@evalgate/sdk 2.0.0

package/dist/cli/upgrade.js

@@ -0,0 +1,492 @@
+"use strict";
+/**
+ * evalgate upgrade --full — Upgrade from Tier 1 (built-in gate) to Tier 2 (full gate)
+ *
+ * What it does:
+ *   1. Adds full regression gate script (scripts/regression-gate.ts)
+ *   2. Adds baseline governance workflow (.github/workflows/baseline-governance.yml)
+ *   3. Updates package.json with eval:regression-gate + eval:baseline-update scripts
+ *   4. Updates .github/workflows/evalgate-gate.yml to use project mode
+ *   5. Prints next steps
+ */
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.parseUpgradeArgs = parseUpgradeArgs;
+exports.runUpgrade = runUpgrade;
+const fs = __importStar(require("node:fs"));
+const path = __importStar(require("node:path"));
+// ── Detect environment ──
+function detectPackageManager(cwd) {
+    if (fs.existsSync(path.join(cwd, "pnpm-lock.yaml")))
+        return "pnpm";
+    if (fs.existsSync(path.join(cwd, "yarn.lock")))
+        return "yarn";
+    return "npm";
+}
+function ok(msg) {
+    console.log(`  ✔ ${msg}`);
+}
+function skip(msg) {
+    console.log(`  – ${msg}`);
+}
+// ── 1. Create scripts/regression-gate.ts ──
+function createGateScript(cwd) {
+    const scriptPath = path.join(cwd, "scripts", "regression-gate.ts");
+    if (fs.existsSync(scriptPath)) {
+        skip("scripts/regression-gate.ts already exists");
+        return true;
+    }
+    const scriptsDir = path.join(cwd, "scripts");
+    if (!fs.existsSync(scriptsDir)) {
+        fs.mkdirSync(scriptsDir, { recursive: true });
+    }
+    const content = `#!/usr/bin/env npx tsx
+/**
+ * Full regression gate — compares current test results against baseline.
+ *
+ * Usage:
+ *   npx tsx scripts/regression-gate.ts                  # run gate
+ *   npx tsx scripts/regression-gate.ts --update-baseline # update baseline with current values
+ *
+ * Generated by: npx evalgate upgrade --full
+ */
+import { readFileSync, writeFileSync, existsSync, mkdirSync } from "node:fs";
+import { execSync, spawnSync } from "node:child_process";
+import { resolve } from "node:path";
+
+const BASELINE_PATH = resolve("evals/baseline.json");
+const REPORT_PATH = resolve("evals/regression-report.json");
+const CONFIDENCE_PATH = resolve("evals/confidence-summary.json");
+
+const isUpdateBaseline = process.argv.includes("--update-baseline");
+
+// ── Helpers ──
+
+function loadJSON(p: string): Record<string, unknown> | null {
+  try {
+    return JSON.parse(readFileSync(p, "utf-8"));
+  } catch {
+    return null;
+  }
+}
+
+function getHeadSha(): string {
+  try {
+    return execSync("git rev-parse --short HEAD").toString().trim();
+  } catch {
+    return "0000000";
+  }
+}
+
+function writeReport(report: Record<string, unknown>): void {
+  const dir = resolve("evals");
+  if (!existsSync(dir)) mkdirSync(dir, { recursive: true });
+  writeFileSync(REPORT_PATH, JSON.stringify(report, null, 2) + "\\n");
+}
+
+// ── Run tests ──
+
+function runTests(): { passed: boolean; total: number; durationMs: number } {
+  const t0 = Date.now();
+  const result = spawnSync("npm", ["test"], {
+    stdio: "pipe",
+    shell: process.platform === "win32",
+    timeout: 300_000,
+  });
+  const durationMs = Date.now() - t0;
+  const passed = result.status === 0;
+  const output = (result.stdout?.toString() ?? "") + (result.stderr?.toString() ?? "");
+
+  let total = 0;
+  const m =
+    output.match(/(\\d+)\\s+(?:tests?|specs?)\\s+(?:passed|completed)/i) ??
