", "inbox");
}
function mdFiles(dir: string): string[] {
  return existsSync(dir) ? readdirSync(dir).filter((f) => f.endsWith(".md")) : [];
}

describe("ingest IO — writePublished / ingestText / scanInbox / listPublished (SC1/SC3/SC4)", () => {
  let root: string;
  const prevEnv = process.env.LINKEDIN_STUDIO_DATA;

  beforeEach(() => {
    root = mkdtempSync(join(tmpdir(), "brain-ingest-"));
    process.env.LINKEDIN_STUDIO_DATA = root;
  });
  afterEach(() => {
    if (prevEnv === undefined) delete process.env.LINKEDIN_STUDIO_DATA;
    else process.env.LINKEDIN_STUDIO_DATA = prevEnv;
    rmSync(root, { recursive: true, force: true });
  });

  test("create-on-demand: ingestText writes ingest/published/<id>.md with no prior init (SC1)", () => {
    const body = "Min første publiserte post.\n\nMed avsnitt.";
    const res = ingestText({ body, captured_at: CAPTURED });
    assert.equal(res.written, true);
    const path = join(publishedDir(root), `${mintContentId(body)}.md`);
    assert.ok(existsSync(path), "record file created at runtime data-path");
    const rec = parsePublishedRecord(readFileSync(path, "utf8"));
    assert.equal(rec.provenance, "published");
    assert.equal(rec.body, body, "verbatim body preserved");
  });

  test("published_date defaults to captured_at when not supplied", () => {
    const res = ingestText({ body: "no date given", captured_at: CAPTURED });
    assert.equal(res.record.published_date, CAPTURED);
  });

  test("published_date is kept when supplied", () => {
    const res = ingestText({ body: "dated", captured_at: CAPTURED, published_date: "2026-05-26" });
    assert.equal(res.record.published_date, "2026-05-26");
  });

  test("re-ingesting an identical body is a no-op (SC3 dedup): {written:false}, no duplicate file", () => {
    const body = "same exact text";
    ingestText({ body, captured_at: CAPTURED });
    const res2 = ingestText({ body, captured_at: "2026-07-01" }); // different day, same body
    assert.equal(res2.written, false);
    assert.equal(mdFiles(publishedDir(root)).length, 1, "still exactly one record");
  });

  test("id-collision with a DIFFERING body disambiguates to <id>-2.md — never silent data loss (B2)", () => {
    const body = "the real body";
    const id = mintContentId(body);
    // Pre-plant a file at <id>.md whose body is DIFFERENT (simulates a hash collision).
    mkdirSync(publishedDir(root), { recursive: true });
    const squatter: PublishedRecord = {
      id, provenance: "published", published_date: CAPTURED, captured_at: CAPTURED,
      source: "manual", specifics: [], trends: [], body: "a DIFFERENT body that happens to share the id",
    };
    writeFileSync(join(publishedDir(root), `${id}.md`), serializePublishedRecord(squatter), "utf8");

    const res = writePublished({ ...squatter, body });
    assert.equal(res.written, true);
    assert.equal(res.collision, true);
    assert.match(res.path, /-2\.md$/);
    assert.equal(mdFiles(publishedDir(root)).length, 2, "both records kept — no data loss");
  });

  test("scanInbox processes top-level *.md only, skipping dotfiles like .DS_Store (SC4)", () => {
    mkdirSync(inboxDir(root), { recursive: true });
    writeFileSync(join(inboxDir(root), "post-a.md"), "Post A body", "utf8");
    writeFileSync(join(inboxDir(root), "post-b.md"), "Post B body", "utf8");
    writeFileSync(join(inboxDir(root), ".DS_Store"), "junk", "utf8");
    writeFileSync(join(inboxDir(root), "notes.txt"), "not markdown", "utf8");

    const res = scanInbox({ captured_at: CAPTURED });
    assert.equal(res.processed.length, 2, "only the two .md files processed");
    assert.equal(mdFiles(publishedDir(root)).length, 2);
  });

  test("scanInbox on an empty/absent inbox is a clean no-op (SC4)", () => {
    const res = scanInbox({ captured_at: CAPTURED });
    assert.deepEqual(res, { processed: [], skipped: [] });
  });

  test("re-running scanInbox re-skips already-published records (idempotent)", () => {
    mkdirSync(inboxDir(root), { recursive: true });
    writeFileSync(join(inboxDir(root), "post.md"), "once", "utf8");
    scanInbox({ captured_at: CAPTURED });
    const res2 = scanInbox({ captured_at: CAPTURED });
    assert.equal(res2.processed.length, 0);
    assert.equal(res2.skipped.length, 1);
    assert.equal(mdFiles(publishedDir(root)).length, 1, "no duplicate");
  });

  test("listPublished surfaces provenance and skips a malformed file without throwing", () => {
    ingestText({ body: "good record", captured_at: CAPTURED });
    mkdirSync(publishedDir(root), { recursive: true });
    writeFileSync(join(publishedDir(root), "broken.md"), "not a valid record at all", "utf8");

    const res = listPublished();
    assert.equal(res.records.length, 1, "the one valid record listed");
    assert.equal(res.records[0].provenance, "published");
    assert.equal(res.skipped, 1, "the malformed file counted as skipped, not a crash");
  });

  // SC4 — ingestText threads specifics/trends onto the written record; absent → [].
  test("SC4: ingestText threads specifics/trends; round-trips off disk", () => {
    const spec = "aaaaaaaaaaaa";
    const trend = "bbbbbbbbbbbb";
    const res = ingestText({ body: "tagged post body", captured_at: CAPTURED, specifics: [spec], trends: [trend] });
    assert.deepEqual(res.record.specifics, [spec]);
    assert.deepEqual(res.record.trends, [trend]);
    const onDisk = parsePublishedRecord(readFileSync(res.path, "utf8"));
    assert.deepEqual(onDisk.specifics, [spec]);
    assert.deepEqual(onDisk.trends, [trend]);
  });

  test("SC4: ingestText with no specifics/trends keeps them empty (regression)", () => {
    const res = ingestText({ body: "plain post", captured_at: CAPTURED });
    assert.deepEqual(res.record.specifics, []);
    assert.deepEqual(res.record.trends, []);
  });

  // SC3 (producer side) — a malformed id thrown fast at the producer boundary.
  test("SC3: ingestText with a non-12-hex specifics id throws at the producer", () => {
    assert.throws(
      () => ingestText({ body: "x", captured_at: CAPTURED, specifics: ["NOTAHEX"] }),
      /bad specifics id/i,
    );
    assert.throws(
      () => ingestText({ body: "y", captured_at: CAPTURED, trends: ["zzz"] }),
      /bad trends id/i,
    );
  });
});