carthagent 1.0.7 → 1.1.0-preview.12

package/tools/native-glyph-pty.py

@@ -0,0 +1,144 @@
+#!/usr/bin/env python3
+"""Paired real-PTY glyph probe. Only known host chrome is normalized/asserted."""
+import argparse
+import fcntl
+import hashlib
+import json
+import os
+from pathlib import Path
+import pty
+import re
+import select
+import signal
+import struct
+import subprocess
+import tempfile
+import termios
+import time
+import importlib.util
+import sys
+
+_helper_path = Path(__file__).with_name('native-glyph-helpers.py')
+_helper_spec = importlib.util.spec_from_file_location('native_glyph_helpers', _helper_path)
+if _helper_spec is None or _helper_spec.loader is None:
+    raise ImportError(f'Could not load native glyph helper module at {_helper_path}')
+_helpers = importlib.util.module_from_spec(_helper_spec)
+sys.modules[_helper_spec.name] = _helpers
+_helper_spec.loader.exec_module(_helpers)
+CHROME_PAIRS = _helpers.CHROME_PAIRS
+USER_TEXT = _helpers.USER_TEXT
+GlyphScreen = _helpers.GlyphScreen
+assert_composer_payload = _helpers.assert_composer_payload
+child_environment = _helpers.child_environment
+
+
+def run_mode(binary, mode, timeout):
+    with tempfile.TemporaryDirectory(prefix='ctg-glyph-pty-') as temp:
+        root = Path(temp)
+        project = root / 'project'
+        config = root / 'config' / 'opencode'
+        for directory in (project, config, root / 'home', root / 'data', root / 'state', root / 'cache'):
+            directory.mkdir(parents=True, exist_ok=True)
+        (config / 'opencode.json').write_text(json.dumps({'update': 'disable', 'model': 'ctg-glyph/fixture',
+            'providers': {'ctg-glyph': {'name': 'Local glyph probe',
+                'package': '@opencode/ai/providers/openai-compatible',
+                'settings': {'baseURL': 'http://127.0.0.1:1/v1', 'apiKey': 'offline-fixture'},
+                'models': {'fixture': {'name': 'Glyph fixture', 'limit': {'context': 32000, 'output': 1024}}}}}}), encoding='utf8')
+        env = child_environment(root, config, mode)
+        master, slave = pty.openpty()
+        fcntl.ioctl(slave, termios.TIOCSWINSZ, struct.pack('HHHH', 32, 100, 0, 0))
+        child = subprocess.Popen([binary, '--standalone'], cwd=project, env=env,
+                                 stdin=slave, stdout=slave, stderr=slave, start_new_session=True)
+        os.close(slave)
+        screen = GlyphScreen()
+        raw = bytearray()
+        started = time.monotonic()
+        pasted = False
+        home_seen = False
+        try:
+            while time.monotonic() - started < timeout:
+                if child.poll() is not None:
+                    raise RuntimeError(f'{mode}: runtime exited {child.returncode}: {screen.visible()[-1200:]}')
+                if select.select([master], [], [], .05)[0]:
+                    try:
+                        chunk = os.read(master, 65536)
+                    except OSError:
+                        break
+                    raw.extend(chunk)
+                    screen.feed(chunk)
+                    if b'\x1b[6n' in chunk:
+                        os.write(master, b'\x1b[1;1R')
+                    if b'\x1b[c' in chunk:
+                        os.write(master, b'\x1b[?1;2c')
+                visible = screen.visible()
+                lower = visible.lower()
+                if not pasted and ('shift+tab agents' in lower and 'ctrl+p commands' in lower):
+                    home_seen = True
+                    # Insert but do not submit: this makes arbitrary user text
+                    # observable in the genuine composer without contacting a provider.
+                    os.write(master, b'\x1b[200~' + USER_TEXT.encode() + b'\x1b[201~')
+                    pasted = True
+                if pasted and any(re.fullmatch(r'\s*[┃|]\s*' + re.escape(USER_TEXT) + r'\s*', row)
+                                  for row in screen.visible().splitlines()):
+                    break
+            visible = screen.visible()
+            if not home_seen or not pasted:
+                raise AssertionError(f'{mode}: home/composer was not reached; screen={visible[-1600:]!r}')
+            composer_row = assert_composer_payload(visible)
+            # Exclude the rendered composer row from static chrome checks.
