Attempt to use the API

ui/src/app/api/hf-jobs/route.ts

@@ -1033,81 +1033,45 @@ async function checkHFJobStatus(token: string, jobId: string, jobNamespace?: str
 }
 
 async function checkHFJobsCapacity(token: string): Promise<any> {
-  return new Promise(async (resolve, reject) => {
-    console.log('Checking HF Jobs capacity for namespace: lora-training-frenzi');
-
-    // Create a temporary file to store the output
-    const tempFile = path.join(tmpdir(), `hf_jobs_ps_${Date.now()}.txt`);
-    console.log(`Writing output to temp file: ${tempFile}`);
-
-    // Use shell redirection to write to file
-    const command = `hf jobs ps --namespace lora-training-frenzi --token "${token}" > "${tempFile}" 2>&1`;
+  try {
+    console.log('Checking HF Jobs capacity for namespace: lora-training-frenzi via API');
 
-    const childProcess = spawn('sh', ['-c', command], {
-      env: {
-        ...process.env,
-        HF_TOKEN: token,
-        TERM: 'dumb',
-        NO_COLOR: '1',
-      }
+    // Use HuggingFace API directly instead of CLI to avoid TTY issues
+    const response = await fetch('https://huggingface.co/api/jobs/lora-training-frenzi', {
+      headers: {
+        'Authorization': `Bearer ${token}`,
+      },
     });
 
-    childProcess.on('close', async (code) => {
-      console.log(`hf jobs ps process exited with code: ${code}`);
-
-      try {
-        // Read the output from the temporary file
-        const output = await readFile(tempFile, 'utf-8');
-
-        console.log('=== RAW OUTPUT START ===');
-        console.log(output);
-        console.log('=== RAW OUTPUT END ===');
-
-        // Count RUNNING jobs in the output
-        // Split by newline and filter out empty lines
-        const lines = output.split(/\r?\n/).filter(line => line.trim().length > 0);
-        let runningCount = 0;
-
-        console.log(`Total non-empty lines in output: ${lines.length}`);
-
-        for (let i = 0; i < lines.length; i++) {
-          const line = lines[i];
-          console.log(`Line ${i}: "${line}"`);
-
-          // Check if line contains RUNNING (case-sensitive as shown in your output)
-          if (line.includes('RUNNING')) {
-            runningCount++;
-            console.log(`  ✓ Line ${i} contains RUNNING (count: ${runningCount})`);
-          }
-        }
+    if (!response.ok) {
+      throw new Error(`API request failed: ${response.status} ${response.statusText}`);
+    }
 
-        const atCapacity = runningCount >= 32;
+    const jobs = await response.json();
+    console.log(`Fetched ${jobs.length} total jobs from API`);
 
-        console.log(`\n=== FINAL COUNT ===`);
-        console.log(`Found ${runningCount} RUNNING jobs. At capacity: ${atCapacity}`);
-        console.log(`==================\n`);
+    // Count jobs with status RUNNING
+    let runningCount = 0;
+    for (const job of jobs) {
+      const status = job.status?.stage || job.status;
+      if (status === 'RUNNING') {
+        runningCount++;
+      }
+    }
 
-        // Clean up temp file
-        try {
-          await unlink(tempFile);
-        } catch (unlinkError) {
-          console.warn('Failed to delete temp file:', unlinkError);
-        }
+    const atCapacity = runningCount >= 32;
 
-        resolve({
-          runningJobs: runningCount,
-          atCapacity,
-          capacityLimit: 32,
-        });
-      } catch (parseError: any) {
-        console.error('Failed to read or parse jobs ps output:', parseError);
-        reject(new Error('Failed to parse capacity status'));
-      }
-    });
+    console.log(`\n=== FINAL COUNT ===`);
+    console.log(`Found ${runningCount} RUNNING jobs. At capacity: ${atCapacity}`);
+    console.log(`==================\n`);
 
-    childProcess.on('error', (err) => {
-      console.error('HF Jobs ps process error:', err);
-      reject(new Error(`Process error: ${err.message}`));
-    });
-  });
+    return {
+      runningJobs: runningCount,
+      atCapacity,
+      capacityLimit: 32,
+    };
+  } catch (error: any) {
+    console.error('Failed to check capacity via API:', error);
+    throw new Error(`Failed to check capacity: ${error.message}`);
+  }
 }