+    output.match(/Tests:\\s+(\\d+)\\s+passed/i) ??
+    output.match(/(\\d+)\\s+passing/i);
+  if (m) total = parseInt(m[1], 10);
+
+  return { passed, total, durationMs };
+}
+
+// ── Main ──
+
+const baseline = loadJSON(BASELINE_PATH);
+if (!baseline) {
+  console.error("❌ Baseline not found. Run: npx evalgate init");
+  const report = {
+    schemaVersion: 1,
+    timestamp: new Date().toISOString(),
+    exitCode: 2,
+    category: "infra_error",
+    passed: false,
+    failures: ["Baseline file not found"],
+    deltas: [],
+    baseline: null,
+    durationMs: 0,
+    command: "npm test",
+    runner: "unknown",
+  };
+  writeReport(report);
+  process.exit(2);
+}
+
+const tests = runTests();
+
+if (isUpdateBaseline) {
+  const user = process.env.USER || process.env.USERNAME || "unknown";
+  const now = new Date().toISOString();
+  const updated = {
+    ...baseline,
+    updatedAt: now,
+    updatedBy: user,
+    commitSha: getHeadSha(),
+    confidenceTests: {
+      ...(baseline.confidenceTests as Record<string, unknown> ?? {}),
+      passed: tests.passed,
+      total: tests.total,
+    },
+  };
+  writeFileSync(BASELINE_PATH, JSON.stringify(updated, null, 2) + "\\n");
+  console.log("✅ Baseline updated with current test results");
+  console.log(\`   Tests: \${tests.total} (\${tests.passed ? "passing" : "FAILING"})\`);
+  process.exit(0);
+}
+
+// ── Compare ──
+
+const bConf = baseline.confidenceTests as { passed?: boolean; total?: number } | undefined;
+const baselinePassed = bConf?.passed ?? true;
+const baselineTotal = bConf?.total ?? 0;
+
+const failures: string[] = [];
+const deltas: Array<Record<string, unknown>> = [];
+
+deltas.push({
+  metric: "tests_passing",
+  baseline: baselinePassed,
+  current: tests.passed,
+  delta: tests.passed === baselinePassed ? "0" : tests.passed ? "+1" : "-1",
+  status: tests.passed ? "pass" : "fail",
+});
+
+if (!tests.passed && baselinePassed) {
+  failures.push("Tests were passing in baseline but are now failing");
+}
+
+if (tests.total > 0 || baselineTotal > 0) {
+  const d = tests.total - baselineTotal;
+  deltas.push({
+    metric: "test_count",
+    baseline: baselineTotal,
+    current: tests.total,
+    delta: d >= 0 ? \`+\${d}\` : \`\${d}\`,
+    status: tests.total >= baselineTotal ? "pass" : "fail",
+  });
+  if (tests.total < baselineTotal) {
+    failures.push(\`Test count dropped from \${baselineTotal} to \${tests.total} (\${d})\`);
+  }
+}
+
+const hasRegression = failures.length > 0;
+const report = {
+  schemaVersion: 1,
+  timestamp: new Date().toISOString(),
+  exitCode: hasRegression ? 1 : 0,
+  category: hasRegression ? "regression" : "pass",
+  passed: !hasRegression,
+  failures,
+  deltas,
+  baseline: {
+    updatedAt: (baseline.updatedAt as string) ?? "unknown",
+    updatedBy: (baseline.updatedBy as string) ?? "unknown",
+  },
+  durationMs: tests.durationMs,
+  command: "npm test",
+  runner: "unknown",
+};
+
+writeReport(report);
+
+if (hasRegression) {
+  console.error("❌ REGRESSION DETECTED");
+  for (const f of failures) console.error(\`   \${f}\`);
+} else {
+  console.log("✅ NO REGRESSION — gate passed");
+}
+
+for (const d of deltas) {
+  const icon = d.status === "pass" ? "✔" : "✖";
+  console.log(\`  \${icon} \${d.metric}: \${d.baseline} → \${d.current} (\${d.delta})\`);
+}
+
+process.exit(report.exitCode);
+`;
+    fs.writeFileSync(scriptPath, content);
+    ok("Created scripts/regression-gate.ts");
