homegraph 1.1.2 → 1.1.3

package/scripts/agent-eval/parse-arms.mjs

@@ -1,116 +1,116 @@
-#!/usr/bin/env node
-// Analyze the tool-surface ablation (/tmp/arms/<repo>/<arm>-r<n>.jsonl).
-// Compares arms A–E on trace adoption, Read/Grep fallback, homegraph payload,
-// round-trips, and duration — averaged across runs per arm.
-//
-// The decisive signal is READS: if removing a tool raises Reads on a question
-// class, that tool was load-bearing for it (not redundant). If removing it
-// changes nothing, it was redundant.
-//
-//   A control       all tools            no steering   (baseline)
-//   B steer         all tools            trace-first   (adoption)
-//   C no-explore    hide explore         trace-first   (is explore redundant?)
-//   D trace-centric hide explore+context trace-first   (is the survey pair redundant?)
-//   E control-probe hide explore+context trace-first   (NON-flow Q — should degrade)
-//
-// Usage: node scripts/agent-eval/parse-arms.mjs [/tmp/arms]
-import { readFileSync, readdirSync, existsSync, statSync } from 'fs';
-import { join } from 'path';
-
-const ROOT = process.argv[2] || '/tmp/arms';
-const cgShort = (n) => n.replace('mcp__homegraph__homegraph_', '').replace('mcp__homegraph__', '');
-
-function parse(file) {
-  if (!existsSync(file)) return null;
-  const lines = readFileSync(file, 'utf8').split('\n').filter(Boolean);
-  const calls = []; let result = null, initCg = 0;
-  for (const l of lines) {
-    let ev; try { ev = JSON.parse(l); } catch { continue; }
-    if (ev.type === 'system' && ev.subtype === 'init') initCg = (ev.tools || []).filter(t => /homegraph/.test(t)).length;
-    if (ev.type === 'assistant') for (const b of (ev.message?.content || [])) if (b.type === 'tool_use')
-      calls.push({ id: b.id, name: b.name, out: 0 });
-    if (ev.type === 'user') for (const b of (ev.message?.content || [])) if (b.type === 'tool_result') {
-      const c = b.content;
-      const txt = typeof c === 'string' ? c : Array.isArray(c) ? c.map(x => x?.text || '').join('') : '';
-      const call = calls.find(k => k.id === b.tool_use_id); if (call) call.out = txt.length;
-    }
-    if (ev.type === 'result') result = ev;
-  }
-  const cg = calls.filter(c => c.name.includes('homegraph'));
-  return {
-    initCg,
-    reads: calls.filter(c => c.name === 'Read').length,
-    greps: calls.filter(c => c.name === 'Grep').length + calls.filter(c => c.name === 'Glob').length,
-    cgCalls: cg.length,
-    cgSeq: cg.map(c => cgShort(c.name)),
-    cgOut: cg.reduce((s, c) => s + c.out, 0),
-    traceUsed: cg.some(c => c.name.includes('trace')),
-    turns: result?.num_turns ?? null,
-    dur: result?.duration_ms ? Math.round(result.duration_ms / 1000) : null,
-    cost: result?.total_cost_usd || 0,
-    ok: result?.subtype === 'success',
-  };
-}
-
-// repo -> arm -> [runs]
-const data = {};
-if (!existsSync(ROOT)) { console.error(`no ${ROOT}`); process.exit(1); }
-for (const repo of readdirSync(ROOT)) {
-  const rdir = join(ROOT, repo);
-  if (!statSync(rdir).isDirectory()) continue;
-  for (const f of readdirSync(rdir)) {
-    const m = f.match(/^([A-I])-r(\d+)\.jsonl$/); if (!m) continue;
-    const p = parse(join(rdir, f)); if (!p || !p.ok) continue;