+            lines = visible.splitlines()
+            marker_rows = [index for index, line in enumerate(lines) if USER_TEXT in line]
+            chrome_region = '\n'.join(line for index, line in enumerate(lines) if index not in marker_rows)
+            chrome = {}
+            for unicode_token, ascii_token in CHROME_PAIRS:
+                chrome[unicode_token] = {'asciiAlternative': ascii_token,
+                                         'unicodePresent': unicode_token in chrome_region,
+                                         'asciiPresent': ascii_token in chrome_region}
+            if mode == 'ascii':
+                if CHROME_PAIRS[0][0] in chrome_region or '▀▀▀' in chrome_region:
+                    raise AssertionError(f'ASCII startup logo still contains Unicode block-art chrome; screen={visible[-1200:]!r}')
+                if 'CARTHAGENT' not in chrome_region.replace(' ', ''):
+                    raise AssertionError(f'ASCII startup logo alternative not observed; screen={visible[-1200:]!r}')
+            else:
+                if CHROME_PAIRS[0][0] not in chrome_region:
+                    raise AssertionError(f'Unicode startup logo not observed; screen={visible[-1200:]!r}')
+            if 'Carthagent' not in screen.title_updates:
+                raise AssertionError(f'{mode}: Carthagent terminal title not observed: {screen.title_updates!r}')
+            if any('opencode' in title.lower() for title in screen.title_updates):
+                raise AssertionError(f'{mode}: upstream terminal title remains: {screen.title_updates!r}')
+            if 'opencode' in chrome_region.lower():
+                raise AssertionError(f'{mode}: upstream branding remains on the home screen')
+            return {'mode': mode, 'homeObserved': home_seen, 'terminalTitles': screen.title_updates, 'composerPayloadExact': True,
+                    'composerRow': composer_row,
+                    'userText': USER_TEXT, 'hostChrome': chrome, 'screen': visible,
+                    'ptyOutputSha256': hashlib.sha256(raw).hexdigest(), 'ptyOutputBytes': len(raw)}
+        finally:
+            if child.poll() is None:
+                os.write(master, b'\x1b')
+                os.killpg(child.pid, signal.SIGTERM)
+                try:
+                    child.wait(timeout=3)
+                except subprocess.TimeoutExpired:
+                    os.killpg(child.pid, signal.SIGKILL)
+                    child.wait()
+            os.close(master)
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--binary', required=True)
+    parser.add_argument('--timeout', type=int, default=45)
+    parser.add_argument('--artifact')
+    args = parser.parse_args()
+    modes = [run_mode(args.binary, mode, args.timeout) for mode in ('ascii', 'unicode')]
+    result = {'status': 'PASS', 'claim': 'paired actual-TUI PTY rendering; only known static chrome is asserted',
+              'provider': 'none; prompt is staged but not submitted', 'physicalClipboardVerified': False,
+              'modes': modes}
+    text = json.dumps(result, ensure_ascii=False, indent=2) + '\n'
+    print(text, end='')
+    if args.artifact:
+        Path(args.artifact).resolve().write_text(text, encoding='utf8')
+
+
+if __name__ == '__main__':
+    main()

package/tools/native-manual-projects.mjs

@@ -0,0 +1,259 @@
+import { mkdir, writeFile, readdir, readFile, rm } from 'node:fs/promises';
+import { createHash } from 'node:crypto';
+import { execFileSync } from 'node:child_process';
+import { dirname, join, resolve } from 'node:path';
+import { pathToFileURL } from 'node:url';
+
+export async function createManualProjects(
+  destination,
+  { packages = 128 } = {},
+) {
+  if (!Number.isInteger(packages) || packages < 2 || packages > 1024)
+    throw Error('packages must be an integer from 2 to 1024');
+  const root = resolve(destination);
+  await mkdir(dirname(root), { recursive: true });
+  await mkdir(root);
+  const write = async (project, file, contents) => {
+    const path = join(root, project, file);
+    await mkdir(dirname(path), { recursive: true });