+    return true;
+}
+// ── 2. Add npm scripts to package.json ──
+function addNpmScripts(cwd) {
+    const pkgPath = path.join(cwd, "package.json");
+    if (!fs.existsSync(pkgPath))
+        return false;
+    let pkg;
+    try {
+        pkg = JSON.parse(fs.readFileSync(pkgPath, "utf-8"));
+    }
+    catch {
+        return false;
+    }
+    const scripts = (pkg.scripts ?? {});
+    let changed = false;
+    if (!scripts["eval:regression-gate"]) {
+        scripts["eval:regression-gate"] = "npx tsx scripts/regression-gate.ts";
+        changed = true;
+    }
+    if (!scripts["eval:baseline-update"]) {
+        scripts["eval:baseline-update"] =
+            "npx tsx scripts/regression-gate.ts --update-baseline";
+        changed = true;
+    }
+    if (changed) {
+        pkg.scripts = scripts;
+        fs.writeFileSync(pkgPath, `${JSON.stringify(pkg, null, 2)}\n`);
+        ok("Added eval:regression-gate and eval:baseline-update scripts to package.json");
+    }
+    else {
+        skip("eval:regression-gate and eval:baseline-update scripts already exist");
+    }
+    return true;
+}
+// ── 3. Create baseline governance workflow ──
+function createGovernanceWorkflow(cwd) {
+    const workflowDir = path.join(cwd, ".github", "workflows");
+    const workflowPath = path.join(workflowDir, "baseline-governance.yml");
+    if (fs.existsSync(workflowPath)) {
+        skip(".github/workflows/baseline-governance.yml already exists");
+        return true;
+    }
+    if (!fs.existsSync(workflowDir)) {
+        fs.mkdirSync(workflowDir, { recursive: true });
+    }
+    const workflow = `# Baseline Governance — requires label + approval for baseline changes
+# Auto-generated by: npx evalgate upgrade --full
+name: Baseline Governance
+
+on:
+  pull_request:
+    paths:
+      - 'evals/baseline.json'
+
+jobs:
+  governance:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Check label
+        run: |
+          LABELS=\${{ toJSON(github.event.pull_request.labels.*.name) }}
+          if echo "$LABELS" | grep -q "baseline-update"; then
+            echo "✅ baseline-update label found"
+          elif echo "$LABELS" | grep -q "baseline-exception"; then
+            echo "⚠️ baseline-exception label found — bypassing delta checks"
+          else
+            echo "❌ Missing 'baseline-update' label"
+            echo "Add the 'baseline-update' label to this PR to update the baseline."
+            exit 1
+          fi
+
+      - name: Show baseline diff
+        run: |
+          echo "## Baseline Changes" >> "$GITHUB_STEP_SUMMARY"
+          echo "" >> "$GITHUB_STEP_SUMMARY"
+          echo "\\\`\\\`\\\`diff" >> "$GITHUB_STEP_SUMMARY"
+          git diff HEAD~1 -- evals/baseline.json >> "$GITHUB_STEP_SUMMARY" || echo "No previous baseline" >> "$GITHUB_STEP_SUMMARY"
+          echo "\\\`\\\`\\\`" >> "$GITHUB_STEP_SUMMARY"
+`;
+    fs.writeFileSync(workflowPath, workflow);
+    ok("Created .github/workflows/baseline-governance.yml");
+    return true;
+}
+// ── 4. Upgrade evalgate-gate.yml to project mode ──
+function upgradeGateWorkflow(cwd) {
+    const pm = detectPackageManager(cwd);
+    const workflowPath = path.join(cwd, ".github", "workflows", "evalgate-gate.yml");
+    if (!fs.existsSync(workflowPath)) {
+        skip("No .github/workflows/evalgate-gate.yml found — run evalgate init first");
+        return false;
+    }
+    const content = fs.readFileSync(workflowPath, "utf-8");
+    // Already upgraded?
