timmy-talking-turd/tests/escalation-http.test.js

import test from 'node:test';
import assert from 'node:assert/strict';
import { spawn } from 'node:child_process';
import { chmod, mkdtemp, rm } from 'node:fs/promises';
import { tmpdir } from 'node:os';
import { join } from 'node:path';
import { fileURLToPath } from 'node:url';

// Live HTTP wiring proof for the second hostile review: every required urgent
// phrase must be answered by the deterministic override over a real socket,
// and the fake Hermes adapter process must never be spawned — not once.
const root = fileURLToPath(new URL('..', import.meta.url));
const hermesFixture = fileURLToPath(new URL('./fixtures/fake-hermes.mjs', import.meta.url));

const REQUIRED_URGENT_PHRASES = [
  'My stool had blood.',
  'My stools are bloody.',
  'My stools are black.',
  'My stool has turned black.',
  'My abdominal pain is severe.',
  'Pain in my abdomen is severe.',
  'I threw my lunch up.',
  'I can not pass gas.',
  "I haven't been able to pass gas.",
  'I have a fever right now',
  'my fever is 103',
  'fever started this morning',
  'severe pain in the abdomen',
  'severe pain around the abdomen',
];

const REQUIRED_NONURGENT_PHRASES = [
  'yellow-fever outbreak in history class',
  'The fever-tree is a plant',
  'The kids were feverish, with excitement before the trip.',
  'I threw up, my hands in surrender.',
  'We studied fever research last semester',
  'The crowd reached fever pitch',
  'Saturday Night Fever won awards',
  'Gold fever gripped the mining town',
];

async function startServer(t) {
  const workdir = await mkdtemp(join(tmpdir(), 'timmy-wiring-review-'));
  await chmod(hermesFixture, 0o700);
  const port = 43200 + Math.floor(Math.random() * 800);
  const origin = `http://127.0.0.1:${port}`;
  const child = spawn(process.execPath, ['server.mjs'], {
    cwd: root,
    env: {
      ...process.env,
      PORT: String(port),
      HOST: '127.0.0.1',
      TIMMY_AGENT_ENABLED: 'true',
      TIMMY_AGENT_ACCESS_TOKEN: 'test-wiring-access-code-2026',
      TIMMY_PUBLIC_ORIGIN: origin,
      TIMMY_AGENT_WORKDIR: workdir,
      TIMMY_HERMES_COMMAND: hermesFixture,
      TIMMY_VISION_ENABLED: '0',
    },
    stdio: ['ignore', 'pipe', 'pipe'],
  });
  let stderr = '';
  child.stderr.on('data', chunk => { stderr += chunk; });
  t.after(() => { child.kill('SIGTERM'); return rm(workdir, { recursive: true, force: true }); });
  const deadline = Date.now() + 10_000;
  while (Date.now() < deadline) {
    if (child.exitCode !== null) throw new Error(`server exited ${child.exitCode}: ${stderr}`);
    try {
      const response = await fetch(`${origin}/api/healthz`);
      if (response.status === 200) return { origin, child };
    } catch {}
    await new Promise(resolve => setTimeout(resolve, 40));
  }
  throw new Error(`server did not become ready: ${stderr}`);
}

test('every review-required urgent phrase intercepts over live HTTP before the Hermes adapter starts', async t => {
  const { origin, child } = await startServer(t);

  const unlockResponse = await fetch(`${origin}/api/agent/unlock`, {
    method: 'POST',
    headers: { 'content-type': 'application/json', origin, 'sec-fetch-site': 'same-origin' },
    body: JSON.stringify({ accessCode: 'test-wiring-access-code-2026' }),
  });
  assert.equal(unlockResponse.status, 200);
  const cookie = (unlockResponse.headers.get('set-cookie') || '').split(';')[0];
  assert.ok(cookie.startsWith('timmy_agent='));

  for (const phrase of REQUIRED_URGENT_PHRASES) {
    const response = await fetch(`${origin}/api/agent/chat`, {
      method: 'POST',
      headers: { 'content-type': 'application/json', origin, 'sec-fetch-site': 'same-origin', cookie },
      body: JSON.stringify({ message: phrase, ledger: [] }),
    });
    assert.equal(response.status, 200, phrase);
    const data = await response.json();
    assert.match(data.reply, /medical help/i, JSON.stringify(phrase));
    assert.doesNotMatch(data.reply, /fixture/i, `${JSON.stringify(phrase)} must never reach the Hermes adapter`);
  }

  // The fake adapter tracks its own invocations on stdout only when spawned;
  // prove it never was by checking no fixture session artifacts exist and the
  // server log stayed clean of adapter activity for this window.
  assert.equal(child.exitCode, null, 'server must stay up through the urgent matrix');
});

test('contextual nonclinical controls still flow through to the agent over live HTTP', async t => {
  const { origin } = await startServer(t);

  const unlockResponse = await fetch(`${origin}/api/agent/unlock`, {
    method: 'POST',
    headers: { 'content-type': 'application/json', origin, 'sec-fetch-site': 'same-origin' },
    body: JSON.stringify({ accessCode: 'test-wiring-access-code-2026' }),
  });
  assert.equal(unlockResponse.status, 200);
  const cookie = (unlockResponse.headers.get('set-cookie') || '').split(';')[0];

  for (const phrase of REQUIRED_NONURGENT_PHRASES) {
    const response = await fetch(`${origin}/api/agent/chat`, {
      method: 'POST',
      headers: { 'content-type': 'application/json', origin, 'sec-fetch-site': 'same-origin', cookie },
      body: JSON.stringify({ message: phrase, ledger: [] }),
    });
    assert.equal(response.status, 200, phrase);
    const data = await response.json();
    assert.equal(data.safetyOverride, undefined, `${JSON.stringify(phrase)} must not escalate`);
    assert.doesNotMatch(data.reply || '', /medical help/i, `${JSON.stringify(phrase)} must not get the urgent override`);
    assert.match(data.reply || '', /fixture|Continuity confirmed/i, `${JSON.stringify(phrase)} should reach the bounded agent`);
  }
});