+    await writeFile(path, contents);
+  };
+  const packageFile = (name, extra = {}) =>
+    JSON.stringify(
+      {
+        name,
+        version: '0.0.0',
+        private: true,
+        type: 'module',
+        scripts: {
+          test: 'node --test test/*.test.mjs',
+          'test:acceptance': 'node --test acceptance/*.test.mjs',
+          ...extra,
+        },
+      },
+      null,
+      2,
+    ) + '\n';
+  const imports =
+    "import {test} from 'node:test';\nimport assert from 'node:assert/strict';\n";
+  try {
+    await write('tiny', 'package.json', packageFile('ctg-invoice-fixture'));
+    await write(
+      'tiny',
+      'src/invoice.mjs',
+      'export function invoiceTotal(lines) { return lines.reduce((total, line) => total + line.quantity * line.unitPrice, 0); }\n',
+    );
+    await write(
+      'tiny',
+      'test/invoice.test.mjs',
+      imports +
+        "import {invoiceTotal} from '../src/invoice.mjs';\ntest('empty invoices',()=>assert.equal(invoiceTotal([]),0));\ntest('integer totals',()=>assert.equal(invoiceTotal([{quantity:2,unitPrice:500}]),1000));\n",
+    );
+    await write(
+      'tiny',
+      'acceptance/discount.test.mjs',
+      imports +
+        "import {invoiceTotal} from '../src/invoice.mjs';\ntest('discount in basis points, rounded once to cents',()=>assert.equal(invoiceTotal([{quantity:3,unitPrice:199}],{discountBps:1500}),507));\ntest('reject invalid discounts',()=>assert.throws(()=>invoiceTotal([],{discountBps:10001}),RangeError));\ntest('immutable inputs',()=>assert.equal(invoiceTotal(Object.freeze([Object.freeze({quantity:1,unitPrice:999})]),{discountBps:1000}),899));\n",
+    );
+    await write(
+      'tiny',
+      'TASK.md',
+      '# Small task\nAdd optional {discountBps} to invoiceTotal. Prices/totals are integer cents. Accept integer discounts 0..10000; reject others with RangeError. Round once after discount. Preserve old calls and never mutate inputs. Run npm test and npm run test:acceptance. Baseline is green; new-feature acceptance is intentionally red.\n',
+    );
+
+    await write(
+      'product',
+      'package.json',
+      packageFile('ctg-forgeboard-fixture', { start: 'node server.mjs' }),
+    );
+    await write(
+      'product',
+      'src/issues.mjs',
+      "export const initialIssues=()=>[{id:1,title:'Design onboarding',status:'todo',revision:1},{id:2,title:'Fix keyboard navigation',status:'doing',revision:1}];\nexport const selectIssues=(issues,status)=>issues.filter(issue=>!status||issue.status===status);\n",
+    );
+    await write(
+      'product',
+      'server.mjs',
+      `import {createServer} from 'node:http';
+import {readFile} from 'node:fs/promises';
+import {pathToFileURL} from 'node:url';
+import {resolve} from 'node:path';
+import {initialIssues,selectIssues} from './src/issues.mjs';
+export function createApp(){
+ const issues=initialIssues();
+ return createServer(async(request,response)=>{
+  const url=new URL(request.url,'http://localhost');
+  if(request.method==='GET'&&url.pathname==='/api/issues'){response.writeHead(200,{'content-type':'application/json'});response.end(JSON.stringify(selectIssues(issues,url.searchParams.get('status'))));return;}
+  if(request.method==='GET'&&url.pathname==='/'){response.writeHead(200,{'content-type':'text/html; charset=utf-8'});response.end(await readFile(new URL('./public/index.html',import.meta.url)));return;}
+  response.writeHead(404,{'content-type':'application/json'});response.end(JSON.stringify({error:'not found'}));
+ });
+}
+if(process.argv[1]&&import.meta.url===pathToFileURL(resolve(process.argv[1])).href){const server=createApp();server.listen(Number(process.env.PORT??3017),'127.0.0.1',()=>console.log('Forgeboard: http://127.0.0.1:'+server.address().port));}
+`,
+    );
+    await write(
+      'product',
+      'public/index.html',
