@sanqianx/project-knowledge 5.0.24 → 5.0.26

package/_site/_test/helpers/parallel-engine-final-harness.js

@@ -0,0 +1,314 @@
+'use strict';
+
+// P-ENG-FINAL lane-exclusive harness (parallel-20261023 batch, engine-final).
+//
+// Self-contained: this lane does NOT require the P-ENG-ADVANCED helpers (those
+// are frozen read-only for the Leader). Everything the drivers need is real:
+//
+//   * real fixture Git repositories (isolated git config, repo-local only)
+//   * a REAL post-commit hook (separate OS process running the real
+//     ai-coding-event-bridge commit-boundary CLI from the pinned package)
+//   * the real ai-coding-event-bridge console server (in-process, isolated
+//     --home journal; pinned @sanqianx package by default, source checkout
+//     only via PK_TEST_BRIDGE_ROOT)
+//   * the real vector-hub server (in-process, isolated root + deterministic
+//     test embeddings: the paid embedding API is NOT exercised; pinned
+//     @sanqianx/vector-hub package by default, source checkout only via
+//     PK_TEST_VECTOR_ROOT)
+//   * the real VENDORED knowledge-engine (in-process, isolated data dir)
+//
+// Isolation: lane temp root at PK_TEST_TEMP_ROOT or os.tmpdir() (never inside
+// the repo), ephemeral 127.0.0.1 ports,
+// scrubbed child env (no KB_*/KNOWLEDGE_*/ANTHROPIC_*/CLAUDE*/GLM/MINIMAX
+// leakage), isolated HOME/USERPROFILE/APPDATA/TEMP, GIT_CONFIG_NOSYSTEM=1 and
+// GIT_CONFIG_GLOBAL pointed at a non-existent file (the user's global git
+// config and hooks are never read or written).
+
+const { spawnSync } = require('node:child_process');
+const fs = require('node:fs');
+const http = require('node:http');
+const os = require('node:os');
+const path = require('node:path');
+
+const REPO_ROOT = path.resolve(__dirname, '..', '..', '..');
+const VENDORED_ENGINE_SERVER = path.join(REPO_ROOT, '_modules', 'knowledge-engine', 'src', 'server.js');
+const VENDORED_ATOMIC_FILE = path.join(REPO_ROOT, '_modules', 'knowledge-engine', 'src', 'runtime', 'atomic-file.js');
+
+// Pinned runtime by default: the root repo's installed
+// versions declared by the root package (identity-checked below), so the drivers are portable to any checkout of this
+// repo. A sibling SOURCE checkout may be substituted only through the explicit
+// env overrides PK_TEST_BRIDGE_ROOT / PK_TEST_VECTOR_ROOT.
+function resolveRuntimeRoot(envVar, pinnedDir, pinnedVersion) {
+  const fromEnv = process.env[envVar];
+  const root = path.resolve(fromEnv || pinnedDir);
+  if (!fs.existsSync(root)) throw new Error(`runtime root ${root} does not exist (${fromEnv ? `${envVar} override` : `pinned ${pinnedVersion} install`} is missing)`);
+  const manifest = JSON.parse(fs.readFileSync(path.join(root, 'package.json'), 'utf8'));
+  if (!fromEnv && manifest.version !== pinnedVersion) {
+    throw new Error(`pinned runtime at ${root} is ${manifest.version}, expected ${pinnedVersion}`);
+  }
+  return { root, version: manifest.version, origin: fromEnv ? `env ${envVar}` : 'pinned package' };
+}
+
+function declaredModuleVersion(name) {
+  const version = JSON.parse(fs.readFileSync(path.join(REPO_ROOT, 'package.json'), 'utf8')).dependencies[name];
+  if (!/^\d+\.\d+\.\d+(?:-rc\.\d+)?$/.test(version || '')) throw new Error(`exact module dependency required: ${name}`);
+  return version;
+}
+const BRIDGE = resolveRuntimeRoot('PK_TEST_BRIDGE_ROOT', path.join(REPO_ROOT, 'node_modules', '@sanqianx', 'ai-coding-event-bridge-console'), declaredModuleVersion('@sanqianx/ai-coding-event-bridge-console'));
+const VECTOR = resolveRuntimeRoot('PK_TEST_VECTOR_ROOT', path.join(REPO_ROOT, 'node_modules', '@sanqianx', 'vector-hub'), declaredModuleVersion('@sanqianx/vector-hub'));
+const BRIDGE_ROOT = BRIDGE.root;
+const BRIDGE_CLI = path.join(BRIDGE_ROOT, 'src', 'bridge', 'bin', 'commit-boundary.js');
+const VECTOR_ROOT = VECTOR.root;
+
+// Temp root: PK_TEST_TEMP_ROOT or the OS temp dir; never inside the repository.
