@claude-flow/cli 3.10.3 → 3.10.5

package/dist/src/benchmarks/gaia-e2e-smoke.d.ts

@@ -0,0 +1,27 @@
+/**
+ * GAIA End-to-End Smoke — ADR-133
+ *
+ * Wires gaia-agent.ts + gaia-judge.ts into a single end-to-end pipeline:
+ *
+ *   for each question in SMOKE_FIXTURE:
+ *     1. runGaiaAgent(question)  — Haiku agent loop, ≤8 turns
+ *     2. judgeAnswer(question, result.finalAnswer)  — exact-match fast-path,
+ *        Sonnet LLM-judge only if exact-match misses
+ *
+ * Reports: pass rate, total cost, mean turn count.
+ * Asserts: ≥ 3/5 questions pass (lenient — smoke fixture is not trivial).
+ *
+ * Cost discipline:
+ *   - Agent: claude-haiku-4-5 at $0.25/$1.25 per M tokens
+ *   - Judge: claude-sonnet-4-6 at $3/$15 per M tokens (only when needed)
+ *   - Expected total for 5 questions × ~2 turns × Haiku + 1-2 Sonnet
+ *     judge calls ≈ $0.02
+ *
+ * Usage:
+ *   ANTHROPIC_API_KEY=sk-ant-... npx tsx src/benchmarks/gaia-e2e-smoke.ts
+ *
+ * Refs: ADR-133, #2156
+ */
+declare function runE2ESmoke(): Promise<void>;
+export { runE2ESmoke };
+//# sourceMappingURL=gaia-e2e-smoke.d.ts.map

package/dist/src/benchmarks/gaia-e2e-smoke.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"gaia-e2e-smoke.d.ts","sourceRoot":"","sources":["../../../src/benchmarks/gaia-e2e-smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AAqDH,iBAAe,WAAW,IAAI,OAAO,CAAC,IAAI,CAAC,CA0G1C;AAiBD,OAAO,EAAE,WAAW,EAAE,CAAC"}

package/dist/src/benchmarks/gaia-e2e-smoke.js

@@ -0,0 +1,136 @@
+/**
+ * GAIA End-to-End Smoke — ADR-133
+ *
+ * Wires gaia-agent.ts + gaia-judge.ts into a single end-to-end pipeline:
+ *
+ *   for each question in SMOKE_FIXTURE:
+ *     1. runGaiaAgent(question)  — Haiku agent loop, ≤8 turns
+ *     2. judgeAnswer(question, result.finalAnswer)  — exact-match fast-path,
+ *        Sonnet LLM-judge only if exact-match misses
+ *
+ * Reports: pass rate, total cost, mean turn count.
+ * Asserts: ≥ 3/5 questions pass (lenient — smoke fixture is not trivial).
+ *
+ * Cost discipline:
+ *   - Agent: claude-haiku-4-5 at $0.25/$1.25 per M tokens
+ *   - Judge: claude-sonnet-4-6 at $3/$15 per M tokens (only when needed)
+ *   - Expected total for 5 questions × ~2 turns × Haiku + 1-2 Sonnet
+ *     judge calls ≈ $0.02
+ *
+ * Usage:
+ *   ANTHROPIC_API_KEY=sk-ant-... npx tsx src/benchmarks/gaia-e2e-smoke.ts
+ *
+ * Refs: ADR-133, #2156
+ */
+import * as os from 'node:os';
+import * as path from 'node:path';
+import { SMOKE_FIXTURE, } from './gaia-loader.js';
+import { runGaiaAgent, } from './gaia-agent.js';
+import { judgeAnswer, } from './gaia-judge.js';
+// ---------------------------------------------------------------------------
+// Configuration
+// ---------------------------------------------------------------------------
+/** Agent model — Haiku only for cost discipline. */
+const AGENT_MODEL = 'claude-haiku-4-5';
+/** Judge model — Sonnet for semantic judgments (only when exact-match fails). */
+const JUDGE_MODEL = 'claude-sonnet-4-6';
+/** Minimum pass rate for smoke to succeed. */
+const MIN_PASS_RATE = 3 / 5;
+// Haiku pricing ($/M tokens)
+const HAIKU_IN = 0.25;
+const HAIKU_OUT = 1.25;
+// Sonnet pricing ($/M tokens)
+const SONNET_IN = 3.0;
+const SONNET_OUT = 15.0;
+// ---------------------------------------------------------------------------
+// Runner
+// ---------------------------------------------------------------------------
+async function runE2ESmoke() {
+    const hasKey = !!(process.env.ANTHROPIC_API_KEY?.trim());
+    if (!hasKey) {
+        console.error('ANTHROPIC_API_KEY is required for the end-to-end smoke.\n' +
+            'Set it with: export ANTHROPIC_API_KEY=sk-ant-...');
+        process.exit(1);
+    }
+    const cacheDir = path.join(os.homedir(), '.cache', 'ruflo', 'gaia', 'judgments');
+    const judgeOpts = { judgeModel: JUDGE_MODEL, cacheDir };
+    const questions = SMOKE_FIXTURE;
+    const rows = [];
+    console.log(`\n=== GAIA End-to-End Smoke (${questions.length} questions) ===\n`);
+    console.log(`Agent: ${AGENT_MODEL}  |  Judge: ${JUDGE_MODEL}\n` +
+        `Questions: ${questions.length}  |  Min pass rate: ${(MIN_PASS_RATE * 100).toFixed(0)}%\n`);
+    for (const q of questions) {
+        process.stdout.write(`[${q.task_id}] "${q.question.slice(0, 60)}..." `);
+        // Run agent
+        const agentResult = await runGaiaAgent(q, { model: AGENT_MODEL });
+        // Judge
+        const judgeResult = await judgeAnswer({ id: q.task_id, expected: q.final_answer, questionText: q.question }, agentResult.finalAnswer, judgeOpts);
+        rows.push({ question: q, agentResult, judgeResult });
+        const verdict = judgeResult.passed ? '\x1b[32mPASS\x1b[0m' : '\x1b[31mFAIL\x1b[0m';
+        const path_ = judgeResult.scoringPath;
+        console.log(`${verdict}  (turns=${agentResult.turns}, ` +
+            `answer="${agentResult.finalAnswer ?? 'null'}", ` +
+            `expected="${q.final_answer}", path=${path_})`);
+    }
+    // ── Summary ──
+    console.log('\n--- Summary ---\n');
+    const passed = rows.filter((r) => r.judgeResult.passed).length;
+    const total = rows.length;
+    const passRate = passed / total;
+    let totalAgentCostUsd = 0;
+    let totalJudgeCostUsd = 0;
+    let totalTurns = 0;
+    for (const row of rows) {
+        const { agentResult, judgeResult } = row;
+        totalTurns += agentResult.turns;
+        totalAgentCostUsd +=
+            (agentResult.totalInputTokens / 1_000_000) * HAIKU_IN +
+                (agentResult.totalOutputTokens / 1_000_000) * HAIKU_OUT;
+        totalJudgeCostUsd += judgeResult.judgeCostUsd ?? 0;
+    }
+    const totalCostUsd = totalAgentCostUsd + totalJudgeCostUsd;
+    const meanTurns = totalTurns / total;
+    console.log(`Pass rate      : ${passed}/${total}  (${(passRate * 100).toFixed(0)}%)`);
+    console.log(`Mean turns     : ${meanTurns.toFixed(1)}`);
+    console.log(`Agent cost     : $${totalAgentCostUsd.toFixed(5)}  (Haiku)`);
+    console.log(`Judge cost     : $${totalJudgeCostUsd.toFixed(5)}  (Sonnet, only when needed)`);
+    console.log(`Total cost     : $${totalCostUsd.toFixed(5)}`);
+    // ── Per-row detail ──
+    console.log('\n--- Per-question detail ---\n');
+    for (const row of rows) {
+        const { question, agentResult, judgeResult } = row;
+        const verdict = judgeResult.passed ? 'PASS' : 'FAIL';
+        console.log(`  ${verdict}  ${question.task_id}  turns=${agentResult.turns}  ` +
+            `path=${judgeResult.scoringPath}  ` +
+            `answer="${agentResult.finalAnswer ?? 'null'}"  ` +
+            `expected="${question.final_answer}"`);
+        if (judgeResult.judgeReason) {
+            console.log(`        reason: ${judgeResult.judgeReason}`);
+        }
+        if (agentResult.error) {
+            console.log(`        error : ${agentResult.error}`);
+        }
+    }
+    // ── Assertion ──
+    console.log('');
+    if (passRate >= MIN_PASS_RATE) {
+        console.log(`\x1b[32mSmoke PASSED\x1b[0m — ${passed}/${total} ≥ ${(MIN_PASS_RATE * 100).toFixed(0)}% required.\n`);
+    }
+    else {
+        console.error(`\x1b[31mSmoke FAILED\x1b[0m — ${passed}/${total} < ${(MIN_PASS_RATE * 100).toFixed(0)}% required.\n`);
+        process.exit(1);
+    }
+}
+// ---------------------------------------------------------------------------
+// Entry point
+// ---------------------------------------------------------------------------
+const isMain = process.argv[1] && (process.argv[1].endsWith('gaia-e2e-smoke.ts') ||
+    process.argv[1].endsWith('gaia-e2e-smoke.js'));
+if (isMain) {
+    runE2ESmoke().catch((err) => {
+        console.error('E2E smoke failed:', err);
+        process.exit(1);
+    });
+}
+export { runE2ESmoke };
+//# sourceMappingURL=gaia-e2e-smoke.js.map

