badgr-cli 1.0.22 → 1.0.25

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "badgr-cli",
-  "version": "1.0.22",
+  "version": "1.0.25",
   "description": "Badgr, run or serve GPU workloads from one command",
   "type": "module",
   "bin": {

package/src/badgr.js

@@ -17,27 +17,27 @@ const HELP = `
 ${chalk.bold('badgr')} — run or serve GPU workloads from one command
 
 ${chalk.bold('COMMANDS')}
-  ${chalk.cyan('badgr login')}                  Authenticate with your API key
-  ${chalk.cyan('badgr run <command>')}          Run a one-off GPU job
-  ${chalk.cyan('badgr serve <model>')}          Serve a model with an OpenAI-compatible endpoint
-  ${chalk.cyan('badgr status')}                 Show what's running and what's billing
-  ${chalk.cyan('badgr logs <id>')}              Stream logs for a running job or endpoint
-  ${chalk.cyan('badgr down <id>')}              Stop a deployment and end billing
-  ${chalk.cyan('badgr receipts')}               Show cost history
-  ${chalk.cyan('badgr test')}                   Run an end-to-end test (provision → run → teardown)
-  ${chalk.cyan('badgr capacity')}               Check what GPU capacity is available right now
+  ${chalk.cyan('badgr login')}               Authenticate with your API key
+  ${chalk.cyan('badgr run <command>')}       Run a one-off GPU job
+  ${chalk.cyan('badgr serve <model>')}       Serve a model with an OpenAI-compatible endpoint
+  ${chalk.cyan('badgr status')}              Show what's running and what's billing
+  ${chalk.cyan('badgr logs <id>')}           Stream logs for a running job or endpoint
+  ${chalk.cyan('badgr down <id>')}           Stop a deployment and end billing
+  ${chalk.cyan('badgr receipts')}            Show cost history
+  ${chalk.cyan('badgr test')}               Run an end-to-end test (provision → run → teardown)
+  ${chalk.cyan('badgr capacity')}            Check what GPU capacity is available right now
 
 ${chalk.bold('EXAMPLES')}
   ${chalk.dim('# Verify the stack works end-to-end:')}
   badgr test
 
-  ${chalk.dim('# Simplest — Badgr picks the GPU (RunPod, reliable):')}
+  ${chalk.dim('# Tier 1 — managed provider routing (default):')}
   badgr run python train.py
   badgr serve meta-llama/Llama-3.1-8B-Instruct
 
-  ${chalk.dim('# Opt into cheaper budget providers (Vast.ai etc.):')}
-  badgr run python train.py --cheap
-  badgr serve meta-llama/Llama-3.1-8B-Instruct --cheap
+  ${chalk.dim('# Tier 2 — marketplace routing, lower-cost options:')}
+  badgr run python train.py --tier 2
+  badgr serve meta-llama/Llama-3.1-8B-Instruct --tier 2
 
   ${chalk.dim('# Pin a specific GPU:')}
   badgr run python train.py --gpu A100
@@ -53,12 +53,12 @@ ${chalk.bold('EXAMPLES')}
   badgr receipts dep-abc123
 
 ${chalk.bold('badgr run OPTIONS')}
-  --gpu <type>           GPU type (default: auto — Badgr picks best available on RunPod)
-  --cheap                Search budget providers too (Vast.ai etc.) for lower prices
+  --gpu <type>           GPU type (default: auto — Badgr picks best available)
+  --tier 1               Managed provider routing (default)
+  --tier 2               Marketplace provider routing, lower-cost options
   --image <image>        Docker image (default: python:3.11-slim)
   --count <n>            Number of GPUs (default: 1)
   --region US|EU|AU      Region preference
-  --tier 1|2             Provider tier: 1 = reliable (default), 2 = budget
   --max-price <$/hr>     Hard spend cap per GPU-hour
   --max-runtime <min>    Auto-stop after N minutes (recommended)
   --max-cost <$>         Auto-stop when spend reaches this amount
@@ -66,10 +66,10 @@ ${chalk.bold('badgr run OPTIONS')}
 