+const TEMP_ROOT = (() => {
+  const base = path.resolve(process.env.PK_TEST_TEMP_ROOT || path.join(os.tmpdir(), 'project-knowledge-engine-test'));
+  if (base === REPO_ROOT || base.startsWith(REPO_ROOT + path.sep)) {
+    throw new Error(`temp root must live outside the repository (got ${base})`);
+  }
+  return base;
+})();
+
+function ensureTempRoot() {
+  fs.mkdirSync(TEMP_ROOT, { recursive: true });
+  return TEMP_ROOT;
+}
+
+function makeLaneDir(label) {
+  ensureTempRoot();
+  return fs.mkdtempSync(path.join(TEMP_ROOT, `${label}-`));
+}
+
+// Scrubbed environment: no profile/quota/credential leakage, isolated HOME and
+// TEMP, isolated git config. Nothing here relies on the caller unsetting vars.
+function isolatedEnv(extra = {}) {
+  const home = extra.ISOLATED_HOME || path.join(ensureTempRoot(), 'isolated-home');
+  fs.mkdirSync(home, { recursive: true });
+  const env = { ...process.env };
+  for (const name of Object.keys(env)) {
+    if (/^(KB_|KNOWLEDGE_|ANTHROPIC_|CLAUDE|OPENAI_|CODEX|MINIMAX|GLM|AI_CODING_EVENT_BRIDGE|PK_TEST_)/i.test(name)) delete env[name];
+  }
+  const tmp = path.join(ensureTempRoot(), 'tmp');
+  fs.mkdirSync(tmp, { recursive: true });
+  env.HOME = home;
+  env.USERPROFILE = home;
+  env.APPDATA = path.join(home, 'AppData', 'Roaming');
+  env.LOCALAPPDATA = path.join(home, 'AppData', 'Local');
+  env.TEMP = tmp;
+  env.TMP = tmp;
+  env.TMPDIR = tmp;
+  env.GIT_CONFIG_NOSYSTEM = '1';
+  env.GIT_CONFIG_GLOBAL = path.join(home, 'does-not-exist.gitconfig');
+  env.AI_CODING_EVENT_BRIDGE_HOME = path.join(home, 'bridge-home');
+  env.BRIDGE_CONSOLE_PICK_FOLDER = '0';
+  return { ...env, ...extra };
+}
+
+// git with the isolated config; returns {code, stdout, stderr} so a driver can
+// assert the real exit code of a commit whose hook failed.
+function gitRun(repoPath, args) {
+  const result = spawnSync('git', args, { cwd: repoPath, windowsHide: true, encoding: 'utf8', env: isolatedEnv() });
+  return { code: result.status, stdout: (result.stdout || '').trim(), stderr: (result.stderr || '').trim() };
+}
+
+function git(repoPath, args) {
+  const result = gitRun(repoPath, args);
+  if (result.code !== 0) throw new Error(`git ${args.join(' ')} failed (${result.code}): ${result.stderr}`);
+  return result.stdout;
+}
+
+// Real fixture repository: init + baseline + one labeled commit per entry.