package/dist/src/benchmarks/gaia-e2e-smoke.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"gaia-e2e-smoke.js","sourceRoot":"","sources":["../../../src/benchmarks/gaia-e2e-smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AAEH,OAAO,KAAK,EAAE,MAAM,SAAS,CAAC;AAC9B,OAAO,KAAK,IAAI,MAAM,WAAW,CAAC;AAClC,OAAO,EACL,aAAa,GAEd,MAAM,kBAAkB,CAAC;AAC1B,OAAO,EACL,YAAY,GAEb,MAAM,iBAAiB,CAAC;AACzB,OAAO,EACL,WAAW,GAGZ,MAAM,iBAAiB,CAAC;AAEzB,8EAA8E;AAC9E,gBAAgB;AAChB,8EAA8E;AAE9E,oDAAoD;AACpD,MAAM,WAAW,GAAG,kBAAkB,CAAC;AAEvC,iFAAiF;AACjF,MAAM,WAAW,GAAG,mBAAmB,CAAC;AAExC,8CAA8C;AAC9C,MAAM,aAAa,GAAG,CAAC,GAAG,CAAC,CAAC;AAE5B,6BAA6B;AAC7B,MAAM,QAAQ,GAAG,IAAI,CAAC;AACtB,MAAM,SAAS,GAAG,IAAI,CAAC;AAEvB,8BAA8B;AAC9B,MAAM,SAAS,GAAG,GAAG,CAAC;AACtB,MAAM,UAAU,GAAG,IAAI,CAAC;AAYxB,8EAA8E;AAC9E,SAAS;AACT,8EAA8E;AAE9E,KAAK,UAAU,WAAW;IACxB,MAAM,MAAM,GAAG,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,iBAAiB,EAAE,IAAI,EAAE,CAAC,CAAC;IACzD,IAAI,CAAC,MAAM,EAAE,CAAC;QACZ,OAAO,CAAC,KAAK,CACX,2DAA2D;YAC3D,kDAAkD,CACnD,CAAC;QACF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,MAAM,QAAQ,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,OAAO,EAAE,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,EAAE,WAAW,CAAC,CAAC;IACjF,MAAM,SAAS,GAAiB,EAAE,UAAU,EAAE,WAAW,EAAE,QAAQ,EAAE,CAAC;IAEtE,MAAM,SAAS,GAAG,aAAa,CAAC;IAChC,MAAM,IAAI,GAAa,EAAE,CAAC;IAE1B,OAAO,CAAC,GAAG,CAAC,gCAAgC,SAAS,CAAC,MAAM,mBAAmB,CAAC,CAAC;IACjF,OAAO,CAAC,GAAG,CACT,UAAU,WAAW,eAAe,WAAW,IAAI;QACnD,cAAc,SAAS,CAAC,MAAM,uBAAuB,CAAC,aAAa,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,CAC3F,CAAC;IAEF,KAAK,MAAM,CAAC,IAAI,SAAS,EAAE,CAAC;QAC1B,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,OAAO,MAAM,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,OAAO,CAAC,CAAC;QAExE,YAAY;QACZ,MAAM,WAAW,GAAG,MAAM,YAAY,CAAC,CAAC,EAAE,EAAE,KAAK,EAAE,WAAW,EAAE,CAAC,CAAC;QAElE,QAAQ;QACR,MAAM,WAAW,GAAG,MAAM,WAAW,CACnC,EAAE,EAAE,EAAE,CAAC,CAAC,OAAO,EAAE,QAAQ,EAAE,CAAC,CAAC,YAAY,EAAE,YAAY,EAAE,CAAC,CAAC,QAAQ,EAAE,EACrE,WAAW,CAAC,WAAW,EACvB,SAAS,CACV,CAAC;QAEF,IAAI,CAAC,IAAI,CAAC,EAAE,QAAQ,EAAE,CAAC,EAAE,WAAW,EAAE,WAAW,EAAE,CAAC,CAAC;QAErD,MAAM,OAAO,GAAG,WAAW,CAAC,MAAM,CAAC,CAAC,CAAC,qBAAqB,CAAC,CAAC,CAAC,qBAAqB,CAAC;QACnF,MAAM,KAAK,GAAG,WAAW,CAAC,WAAW,CAAC;QACtC,OAAO,CAAC,GAAG,CACT,GAAG,OAAO,YAAY,WAAW,CAAC,KAAK,IAAI;YAC3C,WAAW,WAAW,CAAC,WAAW,IAAI,MAAM,KAAK;YACjD,aAAa,CAAC,CAAC,YAAY,WAAW,KAAK,GAAG,CAC/C,CAAC;IACJ,CAAC;IAED,gBAAgB;IAChB,OAAO,CAAC,GAAG,CAAC,qBAAqB,CAAC,CAAC;IAEnC,MAAM,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC;IAC/D,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC;IAC1B,MAAM,QAAQ,GAAG,MAAM,GAAG,KAAK,CAAC;IAEhC,IAAI,iBAAiB,GAAG,CAAC,CAAC;IAC1B,IAAI,iBAAiB,GAAG,CAAC,CAAC;IAC1B,IAAI,UAAU,GAAG,CAAC,CAAC;IAEnB,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,EAAE,WAAW,EAAE,WAAW,EAAE,GAAG,GAAG,CAAC;QACzC,UAAU,IAAI,WAAW,CAAC,KAAK,CAAC;QAChC,iBAAiB;YACf,CAAC,WAAW,CAAC,gBAAgB,GAAG,SAAS,CAAC,GAAG,QAAQ;gBACrD,CAAC,WAAW,CAAC,iBAAiB,GAAG,SAAS,CAAC,GAAG,SAAS,CAAC;QAC1D,iBAAiB,IAAI,WAAW,CAAC,YAAY,IAAI,CAAC,CAAC;IACrD,CAAC;IAED,MAAM,YAAY,GAAG,iBAAiB,GAAG,iBAAiB,CAAC;IAC3D,MAAM,SAAS,GAAG,UAAU,GAAG,KAAK,CAAC;IAErC,OAAO,CAAC,GAAG,CAAC,oBAAoB,MAAM,IAAI,KAAK,MAAM,CAAC,QAAQ,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IACtF,OAAO,CAAC,GAAG,CAAC,oBAAoB,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IACxD,OAAO,CAAC,GAAG,CAAC,qBAAqB,iBAAiB,CAAC,OAAO,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC;IAC1E,OAAO,CAAC,GAAG,CAAC,qBAAqB,iBAAiB,CAAC,OAAO,CAAC,CAAC,CAAC,8BAA8B,CAAC,CAAC;IAC7F,OAAO,CAAC,GAAG,CAAC,qBAAqB,YAAY,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAE5D,uBAAuB;IACvB,OAAO,CAAC,GAAG,CAAC,iCAAiC,CAAC,CAAC;IAC/C,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,EAAE,QAAQ,EAAE,WAAW,EAAE,WAAW,EAAE,GAAG,GAAG,CAAC;QACnD,MAAM,OAAO,GAAG,WAAW,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC;QACrD,OAAO,CAAC,GAAG,CACT,KAAK,OAAO,KAAK,QAAQ,CAAC,OAAO,WAAW,WAAW,CAAC,KAAK,IAAI;YACjE,QAAQ,WAAW,CAAC,WAAW,IAAI;YACnC,WAAW,WAAW,CAAC,WAAW,IAAI,MAAM,KAAK;YACjD,aAAa,QAAQ,CAAC,YAAY,GAAG,CACtC,CAAC;QACF,IAAI,WAAW,CAAC,WAAW,EAAE,CAAC;YAC5B,OAAO,CAAC,GAAG,CAAC,mBAAmB,WAAW,CAAC,WAAW,EAAE,CAAC,CAAC;QAC5D,CAAC;QACD,IAAI,WAAW,CAAC,KAAK,EAAE,CAAC;YACtB,OAAO,CAAC,GAAG,CAAC,mBAAmB,WAAW,CAAC,KAAK,EAAE,CAAC,CAAC;QACtD,CAAC;IACH,CAAC;IAED,kBAAkB;IAClB,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;IAChB,IAAI,QAAQ,IAAI,aAAa,EAAE,CAAC;QAC9B,OAAO,CAAC,GAAG,CACT,iCAAiC,MAAM,IAAI,KAAK,MAAM,CAAC,aAAa,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,eAAe,CACtG,CAAC;IACJ,CAAC;SAAM,CAAC;QACN,OAAO,CAAC,KAAK,CACX,iCAAiC,MAAM,IAAI,KAAK,MAAM,CAAC,aAAa,GAAG,GAAG,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,eAAe,CACtG,CAAC;QACF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;AACH,CAAC;AAED,8EAA8E;AAC9E,cAAc;AACd,8EAA8E;AAE9E,MAAM,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAChC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,mBAAmB,CAAC;IAC7C,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,mBAAmB,CAAC,CAC9C,CAAC;AACF,IAAI,MAAM,EAAE,CAAC;IACX,WAAW,EAAE,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,EAAE;QAC1B,OAAO,CAAC,KAAK,CAAC,mBAAmB,EAAE,GAAG,CAAC,CAAC;QACxC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC,CAAC,CAAC;AACL,CAAC;AAED,OAAO,EAAE,WAAW,EAAE,CAAC"}