+    if (content.includes("eval:regression-gate")) {
+        skip("evalgate-gate.yml already uses project mode");
+        return true;
+    }
+    const installCmd = pm === "pnpm"
+        ? "pnpm install --frozen-lockfile"
+        : pm === "yarn"
+            ? "yarn install --frozen-lockfile"
+            : "npm ci";
+    const setupSteps = pm === "pnpm"
+        ? `      - uses: pnpm/action-setup@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+          cache: pnpm
+      - run: ${installCmd}`
+        : `      - uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+          cache: ${pm}
+      - run: ${installCmd}`;
+    const workflow = `# EvalGate Regression Gate (Full / Tier 2)
+# Upgraded by: npx evalgate upgrade --full
+name: EvalGate Gate
+
+on:
+  pull_request:
+    branches: [main]
+
+concurrency:
+  group: evalgate-\${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  regression-gate:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+${setupSteps}
+      - name: Run regression gate
+        run: ${pm} run eval:regression-gate
+
+      - name: Gate summary
+        if: always()
+        run: npx -y @evalgate/sdk@^2 gate --format github
+
+      - name: Upload report
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: regression-report
+          path: evals/regression-report.json
+          if-no-files-found: ignore
+`;
+    fs.writeFileSync(workflowPath, workflow);
+    ok("Upgraded .github/workflows/evalgate-gate.yml to project mode (Tier 2)");
+    return true;
+}
+// ── 5. Add CODEOWNERS entry ──
+function addCodeowners(cwd) {
+    const codeownersPath = path.join(cwd, ".github", "CODEOWNERS");
+    const entry = "evals/baseline.json";
+    if (fs.existsSync(codeownersPath)) {
+        const content = fs.readFileSync(codeownersPath, "utf-8");
+        if (content.includes(entry)) {
+            skip("CODEOWNERS already has evals/baseline.json entry");
+            return true;
+        }
+        fs.appendFileSync(codeownersPath, `\n# EvalGate baseline — requires approval\n${entry} @YOUR_TEAM\n`);
+    }
+    else {
+        const dir = path.join(cwd, ".github");
+        if (!fs.existsSync(dir))
+            fs.mkdirSync(dir, { recursive: true });
+        fs.writeFileSync(codeownersPath, `# EvalGate baseline — requires approval\n${entry} @YOUR_TEAM\n`);
+    }
+    ok("Added evals/baseline.json to .github/CODEOWNERS (edit @YOUR_TEAM)");
+    return true;
+}
+function parseUpgradeArgs(argv) {
+    return { full: argv.includes("--full") };
+}
+// ── Main ──
+function runUpgrade(argv) {
+    const args = parseUpgradeArgs(argv);
+    const cwd = process.cwd();
+    if (!args.full) {
+        console.log(`evalgate upgrade — Upgrade regression gate
+
+Usage:
+  evalgate upgrade --full    Upgrade from Tier 1 (built-in) to Tier 2 (full gate)
+
+What --full does:
+  1. Creates scripts/regression-gate.ts (full gate script)
+  2. Adds eval:regression-gate + eval:baseline-update npm scripts
+  3. Creates baseline governance workflow
+  4. Upgrades CI workflow to project mode
+  5. Adds CODEOWNERS entry for baseline
+
+After upgrading:
+  - evalgate gate delegates to your eval:regression-gate script
+  - Baseline changes require PR label + approval
+  - Full metric comparison: golden eval, confidence, latency, cost
+`);
+        return argv.includes("--help") || argv.includes("-h") ? 0 : 1;
+    }
+    console.log("");
+    console.log("  evalgate upgrade --full — upgrading to Tier 2\n");
+    // Check preconditions
+    const pkgPath = path.join(cwd, "package.json");
+    if (!fs.existsSync(pkgPath)) {
+        console.error("  ✖ No package.json found. Run this from a Node.js project root.");
+        return 1;
+    }
+    if (!fs.existsSync(path.join(cwd, "evals", "baseline.json"))) {