 ${chalk.bold('badgr serve OPTIONS')}
   --gpu <type>           GPU type (default: auto — inferred from model size)
-  --cheap                Search budget providers too (Vast.ai etc.) for lower prices
+  --tier 1               Managed provider routing (default)
+  --tier 2               Marketplace provider routing, lower-cost options
   --count <n>            Number of GPUs (default: 1)
   --region US|EU|AU      Region preference
-  --tier 1|2             Provider tier: 1 = reliable (default), 2 = budget
   --max-price <$/hr>     Hard spend cap per GPU-hour
   --no-wait              Skip endpoint health check
 
@@ -103,7 +103,7 @@ async function main() {
     case 'receipts': return receiptsCommand(config, rest, chalk);
     case 'models':   return modelsCommand(config, chalk);
     case 'capacity': return capacityCommand(config, rest, chalk);
-    case 'test':     return testCommand(config, chalk);
+    case 'test':     return testCommand(config, rest, chalk);
     // legacy aliases kept for compatibility
     case 'up':       return upCommand(config, rest, chalk);
     case 'config': {

package/src/commands/run.js

@@ -2,7 +2,6 @@ import readline from 'readline';
 import { requireApiKey } from '../config.js';
 import { callApi, terminateDeployment } from '../api.js';
 import { addReceipt, updateReceipt, generateReceiptId } from '../store.js';
-import { rankAlternatives, promptFallback } from '../fallback.js';
 
 /**
  * badgr run python train.py                    # gpu=auto, attached
@@ -19,7 +18,6 @@ export function parseRunArgs(args) {
     if (args[i] === '--count')          { flags.count = parseInt(args[++i], 10); i++; continue; }
     if (args[i] === '--region')         { flags.region = args[++i]; i++; continue; }
     if (args[i] === '--tier')           { flags.tier = args[++i]; i++; continue; }
-    if (args[i] === '--cheap')          { flags.cheap = true; i++; continue; }
     if (args[i] === '--max-price')      { flags.maxPrice = parseFloat(args[++i]); i++; continue; }
     if (args[i] === '--name')           { flags.name = args[++i]; i++; continue; }
     if (args[i] === '--detach')          { flags.detach = true; i++; continue; }
@@ -350,9 +348,8 @@ export async function runCommand(config, args, chalk) {
   const maxRuntimeMs = flags.maxRuntime ? flags.maxRuntime * 60 * 1000 : null;
   const maxCost      = flags.maxCost ?? null;
 
-  // Resolve effective tier: --cheap and --tier 2 opt into budget providers;
-  // everything else defaults to tier 1 (RunPod only) for reliability.
-  const effectiveTier = (flags.cheap || flags.tier === '2' || flags.tier === 'tier2' || flags.tier === 'tier-2')
+  // Tier 1 = managed routing (default). Tier 2 = marketplace routing, opt-in via --tier 2.
+  const effectiveTier = (flags.tier === '2' || flags.tier === 'tier2' || flags.tier === 'tier-2')
     ? '2'
     : (flags.tier || '1');
 
@@ -368,7 +365,7 @@ export async function runCommand(config, args, chalk) {
     console.log(chalk.bold(`\n⚡ Running ${workloadLabel}\n`));
     if (command) console.log(`  ${chalk.bold('Command:')}  ${command.join(' ')}`);
     if (image)   console.log(`  ${chalk.bold('Image:')}    ${image}`);
-    if (effectiveTier === '2') console.log(`  ${chalk.dim('(budget mode — searching all providers)')}`);
+    if (effectiveTier === '2') console.log(`  ${chalk.dim('(tier 2 — marketplace routing)')}`)
     console.log();
     process.stdout.write(chalk.dim('  Finding best GPU...'));
 