package/dist/src/benchmarks/gaia-extract.smoke.d.ts

@@ -0,0 +1,45 @@
+/**
+ * Smoke tests for extractFinalAnswer + buildUserMessage (iter 53a T2 narrowed).
+ *
+ * Iter 53a changes vs iter 52 T2:
+ *   - Stage 2/3 (prose fallback) REMOVED — extractFinalAnswer is Stage 1 only.
+ *   - Commitment prompt no longer has "FINAL_ANSWER: unknown" surrender instruction.
+ *   - Reversed-text preprocessor PRESERVED (not the source of regressions).
+ *
+ * Anti-regression suite (7 cases):
+ *   These are the exact failure modes from iter 52 T2 that must not regress.
+ *   They test that the extraction logic is stable and correct for the 7 questions
+ *   where iter 52 broke but iter 51 was correct.
+ *
+ * Run (after build):
+ *   node dist/benchmarks/gaia-extract.smoke.js
+ *
+ * Exit 0 on all pass, 1 on any failure.
+ *
+ * Refs: ADR-133, ADR-135, iter 52 T2, iter 53a, #2156
+ */
+interface AnthropicResponse {
+    id: string;
+    model: string;
+    stop_reason: string;
+    content: Array<{
+        type: string;
+        text?: string;
+    }>;
+    usage: {
+        input_tokens: number;
+        output_tokens: number;
+    };
+}
+declare let _extractFinalAnswer: (resp: AnthropicResponse) => string | null;
+declare let _buildUserMessage: (question: string) => string;
+declare function makeResp(text: string): AnthropicResponse;
+interface TestCase {
+    label: string;
+    input: string;
+    expected: string | null;
+    isUserMsg?: boolean;
+}
+declare const CASES: TestCase[];
+declare function main(): Promise<void>;
+//# sourceMappingURL=gaia-extract.smoke.d.ts.map

package/dist/src/benchmarks/gaia-extract.smoke.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"gaia-extract.smoke.d.ts","sourceRoot":"","sources":["../../../src/benchmarks/gaia-extract.smoke.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAOH,UAAU,iBAAiB;IACzB,EAAE,EAAE,MAAM,CAAC;IACX,KAAK,EAAE,MAAM,CAAC;IACd,WAAW,EAAE,MAAM,CAAC;IACpB,OAAO,EAAE,KAAK,CAAC;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,IAAI,CAAC,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;IAChD,KAAK,EAAE;QAAE,YAAY,EAAE,MAAM,CAAC;QAAC,aAAa,EAAE,MAAM,CAAA;KAAE,CAAC;CACxD;AAID,QAAA,IAAI,mBAAmB,EAAE,CAAC,IAAI,EAAE,iBAAiB,KAAK,MAAM,GAAG,IAAI,CAAC;AAEpE,QAAA,IAAI,iBAAiB,EAAE,CAAC,QAAQ,EAAE,MAAM,KAAK,MAAM,CAAC;AAEpD,iBAAS,QAAQ,CAAC,IAAI,EAAE,MAAM,GAAG,iBAAiB,CAQjD;AAOD,UAAU,QAAQ;IAChB,KAAK,EAAE,MAAM,CAAC;IACd,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,EAAE,MAAM,GAAG,IAAI,CAAC;IACxB,SAAS,CAAC,EAAE,OAAO,CAAC;CACrB;AAED,QAAA,MAAM,KAAK,EAAE,QAAQ,EAuJpB,CAAC;AAMF,iBAAe,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,CA8DnC"}

