badgr-cli 1.0.22 → 1.0.23

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "badgr-cli",
-  "version": "1.0.22",
+  "version": "1.0.23",
   "description": "Badgr, run or serve GPU workloads from one command",
   "type": "module",
   "bin": {

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
@@ -454,56 +453,48 @@ 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 (RunPod) is out of capacity — offer budget providers.
+      if (process.stdin.isTTY) {
+        const answer = await askConfirm(
+          `\n  RunPod has no suitable capacity. Try budget marketplace GPUs? (Vast.ai, Salad) [${chalk.bold('Enter')}/${chalk.bold('q')}]: `
+        );
+        if (answer.toLowerCase() === 'q') {
+          console.log(chalk.dim('\n  Cancelled.\n'));
+          process.exit(0);
+        }
+      } else {
+        console.log(chalk.dim('\n  RunPod capacity unavailable — falling back to budget providers...\n'));
       }
 
-      console.log(chalk.dim(`\n  Trying ${chosen.gpu} in ${chosen.region}...\n`));
-
+      console.log(chalk.dim('  Searching budget providers (Vast.ai, Salad)...'));
       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`));
           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 budget providers: ${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
@@ -107,46 +112,45 @@ 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 (RunPod) out of capacity — offer budget providers.
+      if (process.stdin.isTTY) {
+        const answer = await askConfirm(
+          `\n  RunPod has no suitable capacity. Try budget marketplace GPUs? (Vast.ai, Salad) [${chalk.bold('Enter')}/${chalk.bold('q')}]: `
+        );
+        if (answer.toLowerCase() === 'q') {
+          console.log(chalk.dim('\n  Cancelled.\n'));
+          process.exit(0);
+        }
+      } else {
+        console.log(chalk.dim('\n  RunPod capacity unavailable — falling back to budget providers...\n'));
       }
 
-      console.log(chalk.dim(`\n  Trying ${chosen.gpu} in ${chosen.region}...\n`));
-
+      console.log(chalk.dim('  Searching budget providers (Vast.ai, Salad)...'));
       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  ✗ Failed to start endpoint on budget providers: ${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

@@ -59,32 +59,50 @@ export async function testCommand(config, chalk) {
   let depId;
 
   // ── 1. Provision ─────────────────────────────────────────────────────────
-  process.stdout.write(chalk.dim('  Provisioning GPU...'));
+  process.stdout.write(chalk.dim('  Provisioning GPU (RunPod)...'));
   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: '1' },
     });
-    depId = dep.deployment_id;
-    process.stdout.write('\n');
-    step(chalk, true, 'Provisioned', `${dep.deployment_id} on ${dep.gpu_type}`);
   } catch (err) {
-    process.stdout.write('\n');
-    step(chalk, false, 'Provisioned', err.message);
-    console.log();
-    console.error(chalk.red('  Test failed — could not provision GPU.\n'));
-    process.exit(1);
+    if (err.errorData?.code === 'NO_CAPACITY_MATCH') {
+      process.stdout.write('\n');
+      process.stdout.write(chalk.dim('  RunPod unavailable, trying budget providers...'));
+      try {
+        dep = await callApi('/run', {
+          method: 'POST',
+          apiKey: config.apiKey,
+          baseUrl: config.baseUrl,
+          body: { ...baseBody, tier: '2' },
+        });
+      } catch (err2) {
+        process.stdout.write('\n');
+        step(chalk, false, 'Provisioned', err2.message);
+        console.log();
+        console.error(chalk.red('  Test failed — no GPU capacity available on any provider.\n'));
+        process.exit(1);
+      }
+    } else {
+      process.stdout.write('\n');
+      step(chalk, false, 'Provisioned', err.message);
+      console.log();
+      console.error(chalk.red('  Test failed — could not provision GPU.\n'));
+      process.exit(1);
+    }
   }
+  depId = dep.deployment_id;
+  process.stdout.write('\n');
+  step(chalk, true, 'Provisioned', `${dep.deployment_id} on ${dep.gpu_type}`);
 
   // ── 2. Container started ─────────────────────────────────────────────────
   process.stdout.write(chalk.dim('  Waiting for container to start...'));