@@ -395,7 +392,7 @@ export async function runCommand(config, args, chalk) {
 
     if (effectiveTier === '2' && process.stdin.isTTY) {
       // Budget mode: confirm because the user is being routed to a less reliable provider.
-      const answer = await askConfirm(`  Press ${chalk.bold('Enter')} to run on budget provider, or ${chalk.bold('q')} to cancel: `);
+      const answer = await askConfirm(`  Press ${chalk.bold('Enter')} to run on tier 2 (marketplace), or ${chalk.bold('q')} to cancel: `);
       if (answer.toLowerCase() === 'q') {
         console.log(chalk.dim('\n  Cancelled.\n'));
         process.exit(0);
@@ -454,56 +451,49 @@ export async function runCommand(config, args, chalk) {
   } catch (err) {
     const d = err.errorData;
     if (d?.code === 'NO_CAPACITY_MATCH') {
-      if (fallbackMode === 'none') {
-        console.error(chalk.red(`\n  ✗ ${gpu} isn't available right now.\n`));
-        console.error(chalk.dim('  Pass --fallback closest to auto-select an alternative.'));
-        process.exit(1);
-      }
-
-      const pool = [
-        ...(Array.isArray(d.same_gpu_other_regions) ? d.same_gpu_other_regions : []),
-        ...(Array.isArray(d.alternatives)            ? d.alternatives            : []),
-      ];
-
-      if (pool.length === 0) {
-        console.error(chalk.red(`\n  ✗ ${gpu} isn't available right now and no alternatives were found.\n`));
+      if (effectiveTier === '2') {
+        console.error(chalk.red('\n  ✗ No GPU capacity available right now on any provider.\n'));
         console.error(chalk.dim('  Try `badgr capacity` to see what\'s available, or try again shortly.'));
         process.exit(1);
       }
 
-      const ranked = rankAlternatives(gpu, pool, fallbackMode);
-      const chosen = await promptFallback(gpu, ranked, chalk);
-
-      if (!chosen) {
-        console.log(chalk.dim('\n  Cancelled.\n'));
-        process.exit(0);
+      // Tier 1 out of capacity — offer tier 2 marketplace routing.
+      if (process.stdin.isTTY) {
+        const answer = await askConfirm(
+          `\n  No tier 1 capacity available. Try tier 2 marketplace routing? [${chalk.bold('Enter')}/${chalk.bold('q')}]: `
+        );
+        if (answer.toLowerCase() === 'q') {
+          console.log(chalk.dim('\n  Cancelled.\n'));
+          process.exit(0);
+        }
+      } else {
+        console.log(chalk.dim('\n  No tier 1 capacity — trying tier 2 marketplace routing...\n'));
       }
 
-      console.log(chalk.dim(`\n  Trying ${chosen.gpu} in ${chosen.region}...\n`));
-
+      console.log(chalk.dim('  Searching tier 2 capacity...'));
       try {
         dep = await callApi('/run', {
           method: 'POST',
           apiKey: config.apiKey,
           baseUrl: config.baseUrl,
-          body: buildBody(chosen.gpu, chosen.region),
+          body: { ...buildBody(), tier: '2' },
         });
       } catch (err2) {
         const d2 = err2.errorData;
-        if (d2?.code === 'PROVISIONING_FAILED' || d2?.code === 'PROVIDER_ADAPTER_ERROR') {
-          if (d2?.low_cost_provider_failed) {
-            console.error(chalk.red(`\n  ✗ Tier 2 unavailable. Tier 1 also unavailable. Try again shortly.\n`));
-          } else {
-            console.error(chalk.red(`\n  ✗ Badgr found ${chosen.gpu} capacity but could not start the machine. Please try again.\n`));
-          }
+        if (d2?.code === 'NO_CAPACITY_MATCH') {
+          console.error(chalk.red('\n  ✗ No GPU capacity available right now on any provider.\n'));
+          console.error(chalk.dim('  Try `badgr capacity` to see what\'s available, or try again shortly.'));
+        } else if (d2?.code === 'PROVISIONING_FAILED' || d2?.code === 'PROVIDER_ADAPTER_ERROR') {
+          console.error(chalk.red(`\n  ✗ Budget provider found capacity but could not start the machine. Please try again.\n`));
+          // (error detail kept below)
           if (process.env.BADGR_DEBUG === '1' || process.env.BADGR_DEBUG === 'true') {
             if (d2?.debug_error) console.error(chalk.dim(`  Provider detail: ${d2.debug_error}`));
+          } else {
+            console.error(chalk.dim(`  Run BADGR_DEBUG=1 badgr run … for a full trace.\n`));
           }
         } else {
-          console.error(chalk.red(`\n  ✗ Job failed to start on ${chosen.gpu}: ${err2.message}\n`));
-        }
-        if (!(process.env.BADGR_DEBUG === '1' || process.env.BADGR_DEBUG === 'true')) {
-          console.error(chalk.dim(`  Run BADGR_DEBUG=1 badgr run … for a full trace.\n`));
+          console.error(chalk.red(`\n  ✗ Could not start job on tier 2: ${err2.message}\n`));
+          console.error(chalk.dim(`  Run BADGR_DEBUG=1 badgr run … for a full trace.`));
         }
         process.exit(1);
       }