-    (((data[repo] ??= {})[m[1]]) ??= []).push(p);
-  }
-}
-
-const avg = (a, f) => a.length ? a.reduce((s, x) => s + (f(x) || 0), 0) / a.length : 0;
-const k = (n) => (n / 1000).toFixed(1);
-const pad = (s, n) => String(s).padEnd(n);
-const ARMS = ['A', 'H', 'I', 'B', 'F', 'G', 'C', 'D', 'E'];
-const LABEL = { A: 'A all/none(old)', H: 'H body-trace/none', I: 'I bodytrace+dest', B: 'B all/steer(thin)', F: 'F all/steer(body)', G: 'G ported(noprompt)', C: 'C no-explore', D: 'D trace-centric', E: 'E nonflow-probe' };
-
-// ---- per repo × arm ----
-console.log('\n=== PER REPO × ARM (avg over runs) ===');
-console.log(pad('repo', 22), pad('arm', 16), 'tools', 'trace', pad('reads', 6), pad('cgOutK', 7), pad('turns', 6), 'dur');
-for (const repo of Object.keys(data).sort()) {
-  for (const arm of ARMS) {
-    const runs = data[repo][arm]; if (!runs?.length) continue;
-    console.log(
-      pad(repo, 22), pad(LABEL[arm], 16),
-      pad(runs[0].initCg, 5),
-      pad(runs.filter(r => r.traceUsed).length + '/' + runs.length, 5),
-      pad(avg(runs, r => r.reads).toFixed(1), 6),
-      pad(k(avg(runs, r => r.cgOut)), 7),
-      pad(avg(runs, r => r.turns).toFixed(1), 6),
-      avg(runs, r => r.dur).toFixed(0) + 's',
-    );
-  }
-}
-
-// ---- aggregate per arm (flow arms A–D over the flow repos; E shown apart) ----
-console.log('\n=== AGGREGATE PER ARM (mean across repos) ===');
-console.log(pad('arm', 16), pad('adoption', 9), pad('reads', 7), pad('greps', 7), pad('cgOutK', 8), pad('turns', 7), pad('dur', 6), 'cost');
-for (const arm of ARMS) {
-  const all = [];
-  for (const repo of Object.keys(data)) for (const r of (data[repo][arm] || [])) all.push({ ...r, repo });
-  if (!all.length) continue;
-  const repos = new Set(all.map(r => r.repo)).size;
-  const adopt = all.filter(r => r.traceUsed).length;
-  console.log(
-    pad(LABEL[arm], 16),
-    pad(`${adopt}/${all.length}`, 9),
-    pad(avg(all, r => r.reads).toFixed(2), 7),
-    pad(avg(all, r => r.greps).toFixed(2), 7),
-    pad(k(avg(all, r => r.cgOut)), 8),
-    pad(avg(all, r => r.turns).toFixed(1), 7),
-    pad(avg(all, r => r.dur).toFixed(0) + 's', 6),
-    '$' + avg(all, r => r.cost).toFixed(3),
-    `  (${repos} repos)`,
-  );
-}
-
-console.log('\nRead the signal: B vs A = does steering alone fix adoption + cut payload.');
-console.log('C vs B = is explore redundant (reads should NOT jump). D vs C = is context redundant.');
-console.log('E = non-flow under trace-centric; reads SHOULD jump (proves survey tools are load-bearing).');
+#!/usr/bin/env node
+// Analyze the tool-surface ablation (/tmp/arms/<repo>/<arm>-r<n>.jsonl).
+// Compares arms A–E on trace adoption, Read/Grep fallback, homegraph payload,
+// round-trips, and duration — averaged across runs per arm.
+//
+// The decisive signal is READS: if removing a tool raises Reads on a question
+// class, that tool was load-bearing for it (not redundant). If removing it
+// changes nothing, it was redundant.
+//
+//   A control       all tools            no steering   (baseline)
+//   B steer         all tools            trace-first   (adoption)
+//   C no-explore    hide explore         trace-first   (is explore redundant?)
+//   D trace-centric hide explore+context trace-first   (is the survey pair redundant?)