package/dist/src/benchmarks/gaia-extract.smoke.js

@@ -0,0 +1,242 @@
+"use strict";
+/**
+ * Smoke tests for extractFinalAnswer + buildUserMessage (iter 53a T2 narrowed).
+ *
+ * Iter 53a changes vs iter 52 T2:
+ *   - Stage 2/3 (prose fallback) REMOVED — extractFinalAnswer is Stage 1 only.
+ *   - Commitment prompt no longer has "FINAL_ANSWER: unknown" surrender instruction.
+ *   - Reversed-text preprocessor PRESERVED (not the source of regressions).
+ *
+ * Anti-regression suite (7 cases):
+ *   These are the exact failure modes from iter 52 T2 that must not regress.
+ *   They test that the extraction logic is stable and correct for the 7 questions
+ *   where iter 52 broke but iter 51 was correct.
+ *
+ * Run (after build):
+ *   node dist/benchmarks/gaia-extract.smoke.js
+ *
+ * Exit 0 on all pass, 1 on any failure.
+ *
+ * Refs: ADR-133, ADR-135, iter 52 T2, iter 53a, #2156
+ */
+// Dynamic import so this file can run either as TS source (tsx) or compiled JS.
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+let _extractFinalAnswer;
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+let _buildUserMessage;
+function makeResp(text) {
+    return {
+        id: 'test',
+        model: 'test',
+        stop_reason: 'end_turn',
+        content: [{ type: 'text', text }],
+        usage: { input_tokens: 0, output_tokens: 0 },
+    };
+}
+const CASES = [
+    // ---------------------------------------------------------------------------
+    // Stage 1: primary FINAL_ANSWER: pattern (iter 53a: this is the ONLY stage)
+    // ---------------------------------------------------------------------------
+    {
+        label: 'Stage1 basic FINAL_ANSWER:',
+        input: 'I searched and found the capital.\nFINAL_ANSWER: Paris',
+        expected: 'Paris',
+    },
+    {
+        label: 'Stage1 case-insensitive final_answer:',
+        input: 'The result is known.\nfinal_answer: 42',
+        expected: '42',
+    },
+    {
+        label: 'Stage1 with leading whitespace',
+        input: 'Done.\n   FINAL_ANSWER:   Tokyo   ',
+        expected: 'Tokyo',
+    },
+    // ---------------------------------------------------------------------------
+    // Prose-only outputs with no FINAL_ANSWER: tag — must return null (iter 53a).
+    // These were previously handled by Stage 2/3 (iter 52 T2).
+    // In iter 53a, we rely on the agent committing with the tag instead.
+    // ---------------------------------------------------------------------------
+    {
+        label: 'Prose-only "The answer is X" — null in iter 53a (no Stage 2)',
+        input: 'After analysis, the answer is Right.',
+        expected: null,
+    },
+    {
+        label: 'Prose-only "Therefore X" — null in iter 53a (no Stage 2)',
+        input: 'Based on the clues, therefore the answer is Berlin.',
+        expected: null,
+    },
+    {
+        label: 'Prose-only "Answer: X" — null in iter 53a (no Stage 2)',
+        input: 'Let me compute this.\nAnswer: 17',
+        expected: null,
+    },
+    {
+        label: 'Prose-only all-caps last line — null in iter 53a (no Stage 3)',
+        input: 'I computed the value based on many steps.\nRIGHT',
+        expected: null,
+    },
+    {
+        label: 'Prose-only numeric last line — null in iter 53a (no Stage 3)',
+        input: 'The final calculation gives us:\n346',
+        expected: null,
+    },
+    {
+        label: 'Prose-only short-phrase last line — null in iter 53a (no Stage 3)',
+        input: 'After extensive research, the result is:\nBerlin, Germany',
+        expected: null,
+    },
+    // Null case: no answer extractable (verbose prose, no commitment)
+    {
+        label: 'Null case: only tool-call reasoning, no commitment',
+        input: 'I tried searching but could not find the specific information requested.',
+        expected: null,
+    },
+    // ---------------------------------------------------------------------------
+    // Reversed text pre-processor (buildUserMessage) — PRESERVED in iter 53a
+    // ---------------------------------------------------------------------------
+    {
+        label: 'Reversed text: adds decoded hint',
+        input: '.rewsna eht sa "tfel" drow eht fo etisoppo eht etirw ,ecnetnes siht dnatsrednu uoy fI',
+        expected: '[NOTE:',
+        isUserMsg: true,
+    },
+    {
+        label: 'Normal text: no hint added',
+        input: 'What is the capital of France?',
+        expected: 'What is the capital of France?',
+        isUserMsg: true,
+    },
+    // ---------------------------------------------------------------------------
+    // Anti-regression suite (iter 53a) — 7 cases for iter 52 regressions.
+    //
+    // These represent the exact regression patterns from iter 52 T2:
+    //   a1e91b78: agent surrendered with "unknown" (commitment prompt over-trigger)
+    //   935e2cff: agent gave empty answer (Stage 2/3 failed after extraction)
+    //   305ac316: agent gave empty answer (Stage 2/3 failed after extraction)
+    //   7673d772: agent gave empty answer (Stage 2/3 failed after extraction)
+    //   3f57289b: Stage 3 grabbed wrong number from prose
+    //   50ec8903: Stage 3 grabbed markdown fragment instead of correct answer
+    //   5a0c1adf: Stage 3 grabbed "Claus Peter" instead of correct "Claus"
+    //
+    // Each test constructs a synthetic response that mimics the failure mode
+    // and asserts the correct Stage 1 extraction behavior.
+    // ---------------------------------------------------------------------------
+    // AR-1: a1e91b78 — agent used to surrender with "FINAL_ANSWER: unknown"