package/src/commands/serve.js

@@ -1,7 +1,12 @@
+import readline from 'readline';
 import { requireApiKey } from '../config.js';
 import { callApi } from '../api.js';
 import { addDeployment, addReceipt, updateReceipt, generateReceiptId } from '../store.js';
-import { rankAlternatives, promptFallback } from '../fallback.js';
+
+function askConfirm(prompt) {
+  const rl = readline.createInterface({ input: process.stdin, output: process.stdout });
+  return new Promise(resolve => rl.question(prompt, ans => { rl.close(); resolve(ans.trim()); }));
+}
 
 /**
  * badgr serve meta-llama/Llama-3.1-8B-Instruct
@@ -18,7 +23,6 @@ export function parseServeArgs(args) {
     if (args[i] === '--count')      { flags.count = parseInt(args[++i], 10); i++; continue; }
     if (args[i] === '--region')     { flags.region = args[++i]; i++; continue; }
     if (args[i] === '--tier')       { flags.tier = args[++i]; i++; continue; }
-    if (args[i] === '--cheap')      { flags.cheap = true; i++; continue; }
     if (args[i] === '--max-price')  { flags.maxPrice = parseFloat(args[++i]); i++; continue; }
     if (args[i] === '--name')       { flags.name = args[++i]; i++; continue; }
     if (args[i] === '--no-wait')    { flags.noWait = true; i++; continue; }
@@ -70,15 +74,15 @@ export async function serveCommand(config, args, chalk) {
   const gpu = flags.gpu ? flags.gpu.toUpperCase().replace('-', '_') : 'AUTO';
   const gpuLabel = gpu === 'AUTO' ? 'auto' : gpu;
 
-  // Default to tier 1 (RunPod) for reliability; --cheap or --tier 2 opts into budget providers.
-  const effectiveTier = (flags.cheap || flags.tier === '2' || flags.tier === 'tier2' || flags.tier === 'tier-2')
+  // Tier 1 = managed routing (default). Tier 2 = marketplace routing, opt-in via --tier 2.
+  const effectiveTier = (flags.tier === '2' || flags.tier === 'tier2' || flags.tier === 'tier-2')
     ? '2'
     : (flags.tier || '1');
 
   console.log(chalk.bold('\nServing model\n'));
   console.log(`  ${chalk.bold('Model:')}  ${model}`);
   console.log(`  ${chalk.bold('GPU:')}    ${gpuLabel}`);
-  if (effectiveTier === '2') console.log(`  ${chalk.dim('(budget mode — searching all providers)')}`);
+  if (effectiveTier === '2') console.log(`  ${chalk.dim('(tier 2 — marketplace routing)')}`);
   console.log();
   process.stdout.write(chalk.dim('  Finding GPU capacity...\n'));
 
@@ -107,46 +111,45 @@ export async function serveCommand(config, args, chalk) {
   } catch (err) {
     const d = err.errorData;
     if (d?.code === 'NO_CAPACITY_MATCH') {
-      const pool = [
-        ...(Array.isArray(d.same_gpu_other_regions) ? d.same_gpu_other_regions : []),
-        ...(Array.isArray(d.alternatives)            ? d.alternatives            : []),
-      ];
-
-      if (pool.length === 0) {
-        console.error(chalk.red(`\n  ✗ No GPU capacity available right now.\n`));
+      if (effectiveTier === '2') {
+        console.error(chalk.red('\n  ✗ No GPU capacity available right now on any provider.\n'));
         console.error(chalk.dim('  Run `badgr capacity` to see alternatives, or try again shortly.'));
         process.exit(1);
       }
 
-      const ranked = rankAlternatives(gpuLabel, pool, 'closest');
-      const chosen = await promptFallback(gpuLabel, ranked, chalk);
-
-      if (!chosen) {
-        console.log(chalk.dim('\n  Cancelled.\n'));
-        process.exit(0);
+      // Tier 1 out of capacity — offer tier 2 marketplace routing.
+      if (process.stdin.isTTY) {
+        const answer = await askConfirm(
+          `\n  No tier 1 capacity available. Try tier 2 marketplace routing? [${chalk.bold('Enter')}/${chalk.bold('q')}]: `
+        );
+        if (answer.toLowerCase() === 'q') {
+          console.log(chalk.dim('\n  Cancelled.\n'));
+          process.exit(0);
+        }
+      } else {
+        console.log(chalk.dim('\n  No tier 1 capacity — trying tier 2 marketplace routing...\n'));
       }
 