+//   E control-probe hide explore+context trace-first   (NON-flow Q — should degrade)
+//
+// Usage: node scripts/agent-eval/parse-arms.mjs [/tmp/arms]
+import { readFileSync, readdirSync, existsSync, statSync } from 'fs';
+import { join } from 'path';
+
+const ROOT = process.argv[2] || '/tmp/arms';
+const cgShort = (n) => n.replace('mcp__homegraph__homegraph_', '').replace('mcp__homegraph__', '');
+
+function parse(file) {
+  if (!existsSync(file)) return null;
+  const lines = readFileSync(file, 'utf8').split('\n').filter(Boolean);
+  const calls = []; let result = null, initCg = 0;
+  for (const l of lines) {
+    let ev; try { ev = JSON.parse(l); } catch { continue; }
+    if (ev.type === 'system' && ev.subtype === 'init') initCg = (ev.tools || []).filter(t => /homegraph/.test(t)).length;
+    if (ev.type === 'assistant') for (const b of (ev.message?.content || [])) if (b.type === 'tool_use')
+      calls.push({ id: b.id, name: b.name, out: 0 });
+    if (ev.type === 'user') for (const b of (ev.message?.content || [])) if (b.type === 'tool_result') {
+      const c = b.content;
+      const txt = typeof c === 'string' ? c : Array.isArray(c) ? c.map(x => x?.text || '').join('') : '';
+      const call = calls.find(k => k.id === b.tool_use_id); if (call) call.out = txt.length;
+    }
+    if (ev.type === 'result') result = ev;
+  }
+  const cg = calls.filter(c => c.name.includes('homegraph'));
+  return {
+    initCg,
+    reads: calls.filter(c => c.name === 'Read').length,
+    greps: calls.filter(c => c.name === 'Grep').length + calls.filter(c => c.name === 'Glob').length,
+    cgCalls: cg.length,
+    cgSeq: cg.map(c => cgShort(c.name)),
+    cgOut: cg.reduce((s, c) => s + c.out, 0),
+    traceUsed: cg.some(c => c.name.includes('trace')),
+    turns: result?.num_turns ?? null,
+    dur: result?.duration_ms ? Math.round(result.duration_ms / 1000) : null,
+    cost: result?.total_cost_usd || 0,
+    ok: result?.subtype === 'success',
+  };
+}
+
+// repo -> arm -> [runs]
+const data = {};
+if (!existsSync(ROOT)) { console.error(`no ${ROOT}`); process.exit(1); }
+for (const repo of readdirSync(ROOT)) {
+  const rdir = join(ROOT, repo);
+  if (!statSync(rdir).isDirectory()) continue;
+  for (const f of readdirSync(rdir)) {
+    const m = f.match(/^([A-I])-r(\d+)\.jsonl$/); if (!m) continue;
+    const p = parse(join(rdir, f)); if (!p || !p.ok) continue;
+    (((data[repo] ??= {})[m[1]]) ??= []).push(p);
+  }
+}
+
+const avg = (a, f) => a.length ? a.reduce((s, x) => s + (f(x) || 0), 0) / a.length : 0;
+const k = (n) => (n / 1000).toFixed(1);
+const pad = (s, n) => String(s).padEnd(n);
+const ARMS = ['A', 'H', 'I', 'B', 'F', 'G', 'C', 'D', 'E'];
+const LABEL = { A: 'A all/none(old)', H: 'H body-trace/none', I: 'I bodytrace+dest', B: 'B all/steer(thin)', F: 'F all/steer(body)', G: 'G ported(noprompt)', C: 'C no-explore', D: 'D trace-centric', E: 'E nonflow-probe' };
+
+// ---- per repo × arm ----
+console.log('\n=== PER REPO × ARM (avg over runs) ===');
+console.log(pad('repo', 22), pad('arm', 16), 'tools', 'trace', pad('reads', 6), pad('cgOutK', 7), pad('turns', 6), 'dur');
+for (const repo of Object.keys(data).sort()) {
+  for (const arm of ARMS) {
+    const runs = data[repo][arm]; if (!runs?.length) continue;
+    console.log(
+      pad(repo, 22), pad(LABEL[arm], 16),
+      pad(runs[0].initCg, 5),
+      pad(runs.filter(r => r.traceUsed).length + '/' + runs.length, 5),
+      pad(avg(runs, r => r.reads).toFixed(1), 6),
+      pad(k(avg(runs, r => r.cgOut)), 7),