+function buildFixtureRepo(root, name, commits = []) {
+  const repoPath = path.join(root, `repo-${name}`);
+  fs.mkdirSync(repoPath, { recursive: true });
+  git(repoPath, ['init', '-q']);
+  git(repoPath, ['config', 'user.email', 'engine-final@test.invalid']);
+  git(repoPath, ['config', 'user.name', 'P-ENG-FINAL']);
+  fs.writeFileSync(path.join(repoPath, 'README.md'), `# ${name}\n\nFixture repository.\n`);
+  git(repoPath, ['add', '.']);
+  git(repoPath, ['commit', '-qm', 'baseline']);
+  const baseline = git(repoPath, ['rev-parse', 'HEAD']);
+  const shas = {};
+  for (const label of commits) {
+    fs.writeFileSync(path.join(repoPath, `${label}.txt`), `change ${name} ${label}\n` + 'x'.repeat(64) + '\n');
+    git(repoPath, ['add', '.']);
+    git(repoPath, ['commit', '-qm', `feature ${label}`]);
+    shas[label] = git(repoPath, ['rev-parse', 'HEAD']);
+  }
+  return { repoPath, baseline, shas };
+}
+
+// Real git hook inside the FIXTURE repository only (.git/hooks/post-commit).
+// The user's global git config and global hooks are never touched.
+function installPostCommitHook(repoPath, { home, mode = 'ok' }) {
+  const hookPath = path.join(repoPath, '.git', 'hooks', 'post-commit');
+  const slash = value => value.replace(/\\/g, '/');
+  const cli = mode === 'ok' ? slash(BRIDGE_CLI) : slash(path.join(BRIDGE_ROOT, 'src', 'bridge', 'bin', 'commit-boundary.missing.js'));
+  const script = [
+    '#!/bin/sh',
+    mode === 'ok' ? `"${slash(process.execPath)}" "${cli}" --home "${slash(home)}" --cwd "${slash(repoPath)}" >/dev/null 2>&1` : `"${slash(process.execPath)}" "${cli}" --home "${slash(home)}" --cwd "${slash(repoPath)}"`,
+    'exit 1',
+    '',
+  ].join('\n');
+  fs.mkdirSync(path.dirname(hookPath), { recursive: true });
+  fs.writeFileSync(hookPath, script, { mode: 0o755 });
+  return { hookPath, mode, content: script };
+}
+
+// Real bridge console + journal, isolated home, ephemeral loopback port.
+function startBridge(home) {
+  fs.mkdirSync(home, { recursive: true });
+  const { createBridge, projectRegistry, resolveRepoContext } = require(path.join(BRIDGE_ROOT, 'src', 'bridge'));
+  const { createConsoleServer } = require(path.join(BRIDGE_ROOT, 'src', 'server'));
+  const bridge = createBridge({ homeDir: home });
+  const server = http.createServer(createConsoleServer({ home }));
+  return {
+    home, bridge, projectRegistry, resolveRepoContext, server,
+    async listen() {
+      await new Promise(resolve => server.listen(0, '127.0.0.1', resolve));
+      return `http://127.0.0.1:${server.address().port}`;
+    },
+    async close() { server.closeAllConnections(); await new Promise(resolve => server.close(resolve)); },
+  };
+}
+
+// Real vector-hub server on an isolated root. Embeddings are deterministic test
+// vectors (the paid embedding API is NOT exercised - stated as a boundary).