+      '<!doctype html><html lang="en"><meta charset="utf-8"><title>Forgeboard</title><style>body{font:18px system-ui;background:#f4f1e9;color:#202923;max-width:900px;margin:64px auto;padding:24px}header{border-bottom:2px solid #202923}li{background:white;padding:20px;margin:12px 0;list-style:none}select{font:inherit;padding:8px}</style><header><small>CARTHAGENT TEST PROJECT / 02</small><h1>Forgeboard</h1><p>A disposable issue board. No external services.</p></header><p><label>Status <select id="status"><option value="">All</option><option>todo</option><option>doing</option><option>done</option></select></label></p><ul id="issues"></ul><script type="module">async function load(){const issues=await(await fetch("/api/issues?status="+encodeURIComponent(document.querySelector("select").value))).json();document.querySelector("ul").replaceChildren(...issues.map(issue=>{const li=document.createElement("li");li.textContent=issue.title+" · "+issue.status+" · revision "+issue.revision;return li}));}document.querySelector("select").onchange=load;load();</script></html>\n',
+    );
+    const apiTest =
+      "import {createApp} from '../server.mjs';\nasync function appFor(t){const app=createApp();await new Promise(r=>app.listen(0,'127.0.0.1',r));t.after(()=>new Promise(r=>app.close(r)));return 'http://127.0.0.1:'+app.address().port;}\n";
+    await write(
+      'product',
+      'test/api.test.mjs',
+      imports +
+        apiTest +
+        "test('real HTTP list, filtering, UI and 404',async t=>{const base=await appFor(t);assert.equal((await(await fetch(base+'/api/issues')).json()).length,2);assert.equal((await(await fetch(base+'/api/issues?status=doing')).json())[0].id,2);assert.match(await(await fetch(base)).text(),/Forgeboard/);assert.equal((await fetch(base+'/missing')).status,404);});\n",
+    );
+    await write(
+      'product',
+      'acceptance/update.test.mjs',
+      imports +
+        apiTest +
+        "test('optimistic concurrency protects competing status updates',async t=>{const base=await appFor(t);const patch=(revision,status)=>fetch(base+'/api/issues/1',{method:'PATCH',headers:{'content-type':'application/json'},body:JSON.stringify({revision,status})});const first=await patch(1,'done');assert.equal(first.status,200);assert.equal((await first.json()).revision,2);assert.equal((await patch(1,'doing')).status,409);assert.equal((await patch(2,'invalid')).status,400);const list=await(await fetch(base+'/api/issues?status=done')).json();assert.equal(list[0].id,1);assert.equal(list[0].status,'done');});\n",
+    );
+    await write(
+      'product',
+      'TASK.md',
+      '# Product task\nImplement PATCH /api/issues/:id accepting JSON {status,revision}. Allow todo/doing/done. Stale revisions ->409; invalid input/JSON ->400; unknown ids ->404; oversized bodies ->413. Increment revision exactly once on success. Add accessible status controls and error feedback to the board. Keep in-memory storage; no external services/dependencies. Preserve safe text rendering. Add malformed/oversized/unknown-id tests beyond supplied acceptance. Run npm test and npm run test:acceptance. Start npm start and inspect http://127.0.0.1:3017 manually.\n',
+    );
+
+    await write(
+      'scale',
+      'package.json',
+      packageFile('ctg-scale-monorepo', {
+        'test:stress': 'node --test --test-concurrency=2 test/*.test.mjs',
+      }),
+    );
+    await write(
+      'scale',
+      'shared/money.mjs',
+      'export function roundMoney(value){return Math.round(value*100)/100;}\n',
+    );
+    await write(
+      'scale',
+      'test/packages.test.mjs',
+      imports +
+        Array.from(
+          { length: packages },
+          (_, i) =>
+            `test('domain-${i} integration',async()=>{const {total}=await import('../packages/domain-${i}/src/index.mjs');assert.equal(total([1,2,3]),6);assert.equal(total([]),0);});`,
+        ).join('\n') +
+        '\n',
+    );
+    for (let i = 0; i < packages; i++) {
+      await write(
+        'scale',
+        `packages/domain-${i}/package.json`,
+        JSON.stringify({
+          name: `@fixture/domain-${i}`,
+          private: true,
+          type: 'module',
+        }) + '\n',
+      );
+      await write(
+        'scale',