+      pad(avg(runs, r => r.turns).toFixed(1), 6),
+      avg(runs, r => r.dur).toFixed(0) + 's',
+    );
+  }
+}
+
+// ---- aggregate per arm (flow arms A–D over the flow repos; E shown apart) ----
+console.log('\n=== AGGREGATE PER ARM (mean across repos) ===');
+console.log(pad('arm', 16), pad('adoption', 9), pad('reads', 7), pad('greps', 7), pad('cgOutK', 8), pad('turns', 7), pad('dur', 6), 'cost');
+for (const arm of ARMS) {
+  const all = [];
+  for (const repo of Object.keys(data)) for (const r of (data[repo][arm] || [])) all.push({ ...r, repo });
+  if (!all.length) continue;
+  const repos = new Set(all.map(r => r.repo)).size;
+  const adopt = all.filter(r => r.traceUsed).length;
+  console.log(
+    pad(LABEL[arm], 16),
+    pad(`${adopt}/${all.length}`, 9),
+    pad(avg(all, r => r.reads).toFixed(2), 7),
+    pad(avg(all, r => r.greps).toFixed(2), 7),
+    pad(k(avg(all, r => r.cgOut)), 8),
+    pad(avg(all, r => r.turns).toFixed(1), 7),
+    pad(avg(all, r => r.dur).toFixed(0) + 's', 6),
+    '$' + avg(all, r => r.cost).toFixed(3),
+    `  (${repos} repos)`,
+  );
+}
+
+console.log('\nRead the signal: B vs A = does steering alone fix adoption + cut payload.');
+console.log('C vs B = is explore redundant (reads should NOT jump). D vs C = is context redundant.');
+console.log('E = non-flow under trace-centric; reads SHOULD jump (proves survey tools are load-bearing).');

package/scripts/agent-eval/parse-bench-readme.mjs

@@ -1,84 +1,84 @@
-#!/usr/bin/env node
-// Aggregate the README A/B (bench-readme.sh output): per repo, median of N runs
-// per arm → time, tool calls, tokens, cost, and % saved. Plus an average row.
-//
-// Tokens = SUM of per-turn assistant `usage` (input + output + cache read +
-// cache creation) — the cumulative "total tokens processed". NOTE: `result.usage`
-// is last-turn-only in current Claude Code, so it under-counts badly; don't use it.
-// `total_cost_usd` and `duration_ms` are already cumulative.
-//
-// Usage: node parse-bench-readme.mjs [/tmp/ab-readme]
-import { readFileSync, existsSync, readdirSync } from 'fs';
-import { join } from 'path';
-const ROOT = process.argv[2] || '/tmp/ab-readme';
-const REPOS = ['vscode', 'excalidraw', 'django', 'tokio', 'okhttp', 'gin', 'alamofire'];
-
-function parse(file) {
-  if (!existsSync(file)) return null;
-  const L = readFileSync(file, 'utf8').split('\n').filter(Boolean);
-  let tools = 0, reads = 0, grep = 0, cg = 0, tokens = 0, r = null, raced = false;
-  for (const l of L) { let e; try { e = JSON.parse(l); } catch { continue; }
-    if (e.type === 'assistant') {
-      const u = e.message?.usage;
-      if (u) tokens += (u.input_tokens || 0) + (u.output_tokens || 0) + (u.cache_read_input_tokens || 0) + (u.cache_creation_input_tokens || 0);
-      for (const b of (e.message?.content || [])) if (b.type === 'tool_use') {
-        const n = b.name;
-        if (n === 'ToolSearch') continue;
-        tools++;
-        if (n === 'Read') reads++;
-        else if (n === 'Grep' || n === 'Glob') grep++;
-        else if (/homegraph/.test(n)) cg++;
-      }
-    }
-    // MCP cold-start race: the headless agent fired before `homegraph serve --mcp`
-    // finished registering its tools, so early calls returned "No such tool
-    // available" and the agent floundered into grep/Read. That measures HomeGraph's
-    // startup latency, NOT its steady-state value — flag the run so the aggregate
-    // can exclude it (an artifact of headless first-turn timing, not the tool).