-      console.log(chalk.dim(`\n  Trying ${chosen.gpu} in ${chosen.region}...\n`));
-
+      console.log(chalk.dim('  Searching tier 2 capacity...'));
       try {
         dep = await callApi('/serve', {
           method: 'POST',
           apiKey: config.apiKey,
           baseUrl: config.baseUrl,
-          body: buildBody(chosen.gpu, chosen.region),
+          body: { ...buildBody(), tier: '2' },
         });
       } catch (err2) {
         const d2 = err2.errorData;
-        if (d2?.code === 'PROVISIONING_FAILED' || d2?.code === 'PROVIDER_ADAPTER_ERROR') {
-          if (d2?.low_cost_provider_failed) {
-            console.error(chalk.red(`\n  ✗ Tier 2 unavailable. Tier 1 also unavailable. Try again shortly.\n`));
-          } else {
-            console.error(chalk.red(`\n  ✗ Badgr found ${chosen.gpu} capacity but could not start the endpoint. Please try again.\n`));
-          }
+        if (d2?.code === 'NO_CAPACITY_MATCH') {
+          console.error(chalk.red('\n  ✗ No GPU capacity available right now on any provider.\n'));
+          console.error(chalk.dim('  Run `badgr capacity` to see alternatives, or try again shortly.'));
+        } else if (d2?.code === 'PROVISIONING_FAILED' || d2?.code === 'PROVIDER_ADAPTER_ERROR') {
+          console.error(chalk.red(`\n  ✗ Budget provider found capacity but could not start the endpoint. Please try again.\n`));
+          console.error(chalk.dim(`  Run BADGR_DEBUG=1 badgr serve … for a full trace.\n`));
         } else {
-          console.error(chalk.red(`\n  ✗ Failed to start endpoint on ${chosen.gpu}: ${err2.message}\n`));
+          console.error(chalk.red(`\n  ✗ Could not start endpoint on tier 2: ${err2.message}\n`));
+          console.error(chalk.dim(`  Run BADGR_DEBUG=1 badgr serve … for a full trace.\n`));
         }
-        console.error(chalk.dim(`  Run BADGR_DEBUG=1 badgr serve … for a full trace.\n`));
         process.exit(1);
       }
     } else if (d?.code === 'PROVISIONING_FAILED' || d?.code === 'PROVIDER_ADAPTER_ERROR') {

package/src/commands/status.js

@@ -47,7 +47,7 @@ export async function statusCommand(config, args, chalk) {
 
   console.log(chalk.bold('\nRunning now:\n'));
   for (const d of running) {
-    const id   = d.name || d.deployment_id;
+    const id   = d.deployment_id || d.name;
     const type = d.workload_type === 'endpoint' ? 'endpoint' : 'job';
     const gpu  = d.gpu_type || '—';
     const rate = d.cost_per_hour > 0 ? chalk.yellow(`$${d.cost_per_hour.toFixed(2)}/hr`) : '';
@@ -74,7 +74,7 @@ export async function statusCommand(config, args, chalk) {
 
   console.log(chalk.bold('Stop billing:\n'));
   for (const d of running) {
-    const id = d.name || d.deployment_id;
+    const id = d.deployment_id || d.name;
     console.log(`  ${chalk.cyan(`badgr down ${id}`)}`);
   }
   console.log();

package/src/commands/test-run.js

@@ -9,6 +9,20 @@ const TEST_COMMAND    = ['python', '-c', "print('hello from badgr')"];
 const TEST_IMAGE      = 'python:3.11-slim';
 const EXPECTED_OUTPUT = 'hello from badgr';
 