+function startVector(root) {
+  fs.mkdirSync(root, { recursive: true });
+  const { HubManager } = require(path.join(VECTOR_ROOT, 'dist', 'cjs', 'server', 'manager'));
+  const { createServer } = require(path.join(VECTOR_ROOT, 'dist', 'cjs', 'server', 'server'));
+  const embeddingsFactory = () => ({
+    maxTokens: 500,
+    async createEmbeddings(inputs) {
+      const texts = Array.isArray(inputs) ? inputs : [inputs];
+      return { status: 'success', output: texts.map((text, index) => [1, 0, index / 1000]) };
+    },
+  });
+  return HubManager.load({
+    rootPath: path.join(root, 'indexes'),
+    configPath: path.join(root, 'vector-config.json'),
+    embeddingsFactory,
+  }).then(async manager => {
+    await manager.saveSettings({ provider: 'minimax', apiKey: 'embedding-test-marker' });
+    const server = http.createServer(createServer(manager));
+    await new Promise(resolve => server.listen(0, '127.0.0.1', resolve));
+    return {
+      manager, server,
+      url: `http://127.0.0.1:${server.address().port}`,
+      async close() { server.closeAllConnections(); await new Promise(resolve => server.close(resolve)); manager.close?.(); },
+    };
+  });
+}
+
+// Real vendored knowledge-engine, in-process, isolated data dir.
+function startEngine({ dataDir, bridgeUrl, vectorUrl, analyzer, debug = true, consume = true, runtimeToken = null }) {
+  fs.mkdirSync(dataDir, { recursive: true });
+  // Fail-closed endpoint guard: the coordinator ignores env and silently
+  // defaults its vector/bridge legs to LIVE 8787/8790, so an engine without
+  // explicit OWN endpoints is refused instead of defaulting live.
+  const { assertOwnedEndpoint } = require('./parallel-qa-engine-endpoint-guard');
+  assertOwnedEndpoint(vectorUrl, { role: 'vectorUrl' });
+  assertOwnedEndpoint(bridgeUrl, { role: 'bridgeUrl' });
+  const { createEngineServer } = require(VENDORED_ENGINE_SERVER);
+  const runtime = createEngineServer({ dataDir, consume, debug, analyzer, bridgeUrl, vectorUrl, runtimeToken });
+  return {
+    runtime,
+    async listen() {
+      await runtime.ready;
+      await new Promise(resolve => runtime.server.listen(0, '127.0.0.1', resolve));
+      return `http://127.0.0.1:${runtime.server.address().port}`;
+    },
+    async close() { await runtime.close(); },
+  };
+}
+
+async function waitUntil(fn, timeoutMs, describe) {
+  const started = Date.now();
+  for (;;) {
+    const value = await fn();
+    if (value) return value;
+    if (Date.now() - started > timeoutMs) throw new Error(`waitUntil timed out after ${timeoutMs}ms: ${describe}`);
+    await new Promise(resolve => setTimeout(resolve, 25));
+  }
+}
+
+async function callJson(base, route, body, init = {}) {
+  const response = await fetch(`${base}${route}`, {
+    ...init,
+    method: body === undefined && !init.method ? 'GET' : init.method || 'POST',
+    headers: { 'Content-Type': 'application/json', ...(init.headers || {}) },
+    body: body === undefined && (!init.method || init.method === 'GET') ? undefined : JSON.stringify(body),
+  });
+  let payload = null;
+  try { payload = await response.json(); } catch { payload = null; }
+  return { status: response.status, body: payload };
+}
+
+// Fire a request and drop the response (simulates a lost network answer).