-    if (e.type === 'user') for (const b of (Array.isArray(e.message?.content) ? e.message.content : [])) {
-      if (b.type === 'tool_result') {
-        const t = Array.isArray(b.content) ? b.content.map(c => c.text || '').join('') : (b.content || '');
-        if (/No such tool available/.test(t)) raced = true;
-      }
-    }
-    if (e.type === 'result') r = e;
-  }
-  if (!r || r.subtype !== 'success') return null;
-  return { dur: r.duration_ms / 1000, tools, reads, grep, cg, tokens, cost: r.total_cost_usd || 0, raced };
-}
-const median = (arr) => { const v = [...arr].sort((a, b) => a - b); const n = v.length; return n === 0 ? 0 : n % 2 ? v[(n - 1) / 2] : (v[n / 2 - 1] + v[n / 2]) / 2; };
-const fmtTime = (s) => s >= 60 ? `${Math.floor(s / 60)}m ${Math.round(s % 60)}s` : `${Math.round(s)}s`;
-const fmtTok = (t) => t >= 1e6 ? `${(t / 1e6).toFixed(1)}M` : `${Math.round(t / 1000)}k`;
-const pct = (w, wo) => wo > 0 ? Math.round((1 - w / wo) * 100) : 0;
-
-console.log('repo        n(w/wo)  time WITH→WITHOUT      tools W→WO   tokens W→WO (saved)     cost W→WO (saved)');
-const savings = { cost: [], tokens: [], time: [], tools: [] };
-for (const repo of REPOS) {
-  const dir = join(ROOT, repo);
-  const runDirs = existsSync(dir) ? readdirSync(dir).filter(d => /^run\d+$/.test(d)) : [];
-  // Exclude MCP-cold-start-raced WITH runs by default — they measure a startup
-  // race, not steady-state value. `CG_INCLUDE_RACED=1` keeps them (to see the raw
-  // distribution). The WITHOUT arm has no MCP, so it's never raced.
-  const includeRaced = process.env.CG_INCLUDE_RACED === '1';
-  const W = [], WO = []; let racedExcluded = 0;
-  for (const rd of runDirs) {
-    const w = parse(join(dir, rd, 'run-headless-with.jsonl'));
-    if (w) { if (w.raced && !includeRaced) racedExcluded++; else W.push(w); }
-    const wo = parse(join(dir, rd, 'run-headless-without.jsonl')); if (wo) WO.push(wo);
-  }
-  if (!W.length || !WO.length) { console.log(`${repo.padEnd(11)} (incomplete: w=${W.length} wo=${WO.length})`); continue; }
-  const m = (arr, k) => median(arr.map(x => x[k]));
-  const wT = m(W, 'dur'), woT = m(WO, 'dur'), wTok = m(W, 'tokens'), woTok = m(WO, 'tokens');
-  const wC = m(W, 'cost'), woC = m(WO, 'cost'), wTl = m(W, 'tools'), woTl = m(WO, 'tools');
-  savings.time.push(pct(wT, woT)); savings.tokens.push(pct(wTok, woTok)); savings.cost.push(pct(wC, woC)); savings.tools.push(pct(wTl, woTl));
-  console.log(
-    `${repo.padEnd(11)} ${W.length}/${WO.length}      ` +
-    `${(fmtTime(wT) + '→' + fmtTime(woT)).padEnd(22)}` +
-    `${(Math.round(wTl) + '→' + Math.round(woTl)).padEnd(12)}` +
-    `${(fmtTok(wTok) + '→' + fmtTok(woTok) + ' (' + pct(wTok, woTok) + '%)').padEnd(24)}` +
-    `$${wC.toFixed(2)}→$${woC.toFixed(2)} (${pct(wC, woC)}%)` +
-    (racedExcluded ? `  [${racedExcluded} raced run${racedExcluded === 1 ? '' : 's'} excluded]` : '')
-  );
-}
-const avg = (a) => a.length ? Math.round(a.reduce((s, x) => s + x, 0) / a.length) : 0;
-console.log(`\nAVERAGE saved:  cost ${avg(savings.cost)}%  ·  tokens ${avg(savings.tokens)}%  ·  time ${avg(savings.time)}%  ·  tool calls ${avg(savings.tools)}%`);
+#!/usr/bin/env node
+// Aggregate the README A/B (bench-readme.sh output): per repo, median of N runs
+// per arm → time, tool calls, tokens, cost, and % saved. Plus an average row.