+// --provider flag resolves to a backend tier value.
+// 'tier1' → managed routing (default), 'tier2' → marketplace routing, 'secondary' → secondary dispatch.
+const PROVIDER_TO_TIER = { tier1: '1', tier2: '2', secondary: 'modal' };
+
+export function parseTestArgs(args) {
+  const flags = {};
+  let i = 0;
+  while (i < args.length) {
+    if (args[i] === '--provider' && args[i + 1]) { flags.provider = args[++i]; i++; continue; }
+    i++;
+  }
+  return flags;
+}
+
 function step(chalk, ok, msg, detail = '') {
   const icon = ok ? chalk.green('✓') : chalk.red('✗');
   const suffix = detail ? chalk.dim(` — ${detail}`) : '';
@@ -46,12 +60,40 @@ async function pollLogs(config, depId, expected, timeoutMs) {
   return false;
 }
 
-export async function testCommand(config, chalk) {
+export async function testCommand(config, args, chalk) {
   requireApiKey(config);
 
+  const flags = parseTestArgs(Array.isArray(args) ? args : []);
+  const providerKey = flags.provider ? flags.provider.toLowerCase() : 'tier1';
+
+  if (providerKey === 'secondary') {
+    // Secondary dispatch provider uses a webhook model, not direct GPU rental.
+    // Verify the backend reports it as configured.
+    console.log(chalk.bold('\n⚡ Testing secondary dispatch provider\n'));
+    let routes;
+    try {
+      routes = await callApi('/compute/routes', { apiKey: config.apiKey, baseUrl: config.baseUrl });
+    } catch {
+      routes = null;
+    }
+    const secondaryRoute = Array.isArray(routes) ? routes.find(r => r.name === 'modal') : null;
+    if (secondaryRoute?.available) {
+      step(chalk, true, 'Secondary provider configured');
+      console.log(chalk.green('\n  ✓ Secondary dispatch provider is ready\n'));
+    } else {
+      step(chalk, false, 'Secondary provider configured', 'contact support to enable secondary dispatch');
+      console.log(chalk.red('\n  ✗ Secondary dispatch provider is not configured\n'));
+      process.exit(1);
+    }
+    return;
+  }
+
+  const tier = PROVIDER_TO_TIER[providerKey] ?? '1';
+  const tierLabel = tier === '1' ? 'tier 1 (managed routing)' : 'tier 2 (marketplace routing)';
+
   console.log(chalk.bold('\n⚡ Running end-to-end test\n'));
   console.log(chalk.dim(`  Command:  ${TEST_COMMAND.join(' ')}`));
-  console.log(chalk.dim(`  Provider: RunPod (Tier 1 — reliable)`));
+  console.log(chalk.dim(`  Routing:  ${tier === '1' ? 'tier 1 — managed provider routing' : 'tier 2 — marketplace routing'}`));
   console.log(chalk.dim(`  Budget:   max $${TEST_MAX_PRICE.toFixed(2)}/hr · 2 minute cap (~$0.05 max)`));
   console.log();
 
@@ -59,32 +101,50 @@ export async function testCommand(config, chalk) {
   let depId;
 
   // ── 1. Provision ─────────────────────────────────────────────────────────
-  process.stdout.write(chalk.dim('  Provisioning GPU...'));
+  process.stdout.write(chalk.dim(`  Provisioning GPU (${tierLabel})...\n`));
   let dep;
+  const baseBody = {
+    command: TEST_COMMAND,
+    image:   TEST_IMAGE,
+    gpu:     'RTX_3080',
+    max_price_per_hour: TEST_MAX_PRICE,
+  };
   try {
     dep = await callApi('/run', {
       method: 'POST',
       apiKey: config.apiKey,
       baseUrl: config.baseUrl,
-      body: {
-        command: TEST_COMMAND,
-        image:   TEST_IMAGE,
-        tier:    '1',
-        // smoke_test workload → cheapest reliable GPU with ≥4 GB VRAM
-        gpu:     'RTX_3080',
-        max_price_per_hour: TEST_MAX_PRICE,
-      },
+      body: { ...baseBody, tier },
     });
-    depId = dep.deployment_id;
-    process.stdout.write('\n');
-    step(chalk, true, 'Provisioned', `${dep.deployment_id} on ${dep.gpu_type}`);
   } catch (err) {
-    process.stdout.write('\n');
-    step(chalk, false, 'Provisioned', err.message);
-    console.log();
-    console.error(chalk.red('  Test failed — could not provision GPU.\n'));
-    process.exit(1);
+    if (err.errorData?.code === 'NO_CAPACITY_MATCH' && tier === '1') {
+      process.stdout.write('\n');
+      process.stdout.write(chalk.dim('  No tier 1 capacity — trying tier 2 marketplace routing...'));
+      try {
+        dep = await callApi('/run', {
+          method: 'POST',
+          apiKey: config.apiKey,
+          baseUrl: config.baseUrl,
+          body: { ...baseBody, tier: '2' },
+        });
+      } catch (err2) {
+        process.stdout.write('\n');
+        step(chalk, false, 'Provisioned', err2.message);
+        console.log();
+        console.error(chalk.red('  Test failed — no GPU capacity available on any provider.\n'));
+        process.exit(1);
+      }
+    } else {
+      process.stdout.write('\n');
+      step(chalk, false, 'Provisioned', err.message);
+      console.log();
+      console.error(chalk.red('  Test failed — could not provision GPU.\n'));
+      process.exit(1);
+    }
   }
+  depId = dep.deployment_id;
+  process.stdout.write('\n');
+  step(chalk, true, 'Provisioned', `${dep.deployment_id} on ${dep.gpu_type}`);
 
