mixdog 1.2.2 → 1.2.4

package/README.md

@@ -134,6 +134,24 @@ is published in [`benchmarks/terminal-bench-2.1/`](benchmarks/terminal-bench-2.1
 and [Context efficiency](docs/context-efficiency.md) explains where the
 savings come from.
 
+<details>
+<summary><b>Benchmark history</b></summary>
+
+Newest first. Each run links to its details, records, and caveats.
+
+| Run | Model · trials | vs | Score | Cost | Final context | Time / trial |
+|---|---|---|---|---|---|---|
+| [2026-10-15](benchmarks/terminal-bench-2.1/#bench-log) | GPT-6.1 Sol xhigh · `k=1` | Codex CLI | 77 vs 78 /89 | **−55%** | **−68%** (10.8K vs 33.6K) | 344s vs 380s |
+| [2026-08-25](benchmarks/terminal-bench-2.1/#gpt-56-sol-xhigh-vs-codex-cli) | GPT-5.6 Sol xhigh · `k=5` | Codex CLI | 385 vs 383 /445 | **−39%** | **−46%** (18.5K vs 34.3K) | 415s vs 437s |
+| [2026-08-25](benchmarks/terminal-bench-2.1/#claude-opus-5-vs-claude-code) | Claude Opus 5 · `k=1` | Claude Code | 79 vs 77 /89 | **−19%** | **−28%** (27.6K vs 38.2K) | 610s vs 708s |
+
+The 2026-10-15 run is the first full run with the
+[effort judge](docs/auto-reasoning-effort.md) active on OpenAI routes, with web
+search on for both sides. It is a single pass, so the `k=5` chart above stays
+the reference.
+
+</details>
+
 ## Documentation
 
 - [Context efficiency](docs/context-efficiency.md)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mixdog",
-  "version": "1.2.2",
+  "version": "1.2.4",
   "private": false,
   "type": "module",
   "description": "Open-source coding agent that matches Codex on Terminal-Bench 2.1 with 63% fewer tokens.",
@@ -161,6 +161,7 @@
     "@mozilla/readability": "^0.6.0",
     "@napi-rs/canvas": "^1.0.8",
     "@pdf-lib/fontkit": "^1.1.1",
+    "ajv": "^8.20.0",
     "chalk": "^5.6.2",
     "cli-highlight": "^2.1.11",
     "diff": "^9.0.0",

package/scripts/gate-local.mjs

@@ -41,11 +41,13 @@ if (build.status !== 0) {
 const logDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mixdog-gate-local-'));
 const resourceDir = fs.mkdtempSync(path.join(os.tmpdir(), 'mixdog-gate-resources-'));
 const budget = gateTestBudget();
+// weight: relative worker share, roughly each leg's measured duration x its 3 workers
+// (runtime ~440s, runtime-slow ~545s, desktop fast ~390s, desktop slow ~280s).
 const legs = [
-  { name: 'runtime', cmds: ['npm test', 'npm run smoke:patch'] },
-  { name: 'desktop-tests-fast', cmds: ['npm test --prefix apps/desktop'] },
-  { name: 'runtime-slow', cmds: ['npm run test:slow'] },
-  { name: 'desktop-tests-slow', cmds: ['npm run test:slow --prefix apps/desktop'] },
+  { name: 'runtime', weight: 3, cmds: ['npm test', 'npm run smoke:patch'] },
+  { name: 'desktop-tests-fast', weight: 3, cmds: ['npm test --prefix apps/desktop'] },
+  { name: 'runtime-slow', weight: 4, cmds: ['npm run test:slow'] },
+  { name: 'desktop-tests-slow', weight: 2, cmds: ['npm run test:slow --prefix apps/desktop'] },
   { name: 'desktop', cmds: ['npm run build --prefix apps/desktop', 'npm run test:daemon:e2e --prefix apps/desktop'] },
 ];
 

package/scripts/lib/gate-scheduler.mjs

@@ -8,17 +8,35 @@ export function gateTestBudget(env = process.env, parallelism = availableParalle
   return argument ? Number(argument.split('=')[1]) : Math.min(6, parallelism);
 }
 
+// Splits the budget across legs in proportion to leg.weight (default 1), at least
+// one worker each. Spare workers go one at a time to the leg with the highest
+// weight per worker, so longer legs get more. Needs budget >= legs.length.
+export function allocateWorkers(legs, budget) {
+  const shares = legs.map(() => 1);
+  for (let spare = budget - legs.length; spare > 0; spare--) {
+    let best = 0;
+    for (let i = 1; i < legs.length; i++) {
+      if ((legs[i].weight ?? 1) / shares[i] > (legs[best].weight ?? 1) / shares[best]) best = i;
+    }
+    shares[best]++;
+  }
+  return shares;
+}
+
+// The test runners read their worker count when spawned, so it cannot grow
+// mid-run: shares are static. With fewer workers than legs, each worker runs
+// legs one after another.
 export async function runGateLegs(legs, budget, runLeg) {
   if (!Number.isInteger(budget) || budget < 1) throw new Error('gate test budget must be a positive integer');
-  const slots = Math.min(legs.length, budget);
-  let next = 0;
-  await Promise.all(
-    Array.from({ length: slots }, async (_, index) => {
-      const concurrency = Math.floor(budget / slots) + (index < budget % slots ? 1 : 0);
-      while (next < legs.length) {
-        const leg = legs[next++];
-        await runLeg(leg, concurrency);
-      }
-    })
-  );
+  if (budget < legs.length) {
+    let next = 0;
+    await Promise.all(
+      Array.from({ length: budget }, async () => {
+        while (next < legs.length) await runLeg(legs[next++], 1);
+      })
+    );
+    return;
+  }
+  const shares = allocateWorkers(legs, budget);
+  await Promise.all(legs.map((leg, index) => runLeg(leg, shares[index])));
 }

package/scripts/probe-provider-catalog.mjs

@@ -0,0 +1,31 @@
+// Prints each named provider's raw listModels() outcome with the real ~/.mixdog data.
+import os from 'node:os';
+import { createMixdogSessionRuntime } from '../src/session-runtime/runtime-core.mjs';
+import { getProvider } from '../src/runtime/agent/orchestrator/providers/registry.mjs';
+import { listCompatModels } from '../src/runtime/agent/orchestrator/providers/openai-compat-models.mjs';
+
+const names = process.argv.slice(2);
+const runtime = await createMixdogSessionRuntime({ provider: 'deepseek', model: 'deepseek-flash', cwd: os.tmpdir() });
+await runtime.listProviderModels?.({ force: true }).catch(() => {});
+for (const name of names) {
+  const provider = getProvider(name);
+  if (!provider) {
+    console.log(name, 'NOT REGISTERED');
+    continue;
+  }
+  try {
+    const models =
+      name === 'xai' ? await listCompatModels(provider, { throwOnError: true }) : await provider.listModels();
+    console.log(
+      name,
+      'models',
+      models.length,
+      models.fallback ? '(fallback)' : '',
+      models.slice(0, 5).map((m) => m.id)
+    );
+  } catch (error) {
+    console.log(name, 'ERROR', error?.status || '', String(error?.message || error).slice(0, 300));
+  }
+}
+await runtime.close('probe-provider-catalog');
+process.exit(0);