+    //   In iter 52, system prompt said to output "FINAL_ANSWER: unknown" if unsure.
+    //   With that instruction removed, the agent reasons further and finds the answer.
+    //   This test verifies: if the agent DOES output a correct FINAL_ANSWER, we extract it.
+    {
+        label: 'AR-1 (a1e91b78): Stage1 extracts correct answer — not "unknown"',
+        input: 'I found the video. The maximum number of bird species seen together is 3.\nFINAL_ANSWER: 3',
+        expected: '3',
+    },
+    // AR-2: 935e2cff — agent had the answer but Stage 2 produced empty string
+    //   In iter 53a, agent should output FINAL_ANSWER: Research correctly.
+    {
+        label: 'AR-2 (935e2cff): Stage1 extracts "Research" correctly',
+        input: 'After reviewing the policy document, R stands for Research.\nFINAL_ANSWER: Research',
+        expected: 'Research',
+    },
+    // AR-3: 305ac316 — agent had "Wojciech" but extraction failed
+    {
+        label: 'AR-3 (305ac316): Stage1 extracts Polish name correctly',
+        input: 'The actor played Wojciech in the movie.\nFINAL_ANSWER: Wojciech',
+        expected: 'Wojciech',
+    },
+    // AR-4: 7673d772 — agent had "inference" but extraction failed
+    {
+        label: 'AR-4 (7673d772): Stage1 extracts legal term correctly',
+        input: 'The fifth section of the federal rule uses the word "inference".\nFINAL_ANSWER: inference',
+        expected: 'inference',
+    },
+    // AR-5: 3f57289b — Stage 3 grabbed wrong number (525 instead of 519)
+    //   The regression was caused by Stage 3 picking a wrong number from prose.
+    //   With Stage 3 removed, Stage 1 is the only source — agent must use the tag.
+    {
+        label: 'AR-5 (3f57289b): Stage1 extracts correct at-bat count',
+        input: 'Reggie Jackson had 525 at-bats but the player with most walks had 519.\nFINAL_ANSWER: 519',
+        expected: '519',
+    },
+    // AR-6: 50ec8903 — Stage 3 grabbed "- Orange-Green edge →" (markdown fragment)
+    //   With Stage 3 removed, only FINAL_ANSWER: tag is extracted.
+    {
+        label: 'AR-6 (50ec8903): Stage1 extracts Rubik cube colors correctly',
+        input: 'After solving the cube configuration, the answer is green, white.\nFINAL_ANSWER: green, white',
+        expected: 'green, white',
+    },
+    // AR-7: 5a0c1adf — Stage 3 grabbed "Claus Peter" (too many words)
+    //   With Stage 3 removed, only FINAL_ANSWER: tag is extracted.
+    {
+        label: 'AR-7 (5a0c1adf): Stage1 extracts first name only, not "Claus Peter"',
+        input: 'The conductor\'s full name is Claus Peter Flor, but his first name is Claus.\nFINAL_ANSWER: Claus',
+        expected: 'Claus',
+    },
+];
+// ---------------------------------------------------------------------------
+// Runner
+// ---------------------------------------------------------------------------
+async function main() {
+    // Import at runtime to work with both tsx and compiled JS.
+    const mod = await import('./gaia-agent.js');
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    const m = mod;
+    _extractFinalAnswer = m._extractFinalAnswerForTest;
+    _buildUserMessage = m._buildUserMessageForTest;
+    if (typeof _extractFinalAnswer !== 'function' || typeof _buildUserMessage !== 'function') {
+        console.error('ERROR: _extractFinalAnswerForTest / _buildUserMessageForTest not exported from gaia-agent.' +
+            '\nAdd `export { extractFinalAnswer as _extractFinalAnswerForTest, ' +
+            'buildUserMessage as _buildUserMessageForTest }` to gaia-agent.ts.');
+        process.exit(1);
+    }
+    let failures = 0;
+    const PASS = '\x1b[32mPASS\x1b[0m';
+    const FAIL = '\x1b[31mFAIL\x1b[0m';
+    console.log('\n=== gaia-extract smoke (iter 53a T2 narrowed) ===\n');
+    console.log('Suite: 12 original + 7 anti-regression = 19 cases\n');
+    for (const tc of CASES) {
+        let actual;
+        if (tc.isUserMsg) {
+            actual = _buildUserMessage(tc.input);
+            // For user message tests, check startsWith instead of exact equality.
+            const pass = tc.expected === null
+                ? actual === null
+                : actual !== null && (tc.expected === actual || actual.startsWith(tc.expected));
+            if (pass) {
+                console.log(`  ${PASS}  ${tc.label}`);
+            }
+            else {
+                console.log(`  ${FAIL}  ${tc.label}`);
+                console.log(`         expected starts-with: ${JSON.stringify(tc.expected)}`);
+                console.log(`         actual:               ${JSON.stringify((actual ?? '').slice(0, 80))}`);
+                failures++;
+            }
+        }
+        else {
+            actual = _extractFinalAnswer(makeResp(tc.input));
+            const pass = tc.expected === null
+                ? actual === null
+                : actual !== null && actual === tc.expected;
+            if (pass) {
+                console.log(`  ${PASS}  ${tc.label}`);
+            }
+            else {
+                console.log(`  ${FAIL}  ${tc.label}`);
+                console.log(`         expected: ${JSON.stringify(tc.expected)}`);
+                console.log(`         actual:   ${JSON.stringify(actual)}`);
+                failures++;
+            }
+        }
+    }
+    const total = CASES.length;
+    const passed = total - failures;
+    console.log(`\n=== ${failures === 0 ? 'ALL PASSED' : `${failures} FAILED`} (${passed}/${total} cases) ===\n`);
+    if (failures > 0)
+        process.exit(1);
+}
+main().catch((err) => {
+    console.error('Smoke crashed:', err);
+    process.exit(2);
+});
+//# sourceMappingURL=gaia-extract.smoke.js.map