+        console.error("  ✖ No evals/baseline.json found. Run 'npx evalgate init' first.");
+        return 1;
+    }
+    createGateScript(cwd);
+    addNpmScripts(cwd);
+    createGovernanceWorkflow(cwd);
+    upgradeGateWorkflow(cwd);
+    addCodeowners(cwd);
+    console.log("");
+    console.log("  Done! Your repo is now Tier 2.\n");
+    console.log("  What changed:");
+    console.log("    - scripts/regression-gate.ts    Full gate script");
+    console.log("    - package.json                  eval:regression-gate + eval:baseline-update");
+    console.log("    - .github/workflows/            Gate + governance workflows");
+    console.log("    - .github/CODEOWNERS            Baseline requires approval\n");
+    console.log("  Next:");
+    console.log("    git add -A");
+    console.log("    git commit -m 'chore: upgrade EvalGate gate to Tier 2'");
+    console.log("    git push\n");
+    console.log("  Commands:");
+    console.log("    npx evalgate gate                 Run full gate locally");
+    console.log("    npx evalgate baseline update      Update baseline with real scores");
+    console.log("");
+    return 0;
+}

package/dist/cli/workspace.d.ts

@@ -0,0 +1,31 @@
+/**
+ * CORE-402: Centralized .evalgate workspace resolution
+ *
+ * Provides unified workspace path resolution for all EvalGate CLI commands.
+ * Prefers .evalgate/; falls back to .evalai/ for backward compatibility.
+ */
+/**
+ * EvalGate workspace paths
+ */
+export interface EvalWorkspace {
+    /** Project root directory */
+    root: string;
+    /** .evalgate directory (or .evalai for legacy projects) */
+    evalDir: string;
+    /** @deprecated Use evalDir */
+    evalgateDir: string;
+    /** runs directory */
+    runsDir: string;
+    /** manifest.json path */
+    manifestPath: string;
+    /** last-run.json path */
+    lastRunPath: string;
+    /** runs/index.json path */
+    indexPath: string;
+    /** baseline-run.json path */
+    baselinePath: string;
+}
+/**
+ * Resolve EvalGate workspace paths. Prefers .evalgate/, falls back to .evalai/.
+ */
+export declare function resolveEvalWorkspace(projectRoot?: string): EvalWorkspace;

package/dist/cli/workspace.js

@@ -0,0 +1,68 @@
+"use strict";
+/**
+ * CORE-402: Centralized .evalgate workspace resolution
+ *
+ * Provides unified workspace path resolution for all EvalGate CLI commands.
+ * Prefers .evalgate/; falls back to .evalai/ for backward compatibility.
+ */
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.resolveEvalWorkspace = resolveEvalWorkspace;
+const fs = __importStar(require("node:fs"));
+const path = __importStar(require("node:path"));
+/**
+ * Resolve EvalGate workspace paths. Prefers .evalgate/, falls back to .evalai/.
+ */
+function resolveEvalWorkspace(projectRoot = process.cwd()) {
+    const evalgateDir = path.join(projectRoot, ".evalgate");
+    const evalaiDir = path.join(projectRoot, ".evalai");
+    const useLegacy = fs.existsSync(evalaiDir) && !fs.existsSync(evalgateDir);
+    const evalDir = useLegacy ? evalaiDir : evalgateDir;
+    if (useLegacy && !process.__EVALGATE_LEGACY_EVALAI_WARNED) {
+        console.warn("[EvalGate] Deprecation: .evalai/ is deprecated. Migrate to .evalgate/ (e.g. mv .evalai .evalgate).");
+        process.__EVALGATE_LEGACY_EVALAI_WARNED = true;
+    }
+    const runsDir = path.join(evalDir, "runs");
+    return {
+        root: projectRoot,
+        evalDir,
+        evalgateDir: evalDir,
+        runsDir,
+        manifestPath: path.join(evalDir, "manifest.json"),
+        lastRunPath: path.join(evalDir, "last-run.json"),
+        indexPath: path.join(runsDir, "index.json"),
+        baselinePath: path.join(evalDir, "baseline-run.json"),
+    };
+}