Skip to content

DEBUGGING_GUIDE.md

Latest commit

 

History

History
110 lines (86 loc) · 2.87 KB

DEBUGGING_GUIDE.md

File metadata and controls

110 lines (86 loc) · 2.87 KB

Debugging Guide - Application Page Stuck Loading

Issue Summary

The application page gets stuck on "Generating your application..." when the workflow is in "waiting_for_input" status.

Root Cause

The application page only handled two statuses:

  • "complete" → Show data
  • "processing" → Poll again

But didn't handle:

  • "waiting_for_input" → Interview required
  • "processing_resume" → Optimizer/Ghostwriter running

Fix Applied

Added redirect logic in frontend/app/application/page.tsx (lines 49-54):

} else if (data.status === "waiting_for_input") {
    // Interview required - redirect to AI-Help page
    router.push(`/ai-help?session=${session_id}`);
} else if (data.status === "processing" || data.status === "processing_resume") {
    // Still processing, poll again
    setTimeout(fetchData, 2000);
}

How to Debug in the Future

1. Check Session Status

curl http://localhost:8000/api/workflow/status/<SESSION_ID> | python3 -m json.tool

2. Workflow Status Flow

"processing" → Initial workflow running (Scout, Profiler, Decoder, Matchmaker)
    ↓
"waiting_for_input" → Interview required (match_score < 0.8)
    ↓
User completes interview
    ↓
"processing_resume" → Optimizer + Ghostwriter running
    ↓
"complete" → Ready to display on application page

3. Common Issues

Issue: Stuck at "waiting_for_input"

Solution: Complete the interview at /ai-help?session=<ID>

Issue: Stuck at "processing"

Possible causes:

  • Backend crashed
  • Agent error
  • ChromaDB not running

Check backend logs:

# Look for errors in the terminal running uvicorn

Issue: Stuck at "processing_resume"

Possible causes:

  • Optimizer agent error
  • Ghostwriter agent error
  • LLM API timeout

Check:

# Check if LLM client is working
# Check API keys in .env

4. Force Complete Interview (Testing Only)

If you want to bypass the interview for testing:

curl -X POST http://localhost:8000/api/workflow/resume \
  -F "session_id=<SESSION_ID>" \
  -F "bridge_story=Test story for debugging"

5. Restart Workflow

If a session is stuck, start a new one:

  1. Go to http://localhost:3000/start
  2. Upload resume and scholarship URL again
  3. You'll get a new session ID

Verification Checklist

After the fix:

  • Application page redirects to AI-Help if interview needed
  • Application page polls while processing_resume
  • Application page shows data when complete
  • Matchmaker page loads correctly
  • AI-Help page works correctly

Testing the Full Flow

  1. Start: Upload resume + URL
  2. Matchmaker: See match score, click "Continue to AI Help"
  3. AI-Help: Answer interview questions
  4. Processing: Wait for optimizer/ghostwriter (will show loading)
  5. Application: See essay + resume optimizations

Estimated time: 30-60 seconds for full workflow.