package/dist/src/benchmarks/gaia-extract.smoke.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"gaia-extract.smoke.js","sourceRoot":"","sources":["../../../src/benchmarks/gaia-extract.smoke.ts"],"names":[],"mappings":";AAAA;;;;;;;;;;;;;;;;;;;GAmBG;AAeH,gFAAgF;AAChF,8DAA8D;AAC9D,IAAI,mBAA+D,CAAC;AACpE,8DAA8D;AAC9D,IAAI,iBAA+C,CAAC;AAEpD,SAAS,QAAQ,CAAC,IAAY;IAC5B,OAAO;QACL,EAAE,EAAE,MAAM;QACV,KAAK,EAAE,MAAM;QACb,WAAW,EAAE,UAAU;QACvB,OAAO,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC;QACjC,KAAK,EAAE,EAAE,YAAY,EAAE,CAAC,EAAE,aAAa,EAAE,CAAC,EAAE;KAC7C,CAAC;AACJ,CAAC;AAcD,MAAM,KAAK,GAAe;IACxB,8EAA8E;IAC9E,4EAA4E;IAC5E,8EAA8E;IAC9E;QACE,KAAK,EAAE,4BAA4B;QACnC,KAAK,EAAE,wDAAwD;QAC/D,QAAQ,EAAE,OAAO;KAClB;IACD;QACE,KAAK,EAAE,uCAAuC;QAC9C,KAAK,EAAE,wCAAwC;QAC/C,QAAQ,EAAE,IAAI;KACf;IACD;QACE,KAAK,EAAE,gCAAgC;QACvC,KAAK,EAAE,oCAAoC;QAC3C,QAAQ,EAAE,OAAO;KAClB;IAED,8EAA8E;IAC9E,8EAA8E;IAC9E,2DAA2D;IAC3D,qEAAqE;IACrE,8EAA8E;IAC9E;QACE,KAAK,EAAE,8DAA8D;QACrE,KAAK,EAAE,sCAAsC;QAC7C,QAAQ,EAAE,IAAI;KACf;IACD;QACE,KAAK,EAAE,0DAA0D;QACjE,KAAK,EAAE,qDAAqD;QAC5D,QAAQ,EAAE,IAAI;KACf;IACD;QACE,KAAK,EAAE,wDAAwD;QAC/D,KAAK,EAAE,kCAAkC;QACzC,QAAQ,EAAE,IAAI;KACf;IACD;QACE,KAAK,EAAE,+DAA+D;QACtE,KAAK,EAAE,kDAAkD;QACzD,QAAQ,EAAE,IAAI;KACf;IACD;QACE,KAAK,EAAE,8DAA8D;QACrE,KAAK,EAAE,sCAAsC;QAC7C,QAAQ,EAAE,IAAI;KACf;IACD;QACE,KAAK,EAAE,mEAAmE;QAC1E,KAAK,EAAE,2DAA2D;QAClE,QAAQ,EAAE,IAAI;KACf;IAED,kEAAkE;IAClE;QACE,KAAK,EAAE,oDAAoD;QAC3D,KAAK,EAAE,0EAA0E;QACjF,QAAQ,EAAE,IAAI;KACf;IAED,8EAA8E;IAC9E,yEAAyE;IACzE,8EAA8E;IAC9E;QACE,KAAK,EAAE,kCAAkC;QACzC,KAAK,EAAE,uFAAuF;QAC9F,QAAQ,EAAE,QAAQ;QAClB,SAAS,EAAE,IAAI;KAChB;IACD;QACE,KAAK,EAAE,4BAA4B;QACnC,KAAK,EAAE,gCAAgC;QACvC,QAAQ,EAAE,gCAAgC;QAC1C,SAAS,EAAE,IAAI;KAChB;IAED,8EAA8E;IAC9E,sEAAsE;IACtE,EAAE;IACF,iEAAiE;IACjE,gFAAgF;IAChF,0EAA0E;IAC1E,0EAA0E;IAC1E,0EAA0E;IAC1E,sDAAsD;IACtD,0EAA0E;IAC1E,uEAAuE;IACvE,EAAE;IACF,yEAAyE;IACzE,uDAAuD;IACvD,8EAA8E;IAE9E,wEAAwE;IACxE,gFAAgF;IAChF,mFAAmF;IACnF,wFAAwF;IACxF;QACE,KAAK,EAAE,iEAAiE;QACxE,KAAK,EAAE,4FAA4F;QACnG,QAAQ,EAAE,GAAG;KACd;IAED,0EAA0E;IAC1E,uEAAuE;IACvE;QACE,KAAK,EAAE,uDAAuD;QAC9D,KAAK,EAAE,qFAAqF;QAC5F,QAAQ,EAAE,UAAU;KACrB;IAED,8DAA8D;IAC9D;QACE,KAAK,EAAE,wDAAwD;QAC/D,KAAK,EAAE,iEAAiE;QACxE,QAAQ,EAAE,UAAU;KACrB;IAED,+DAA+D;IAC/D;QACE,KAAK,EAAE,uDAAuD;QAC9D,KAAK,EAAE,2FAA2F;QAClG,QAAQ,EAAE,WAAW;KACtB;IAED,qEAAqE;IACrE,4EAA4E;IAC5E,+EAA+E;IAC/E;QACE,KAAK,EAAE,uDAAuD;QAC9D,KAAK,EAAE,2FAA2F;QAClG,QAAQ,EAAE,KAAK;KAChB;IAED,+EAA+E;IAC/E,+DAA+D;IAC/D;QACE,KAAK,EAAE,8DAA8D;QACrE,KAAK,EAAE,+FAA+F;QACtG,QAAQ,EAAE,cAAc;KACzB;IAED,kEAAkE;IAClE,+DAA+D;IAC/D;QACE,KAAK,EAAE,qEAAqE;QAC5E,KAAK,EAAE,mGAAmG;QAC1G,QAAQ,EAAE,OAAO;KAClB;CACF,CAAC;AAEF,8EAA8E;AAC9E,SAAS;AACT,8EAA8E;AAE9E,KAAK,UAAU,IAAI;IACjB,2DAA2D;IAC3D,MAAM,GAAG,GAAG,MAAM,MAAM,CAAC,iBAAiB,CAAC,CAAC;IAC5C,8DAA8D;IAC9D,MAAM,CAAC,GAAG,GAAU,CAAC;IAErB,mBAAmB,GAAG,CAAC,CAAC,0BAA0B,CAAC;IACnD,iBAAiB,GAAG,CAAC,CAAC,wBAAwB,CAAC;IAE/C,IAAI,OAAO,mBAAmB,KAAK,UAAU,IAAI,OAAO,iBAAiB,KAAK,UAAU,EAAE,CAAC;QACzF,OAAO,CAAC,KAAK,CACX,4FAA4F;YAC5F,oEAAoE;YACpE,mEAAmE,CACpE,CAAC;QACF,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAClB,CAAC;IAED,IAAI,QAAQ,GAAG,CAAC,CAAC;IACjB,MAAM,IAAI,GAAG,qBAAqB,CAAC;IACnC,MAAM,IAAI,GAAG,qBAAqB,CAAC;IAEnC,OAAO,CAAC,GAAG,CAAC,uDAAuD,CAAC,CAAC;IACrE,OAAO,CAAC,GAAG,CAAC,qDAAqD,CAAC,CAAC;IAEnE,KAAK,MAAM,EAAE,IAAI,KAAK,EAAE,CAAC;QACvB,IAAI,MAAqB,CAAC;QAE1B,IAAI,EAAE,CAAC,SAAS,EAAE,CAAC;YACjB,MAAM,GAAG,iBAAiB,CAAC,EAAE,CAAC,KAAK,CAAC,CAAC;YACrC,sEAAsE;YACtE,MAAM,IAAI,GAAG,EAAE,CAAC,QAAQ,KAAK,IAAI;gBAC/B,CAAC,CAAC,MAAM,KAAK,IAAI;gBACjB,CAAC,CAAC,MAAM,KAAK,IAAI,IAAI,CAAC,EAAE,CAAC,QAAQ,KAAK,MAAM,IAAI,MAAM,CAAC,UAAU,CAAC,EAAE,CAAC,QAAQ,CAAC,CAAC,CAAC;YAClF,IAAI,IAAI,EAAE,CAAC;gBACT,OAAO,CAAC,GAAG,CAAC,KAAK,IAAI,KAAK,EAAE,CAAC,KAAK,EAAE,CAAC,CAAC;YACxC,CAAC;iBAAM,CAAC;gBACN,OAAO,CAAC,GAAG,CAAC,KAAK,IAAI,KAAK,EAAE,CAAC,KAAK,EAAE,CAAC,CAAC;gBACtC,OAAO,CAAC,GAAG,CAAC,kCAAkC,IAAI,CAAC,SAAS,CAAC,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;gBAC7E,OAAO,CAAC,GAAG,CAAC,kCAAkC,IAAI,CAAC,SAAS,CAAC,CAAC,MAAM,IAAI,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC;gBAC7F,QAAQ,EAAE,CAAC;YACb,CAAC;QACH,CAAC;aAAM,CAAC;YACN,MAAM,GAAG,mBAAmB,CAAC,QAAQ,CAAC,EAAE,CAAC,KAAK,CAAC,CAAC,CAAC;YACjD,MAAM,IAAI,GAAG,EAAE,CAAC,QAAQ,KAAK,IAAI;gBAC/B,CAAC,CAAC,MAAM,KAAK,IAAI;gBACjB,CAAC,CAAC,MAAM,KAAK,IAAI,IAAI,MAAM,KAAK,EAAE,CAAC,QAAQ,CAAC;YAC9C,IAAI,IAAI,EAAE,CAAC;gBACT,OAAO,CAAC,GAAG,CAAC,KAAK,IAAI,KAAK,EAAE,CAAC,KAAK,EAAE,CAAC,CAAC;YACxC,CAAC;iBAAM,CAAC;gBACN,OAAO,CAAC,GAAG,CAAC,KAAK,IAAI,KAAK,EAAE,CAAC,KAAK,EAAE,CAAC,CAAC;gBACtC,OAAO,CAAC,GAAG,CAAC,sBAAsB,IAAI,CAAC,SAAS,CAAC,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;gBACjE,OAAO,CAAC,GAAG,CAAC,sBAAsB,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,EAAE,CAAC,CAAC;gBAC5D,QAAQ,EAAE,CAAC;YACb,CAAC;QACH,CAAC;IACH,CAAC;IAED,MAAM,KAAK,GAAG,KAAK,CAAC,MAAM,CAAC;IAC3B,MAAM,MAAM,GAAG,KAAK,GAAG,QAAQ,CAAC;IAChC,OAAO,CAAC,GAAG,CAAC,SAAS,QAAQ,KAAK,CAAC,CAAC,CAAC,CAAC,YAAY,CAAC,CAAC,CAAC,GAAG,QAAQ,SAAS,KAAK,MAAM,IAAI,KAAK,eAAe,CAAC,CAAC;IAC9G,IAAI,QAAQ,GAAG,CAAC;QAAE,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AACpC,CAAC;AAED,IAAI,EAAE,CAAC,KAAK,CAAC,CAAC,GAAG,EAAE,EAAE;IACnB,OAAO,CAAC,KAAK,CAAC,gBAAgB,EAAE,GAAG,CAAC,CAAC;IACrC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;AAClB,CAAC,CAAC,CAAC"}