+//
+// Tokens = SUM of per-turn assistant `usage` (input + output + cache read +
+// cache creation) — the cumulative "total tokens processed". NOTE: `result.usage`
+// is last-turn-only in current Claude Code, so it under-counts badly; don't use it.
+// `total_cost_usd` and `duration_ms` are already cumulative.
+//
+// Usage: node parse-bench-readme.mjs [/tmp/ab-readme]
+import { readFileSync, existsSync, readdirSync } from 'fs';
+import { join } from 'path';
+const ROOT = process.argv[2] || '/tmp/ab-readme';
+const REPOS = ['vscode', 'excalidraw', 'django', 'tokio', 'okhttp', 'gin', 'alamofire'];
+
+function parse(file) {
+  if (!existsSync(file)) return null;
+  const L = readFileSync(file, 'utf8').split('\n').filter(Boolean);
+  let tools = 0, reads = 0, grep = 0, cg = 0, tokens = 0, r = null, raced = false;
+  for (const l of L) { let e; try { e = JSON.parse(l); } catch { continue; }
+    if (e.type === 'assistant') {
+      const u = e.message?.usage;
+      if (u) tokens += (u.input_tokens || 0) + (u.output_tokens || 0) + (u.cache_read_input_tokens || 0) + (u.cache_creation_input_tokens || 0);
+      for (const b of (e.message?.content || [])) if (b.type === 'tool_use') {
+        const n = b.name;
+        if (n === 'ToolSearch') continue;
+        tools++;
+        if (n === 'Read') reads++;
+        else if (n === 'Grep' || n === 'Glob') grep++;
+        else if (/homegraph/.test(n)) cg++;
+      }
+    }
+    // MCP cold-start race: the headless agent fired before `homegraph serve --mcp`
+    // finished registering its tools, so early calls returned "No such tool
+    // available" and the agent floundered into grep/Read. That measures HomeGraph's
+    // startup latency, NOT its steady-state value — flag the run so the aggregate
+    // can exclude it (an artifact of headless first-turn timing, not the tool).
+    if (e.type === 'user') for (const b of (Array.isArray(e.message?.content) ? e.message.content : [])) {
+      if (b.type === 'tool_result') {
+        const t = Array.isArray(b.content) ? b.content.map(c => c.text || '').join('') : (b.content || '');
+        if (/No such tool available/.test(t)) raced = true;
+      }
+    }
+    if (e.type === 'result') r = e;
+  }
+  if (!r || r.subtype !== 'success') return null;
+  return { dur: r.duration_ms / 1000, tools, reads, grep, cg, tokens, cost: r.total_cost_usd || 0, raced };
+}
+const median = (arr) => { const v = [...arr].sort((a, b) => a - b); const n = v.length; return n === 0 ? 0 : n % 2 ? v[(n - 1) / 2] : (v[n / 2 - 1] + v[n / 2]) / 2; };
+const fmtTime = (s) => s >= 60 ? `${Math.floor(s / 60)}m ${Math.round(s % 60)}s` : `${Math.round(s)}s`;
+const fmtTok = (t) => t >= 1e6 ? `${(t / 1e6).toFixed(1)}M` : `${Math.round(t / 1000)}k`;
+const pct = (w, wo) => wo > 0 ? Math.round((1 - w / wo) * 100) : 0;
+
+console.log('repo        n(w/wo)  time WITH→WITHOUT      tools W→WO   tokens W→WO (saved)     cost W→WO (saved)');
+const savings = { cost: [], tokens: [], time: [], tools: [] };
+for (const repo of REPOS) {
+  const dir = join(ROOT, repo);
+  const runDirs = existsSync(dir) ? readdirSync(dir).filter(d => /^run\d+$/.test(d)) : [];
+  // Exclude MCP-cold-start-raced WITH runs by default — they measure a startup
+  // race, not steady-state value. `CG_INCLUDE_RACED=1` keeps them (to see the raw
+  // distribution). The WITHOUT arm has no MCP, so it's never raced.