+function callJsonAndLoseResponse(base, route, body) {
+  return new Promise(resolve => {
+    const url = new URL(route, base);
+    const payload = JSON.stringify(body || {});
+    const request = http.request({
+      hostname: url.hostname, port: url.port, path: url.pathname + url.search, method: 'POST',
+      headers: { 'Content-Type': 'application/json', 'Content-Length': Buffer.byteLength(payload) },
+    });
+    request.on('error', () => resolve({ lost: true, error: 'socket-error' }));
+    request.end(payload);
+    setTimeout(() => { request.destroy(); resolve({ lost: true, error: 'response-dropped' }); }, 120);
+  });
+}
+
+function listFilesRecursive(root) {
+  const out = [];
+  if (!fs.existsSync(root)) return out;
+  for (const entry of fs.readdirSync(root, { withFileTypes: true })) {
+    const absolute = path.join(root, entry.name);
+    if (entry.isDirectory()) out.push(...listFilesRecursive(absolute).map(sub => `${entry.name}/${sub}`));
+    else out.push(entry.name);
+  }
+  return out.sort();
+}
+
+function readJsonl(file) {
+  if (!fs.existsSync(file)) return [];
+  return fs.readFileSync(file, 'utf8').split('\n').filter(Boolean).map(line => JSON.parse(line));
+}
+
+function saveEvidence(laneDir, name, value) {
+  fs.mkdirSync(laneDir, { recursive: true });
+  fs.writeFileSync(path.join(laneDir, name), JSON.stringify(value, null, 2));
+}
+
+function hashText(value) {
+  return require('node:crypto').createHash('sha256').update(value).digest('hex');
+}
+
+module.exports = {
+  REPO_ROOT,
+  VENDORED_ENGINE_SERVER,
+  VENDORED_ATOMIC_FILE,
+  RUNTIME_IDENTITY: { bridge: BRIDGE, vector: VECTOR },
+  BRIDGE_ROOT,
+  BRIDGE_CLI,
+  VECTOR_ROOT,
+  TEMP_ROOT,
+  ensureTempRoot,
+  makeLaneDir,
+  isolatedEnv,
+  git,
+  gitRun,
+  buildFixtureRepo,
+  installPostCommitHook,
+  startBridge,
+  startVector,
+  startEngine,
+  waitUntil,
+  callJson,
+  callJsonAndLoseResponse,
+  listFilesRecursive,
+  readJsonl,
+  saveEvidence,
+  hashText,
+};

package/_site/_test/helpers/parallel-engine-final-standin-analyzer.js

@@ -0,0 +1,159 @@
+'use strict';
+
+// P-ENG-FINAL deterministic analyzer stand-in.
+//
+// SCOPE (declared): this stand-in writes REAL staged Markdown bytes and a REAL
+// staging manifest through the engine's real promotion path (journal, backups,
+// hashing, state advance, index sync). It does NOT exercise a real model, a
+// real prompt, or any paid API - the promoted text is fixture content, and the
+// memory candidates it emits are engine-validated against the frozen
+// conversation snapshot exactly like a real analyzer's would be.
+//
+// The stand-in exists to make timing and failure deterministic; every engine
+// behaviour asserted by the drivers (journal, claim binding, promotion,
+// eligibility gate, dedup, retry) is the production code path.
+
+const fs = require('node:fs');
+const path = require('node:path');
+const crypto = require('node:crypto');
+
+function hashBuffer(value) {
+  return `sha256:${crypto.createHash('sha256').update(value).digest('hex')}`;
+}
+
+function createStandinAnalyzer(options = {}) {
+  const {
+    eventsFile = null,
+    failShas = new Set(),
+    // Transient fault model: the FIRST invocation for a listed SHA fails, the
+    // retry succeeds (models a one-off fault, not a permanently broken commit).