+        `packages/domain-${i}/src/index.mjs`,
+        "import {roundMoney} from '../../../shared/money.mjs';\nexport const total=values=>roundMoney(values.reduce((sum,value)=>sum+value,0));\n",
+      );
+      for (let j = 0; j < 8; j++)
+        await write(
+          'scale',
+          `packages/domain-${i}/src/metric-${j}.mjs`,
+          `import {total} from './index.mjs';\nexport const metric${j}=values=>({domain:${i},metric:${j},total:total(values)});\n`,
+        );
+      await write(
+        'scale',
+        `docs/domain-${i}.md`,
+        `# Domain ${i}\nUses shared/money.mjs through its public index. Preserve domain identity.\n`,
+      );
+    }
+    await write(
+      'scale',
+      'acceptance/rounding.test.mjs',
+      imports +
+        "import {roundMoney} from '../shared/money.mjs';\nimport {total} from '../packages/domain-0/src/index.mjs';\ntest('decimal halves away from zero across shared consumers',()=>{assert.equal(roundMoney(1.005),1.01);assert.equal(roundMoney(-1.005),-1.01);assert.equal(total([1.005]),1.01);});\ntest('reject nonfinite amounts',()=>{for(const v of [NaN,Infinity,-Infinity])assert.throws(()=>roundMoney(v),RangeError);});\n",
+    );
+    await write(
+      'scale',
+      'TASK.md',
+      `# Large monorepo task\n${packages} packages share money rounding; do not rewrite generated packages. Fix representable decimal halves (1.005 ->1.01, -1.005 ->-1.01, ties away from zero); reject nonfinite numbers with RangeError. Preserve existing behavior, avoid dependencies, and add boundary regressions. Run npm test and npm run test:acceptance. Use code navigation to find affected consumers and verify at least three distinct domains.\n`,
+    );
+    const inventory = {};
+    for (const name of ['tiny', 'product', 'scale']) {
+      await write(name, '.gitignore', 'node_modules/\n.harness/\n.env\n');
+      const files = [];
+      const visit = async (directory) => {
+        for (const entry of (
+          await readdir(join(root, name, directory), { withFileTypes: true })
+        ).sort((a, b) => a.name.localeCompare(b.name))) {
+          const file = directory ? `${directory}/${entry.name}` : entry.name;
+          if (entry.isDirectory()) await visit(file);
+          else
+            files.push({
+              path: file,
+              sha256: createHash('sha256')
+                .update(await readFile(join(root, name, file)))
+                .digest('hex'),
+            });
+        }
+      };
+      await visit('');
+      inventory[name] = { files: files.length, source: files };
+      const git = (...args) =>
+        execFileSync('git', ['-c', 'core.hooksPath=disabled-hooks', ...args], {
+          cwd: join(root, name),
+          stdio: 'pipe',
+        });
+      git('init', '-b', 'main');
+      git('add', '.');
+      git(
+        '-c',
+        'user.name=CTG manual fixture',
+        '-c',
+        'user.email=fixture@localhost',
+        'commit',
+        '-m',
+        'Initialize disposable manual-test project\n\nCo-authored-by: openhands <openhands@all-hands.dev>',
+      );
+    }
+    await writeFile(
+      join(root, 'inventory.json'),
+      JSON.stringify({ version: 1, packages, projects: inventory }, null, 2) +
+        '\n',
+    );
+    return {
+      root,
+      packages,
+      projects: Object.fromEntries(
+        Object.entries(inventory).map(([name, value]) => [
+          name,
+          { files: value.files },
+        ]),
+      ),
+    };
+  } catch (error) {
+    await rm(root, { recursive: true, force: true });
+    throw error;
+  }
+}
+
+if (
+  process.argv[1] &&
+  import.meta.url === pathToFileURL(resolve(process.argv[1])).href
+) {
+  const [destination, packages] = process.argv.slice(2);
+  if (!destination)
+    throw Error(
+      'Usage: node tools/native-manual-projects.mjs <new-directory> [packages=128]',
+    );
+  console.log(
+    JSON.stringify(
+      await createManualProjects(destination, {
+        packages: Number(packages ?? 128),
+      }),
+      null,
+      2,
+    ),
+  );
+}