+  const includeRaced = process.env.CG_INCLUDE_RACED === '1';
+  const W = [], WO = []; let racedExcluded = 0;
+  for (const rd of runDirs) {
+    const w = parse(join(dir, rd, 'run-headless-with.jsonl'));
+    if (w) { if (w.raced && !includeRaced) racedExcluded++; else W.push(w); }
+    const wo = parse(join(dir, rd, 'run-headless-without.jsonl')); if (wo) WO.push(wo);
+  }
+  if (!W.length || !WO.length) { console.log(`${repo.padEnd(11)} (incomplete: w=${W.length} wo=${WO.length})`); continue; }
+  const m = (arr, k) => median(arr.map(x => x[k]));
+  const wT = m(W, 'dur'), woT = m(WO, 'dur'), wTok = m(W, 'tokens'), woTok = m(WO, 'tokens');
+  const wC = m(W, 'cost'), woC = m(WO, 'cost'), wTl = m(W, 'tools'), woTl = m(WO, 'tools');
+  savings.time.push(pct(wT, woT)); savings.tokens.push(pct(wTok, woTok)); savings.cost.push(pct(wC, woC)); savings.tools.push(pct(wTl, woTl));
+  console.log(
+    `${repo.padEnd(11)} ${W.length}/${WO.length}      ` +
+    `${(fmtTime(wT) + '→' + fmtTime(woT)).padEnd(22)}` +
+    `${(Math.round(wTl) + '→' + Math.round(woTl)).padEnd(12)}` +
+    `${(fmtTok(wTok) + '→' + fmtTok(woTok) + ' (' + pct(wTok, woTok) + '%)').padEnd(24)}` +
+    `$${wC.toFixed(2)}→$${woC.toFixed(2)} (${pct(wC, woC)}%)` +
+    (racedExcluded ? `  [${racedExcluded} raced run${racedExcluded === 1 ? '' : 's'} excluded]` : '')
+  );
+}
+const avg = (a) => a.length ? Math.round(a.reduce((s, x) => s + x, 0) / a.length) : 0;
+console.log(`\nAVERAGE saved:  cost ${avg(savings.cost)}%  ·  tokens ${avg(savings.tokens)}%  ·  time ${avg(savings.time)}%  ·  tool calls ${avg(savings.tools)}%`);

package/scripts/agent-eval/parse-run.mjs

@@ -1,45 +1,45 @@
-#!/usr/bin/env node
-// Parse a Claude Code stream-json run log: tool-call sequence + token usage.