+    failOnceShas = new Set(),
+    readSnapshot = null,
+    humanQuote = null,
+    delegatedQuote = null,
+    emitMemoryCandidates = false,
+    extraOperations = [],
+  } = options;
+
+  function record(entry) {
+    if (!eventsFile) return;
+    fs.mkdirSync(path.dirname(eventsFile), { recursive: true });
+    fs.appendFileSync(eventsFile, `${JSON.stringify({ ...entry, at: new Date().toISOString(), pid: process.pid })}\n`);
+  }
+
+  const attemptsBySha = new Map();
+
+  function pickEvent(snapshot, origin) {
+    const turns = snapshot?.turns || [];
+    for (const turn of turns) {
+      for (const event of turn.userEvents || []) {
+        if ((event.inputOrigin || 'unknown') === origin) return event;
+      }
+    }
+    return null;
+  }
+
+  return {
+    sourceReadMode: 'isolated-frozen-evidence',
+    async runClaim(input) {
+      const claim = input.claim;
+      const sha = claim.commitSha;
+      const attempt = (attemptsBySha.get(sha) || 0) + 1;
+      attemptsBySha.set(sha, attempt);
+      record({ kind: 'start', projectId: input.projectId, runId: claim.runId, commitSha: sha, attempt });
+      try {
+        if (failShas.has(sha) || (failOnceShas.has(sha) && attempt === 1)) {
+          const error = new Error('stand-in analyzer failure (injected)');
+          error.code = 'STANDIN_ANALYZER_FAILURE';
+          throw error;
+        }
+        const filename = `changes/${sha.slice(0, 12)}.md`;
+        const lines = [
+          '---',
+          `commit: ${sha}`,
+          `run: ${claim.runId}`,
+          '---',
+          `# Fixture knowledge ${input.projectId}`,
+          '',
+          `Commit ${sha} promoted by the deterministic stand-in analyzer (no real model output).`,
+          '',
+        ];
+        if (humanQuote) lines.push(`用户原话：${humanQuote}`, '');
+        if (delegatedQuote) lines.push(`维护指令：${delegatedQuote}`, '');
+        const markdown = lines.join('\n');
+
+        const AtomicFile = require('../../../_modules/knowledge-engine/src/runtime/atomic-file.js');
+        AtomicFile.writeFileAtomic(path.join(input.stagingPath, 'files', ...filename.split('/')), markdown);
+
+        const references = [
+          `commit:${claim.commitSha}`,
+          `patch:${claim.patchHash}`,
+          `conversation:${claim.conversationSnapshotHash}`,
+          `retrieval:${claim.retrievalManifestHash}`,
+        ];
+        const operations = [{
+          operation: 'create',
+          path: filename,
+          reason: 'Deterministic stand-in knowledge for acceptance coverage',
+          sha256: hashBuffer(markdown),
+          evidenceReferences: references,
+        }];
+        for (const extra of extraOperations) {
+          operations.push({
+            operation: extra.operation,
+            path: extra.path,
+            reason: extra.reason || 'Deterministic stand-in extra operation',
+            sha256: extra.operation === 'delete' ? undefined : hashBuffer(extra.content || ''),
+            evidenceReferences: references,
+          });
+          if (extra.operation !== 'delete' && extra.content) {
+            AtomicFile.writeFileAtomic(path.join(input.stagingPath, 'files', ...extra.path.split('/')), extra.content);
+          }
+        }
+
+        const memoryCandidates = [];
+        if (emitMemoryCandidates && readSnapshot) {
+          const snapshot = readSnapshot(input.projectId, sha);
+          const human = pickEvent(snapshot, 'human');
+          const delegated = pickEvent(snapshot, 'agent-delegation');
+          if (human && humanQuote) {
+            memoryCandidates.push({
+              text: humanQuote,
+              entryId: filename,
+              quote: humanQuote,
+              sourceEventId: human.eventId,
+              sourceHash: human.contentHash,
+              scopeType: 'global',
+              eligibility: { durable: true, crossProject: true, reason: 'fixture durable owner preference', scopeReason: 'fixture scope reason' },
+            });
+          }
+          if (delegated && delegatedQuote) {
+            memoryCandidates.push({
+              text: delegatedQuote,
+              entryId: filename,
+              quote: delegatedQuote,
+              sourceEventId: delegated.eventId,
+              sourceHash: delegated.contentHash,
+              scopeType: 'global',
+              eligibility: { durable: true, crossProject: true, reason: 'fixture delegated maintenance instruction', scopeReason: 'fixture scope reason' },
+            });
+          }
+        }
+
+        AtomicFile.writeJsonAtomic(input.manifestPath, {
+          schema: 'knowledge-staging-manifest/v1',
+          projectId: input.projectId,
+          runId: claim.runId,
+          commitSha: claim.commitSha,
+          operations,
+          memoryCandidates,
+        });
+        return { ok: true };
+      } finally {
+        record({ kind: 'finish', projectId: input.projectId, runId: claim.runId, commitSha: sha });
+      }
+    },
+  };
+}
+
+module.exports = { createStandinAnalyzer };