   // ── 2. Container started ─────────────────────────────────────────────────
   process.stdout.write(chalk.dim('  Waiting for container to start...'));

package/tests/commands.test.js

@@ -1,7 +1,7 @@
 import { describe, it, expect, vi } from 'vitest';
 import { parseRunArgs, classifyFailure, inferWorkload } from '../src/commands/run.js';
 import { parseServeArgs } from '../src/commands/serve.js';
-import { testCommand } from '../src/commands/test-run.js';
+import { testCommand, parseTestArgs } from '../src/commands/test-run.js';
 import { rankAlternatives, diffDescription, promptFallback } from '../src/fallback.js';
 
 describe('parseRunArgs', () => {
@@ -73,20 +73,15 @@ describe('parseRunArgs', () => {
     expect(flags.maxCost).toBe(5.0);
   });
 
-  it('parses --cheap flag', () => {
-    const { flags } = parseRunArgs(['python', 'train.py', '--cheap']);
-    expect(flags.cheap).toBe(true);
-  });
-
-  it('--cheap defaults to falsy when not passed', () => {
-    const { flags } = parseRunArgs(['python', 'train.py']);
-    expect(flags.cheap).toBeFalsy();
-  });
-
   it('parses --tier 2 flag', () => {
     const { flags } = parseRunArgs(['python', 'train.py', '--tier', '2']);
     expect(flags.tier).toBe('2');
   });
+
+  it('--tier defaults to undefined when not passed', () => {
+    const { flags } = parseRunArgs(['python', 'train.py']);
+    expect(flags.tier).toBeUndefined();
+  });
 });
 
 describe('classifyFailure', () => {
@@ -204,14 +199,14 @@ describe('parseServeArgs', () => {
     expect(model).toBeNull();
   });
 
-  it('parses --cheap flag', () => {
-    const { flags } = parseServeArgs(['my/model', '--cheap']);
-    expect(flags.cheap).toBe(true);
+  it('parses --tier 2 flag', () => {
+    const { flags } = parseServeArgs(['my/model', '--tier', '2']);
+    expect(flags.tier).toBe('2');
   });
 
-  it('--cheap defaults to falsy when not passed', () => {
+  it('--tier defaults to undefined when not passed', () => {
     const { flags } = parseServeArgs(['my/model']);
-    expect(flags.cheap).toBeFalsy();
+    expect(flags.tier).toBeUndefined();
   });
 });
 
@@ -221,6 +216,24 @@ describe('testCommand', () => {
   });
 });
 
+describe('parseTestArgs', () => {
+  it('returns empty flags for no args', () => {
+    expect(parseTestArgs([])).toEqual({});
+  });
+
+  it('parses --provider tier1', () => {
+    expect(parseTestArgs(['--provider', 'tier1'])).toEqual({ provider: 'tier1' });
+  });
+
+  it('parses --provider tier2', () => {
+    expect(parseTestArgs(['--provider', 'tier2'])).toEqual({ provider: 'tier2' });
+  });
+
+  it('parses --provider secondary', () => {
+    expect(parseTestArgs(['--provider', 'secondary'])).toEqual({ provider: 'secondary' });
+  });
+});
+
 describe('inferWorkload', () => {
   it('returns general for empty command', () => {
     expect(inferWorkload([])).toBe('general');