-import { readFileSync } from 'fs';
-const file = process.argv[2];
-const lines = readFileSync(file, 'utf8').split('\n').filter(Boolean);
-
-const toolCalls = [];
-let result = null;
-let initTools = null;
-
-for (const line of lines) {
-  let ev;
-  try { ev = JSON.parse(line); } catch { continue; }
-  if (ev.type === 'system' && ev.subtype === 'init') {
-    initTools = (ev.tools || []).filter(t => /homegraph/.test(t));
-  }
-  if (ev.type === 'assistant' && ev.message?.content) {
-    for (const block of ev.message.content) {
-      if (block.type === 'tool_use') {
-        let detail = '';
-        if (block.name === 'Task') detail = ` [subagent_type=${block.input?.subagent_type ?? '?'}] ${(block.input?.description ?? '').slice(0,40)}`;
-        else if (/homegraph/.test(block.name)) detail = ` ${JSON.stringify(block.input?.query ?? block.input?.task ?? block.input?.symbol ?? '').slice(0,60)}`;
-        else if (block.name === 'Bash') detail = ` ${(block.input?.command ?? '').slice(0,50)}`;
-        else if (block.name === 'Read') detail = ` ${(block.input?.file_path ?? '').split('/').slice(-1)[0]}`;
-        toolCalls.push(`${block.name}${detail}`);
-      }
-    }
-  }
-  if (ev.type === 'result') result = ev;
-}
-
-console.log(`\n=== ${file.split('/').pop()} ===`);
-console.log(`homegraph tools exposed: ${initTools ? initTools.length : '?'}`);
-console.log(`\nTool calls (${toolCalls.length}):`);
-const counts = {};
-for (const tc of toolCalls) { const n = tc.split(' ')[0]; counts[n] = (counts[n]||0)+1; }
-console.log('  by type:', JSON.stringify(counts));
-toolCalls.forEach((tc, i) => console.log(`  ${i+1}. ${tc}`));
-
-if (result) {
-  const u = result.usage || {};
-  const totalIn = (u.input_tokens||0) + (u.cache_read_input_tokens||0) + (u.cache_creation_input_tokens||0);
-  console.log(`\nResult: ${result.subtype} | duration ${(result.duration_ms/1000).toFixed(0)}s | turns ${result.num_turns}`);
-  console.log(`  tokens: in=${totalIn} out=${u.output_tokens||0} | cost $${(result.total_cost_usd||0).toFixed(3)}`);
-}
+#!/usr/bin/env node
+// Parse a Claude Code stream-json run log: tool-call sequence + token usage.
+import { readFileSync } from 'fs';
+const file = process.argv[2];
+const lines = readFileSync(file, 'utf8').split('\n').filter(Boolean);
+
+const toolCalls = [];
+let result = null;
+let initTools = null;
+
+for (const line of lines) {
+  let ev;
+  try { ev = JSON.parse(line); } catch { continue; }
+  if (ev.type === 'system' && ev.subtype === 'init') {
+    initTools = (ev.tools || []).filter(t => /homegraph/.test(t));
+  }
+  if (ev.type === 'assistant' && ev.message?.content) {
+    for (const block of ev.message.content) {
+      if (block.type === 'tool_use') {
+        let detail = '';
+        if (block.name === 'Task') detail = ` [subagent_type=${block.input?.subagent_type ?? '?'}] ${(block.input?.description ?? '').slice(0,40)}`;
+        else if (/homegraph/.test(block.name)) detail = ` ${JSON.stringify(block.input?.query ?? block.input?.task ?? block.input?.symbol ?? '').slice(0,60)}`;
+        else if (block.name === 'Bash') detail = ` ${(block.input?.command ?? '').slice(0,50)}`;
+        else if (block.name === 'Read') detail = ` ${(block.input?.file_path ?? '').split('/').slice(-1)[0]}`;
+        toolCalls.push(`${block.name}${detail}`);
+      }
+    }
+  }
+  if (ev.type === 'result') result = ev;
+}
+
+console.log(`\n=== ${file.split('/').pop()} ===`);
+console.log(`homegraph tools exposed: ${initTools ? initTools.length : '?'}`);
+console.log(`\nTool calls (${toolCalls.length}):`);
+const counts = {};
+for (const tc of toolCalls) { const n = tc.split(' ')[0]; counts[n] = (counts[n]||0)+1; }
+console.log('  by type:', JSON.stringify(counts));
+toolCalls.forEach((tc, i) => console.log(`  ${i+1}. ${tc}`));
+
+if (result) {
+  const u = result.usage || {};
+  const totalIn = (u.input_tokens||0) + (u.cache_read_input_tokens||0) + (u.cache_creation_input_tokens||0);
+  console.log(`\nResult: ${result.subtype} | duration ${(result.duration_ms/1000).toFixed(0)}s | turns ${result.num_turns}`);
+  console.log(`  tokens: in=${totalIn} out=${u.output_tokens||0} | cost $${(result.total_cost_usd||0).toFixed(3)}`);
+}