package/dist/src/benchmarks/gaia-hardness/features.d.ts

@@ -0,0 +1,46 @@
+/**
+ * GAIA Hardness Predictor — Feature Extraction (ADR-136 Track Q)
+ *
+ * Extracts a 17-dimensional feature vector from a GaiaQuestion for use
+ * in the linear hardness classifier.  Deliberately avoids external
+ * dependencies (no spacy, no heavy NLP) — all extraction is regex-based
+ * with O(1) per question.
+ *
+ * Feature vector layout (17 dims):
+ *   [0]  question length in characters (normalised / 500)
+ *   [1]  question length in words      (normalised / 100)
+ *   [2]  sentence count                (normalised / 5)
+ *   [3]  question word: "what"         (0/1)
+ *   [4]  question word: "how"          (0/1)
+ *   [5]  question word: "who/when/where" (0/1)
+ *   [6]  question word: "calculate/compute/how many/what percentage" (0/1)
+ *   [7]  has numeric token             (0/1)
+ *   [8]  has year token (4-digit)      (0/1)
+ *   [9]  has comparison keyword        (0/1)
+ *   [10] estimated named-entity count  (normalised / 5)
+ *   [11] digit-token count             (normalised / 5)
+ *   [12] multi-hop signal              (0/1)  — bridge/relative clause pattern
+ *   [13] requires math                 (0/1)  — "how many/much/percentage/calculate/compute"
+ *   [14] temporal chain                (0/1)  — "before/after X happened / since / until"
+ *   [15] tool implication count        (normalised / 4)  — PDF/image/video/URL markers
+ *   [16] file attachment present       (0/1)
+ *
+ * All continuous values are min-max normalised to [0, 1] using fixed
+ * divisors chosen so typical GAIA questions fall in [0.1, 0.9].
+ *
+ * Refs: ADR-136, #2156
+ */
+import type { GaiaQuestion } from '../gaia-loader.js';
+export interface FeatureVector {
+    /** Raw 17-element array in [0, 1]. */
+    values: number[];
+    /** Human-readable labels matching values index. */
+    labels: string[];
+}
+export declare const FEATURE_LABELS: readonly string[];
+/**
+ * Extract the 17-dimensional feature vector from a GaiaQuestion.
+ * All features are in [0, 1].  Never throws.
+ */
+export declare function extractFeatures(q: GaiaQuestion): FeatureVector;
+//# sourceMappingURL=features.d.ts.map

package/dist/src/benchmarks/gaia-hardness/features.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"features.d.ts","sourceRoot":"","sources":["../../../../src/benchmarks/gaia-hardness/features.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA+BG;AAEH,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAC;AAMtD,MAAM,WAAW,aAAa;IAC5B,sCAAsC;IACtC,MAAM,EAAE,MAAM,EAAE,CAAC;IACjB,mDAAmD;IACnD,MAAM,EAAE,MAAM,EAAE,CAAC;CAClB;AAMD,eAAO,MAAM,cAAc,EAAE,SAAS,MAAM,EAkBlC,CAAC;AA+CX;;;GAGG;AACH,wBAAgB,eAAe,CAAC,CAAC,EAAE,YAAY,GAAG,aAAa,CAmF9D"}