Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.agentbot.raveculture.xyz/llms.txt

Use this file to discover all available pages before exploring further.

​
Troubleshooting

Common issues when building, deploying, and running Agentbot agents. Agentbot troubleshooting

​
Installation Issues

npm install --legacy-peer-deps
Agentbot uses React 19 which some packages haven’t updated for yet. The --legacy-peer-deps flag resolves most conflicts.
npx prisma generate
Prisma client must be generated before build. If you see @prisma/client did not initialize, run the generate command manually.
lsof -i :3000
kill -9 <PID>
Or use a different port:
PORT=3001 npm run dev

​
Deployment Issues

Check that all imports use the @/ alias correctly. The @/ prefix maps to the web/ directory. Common mistake:
// ❌ Wrong
import { auth } from '../../../lib/auth'

// ✅ Right
import { auth } from '@/app/lib/auth'
The health check at /health must respond within 60 seconds. Common causes:
  • Database connection string is wrong (check DATABASE_URL)
  • Missing environment variables causing startup crash
Check Railway logs for the actual error.
Agentbot reads env vars at startup. After changing them:
  • Vercel: Redeploy from dashboard or vercel --prod
  • Railway: Service auto-restarts on env change
  • Local: Restart the dev server (npm run dev)

​
Agent Issues

  1. Check agent status: GET /api/agents/{id}
  2. Verify the channel token (Telegram/Discord/WhatsApp) is valid
  3. Check the agent container logs
  4. Ensure the AI provider API key is set and has credits
Each agent has a memory limit based on plan:
PlanMemoryCPUDescription
Solo2 GB1 vCPUTrial / light workloads only
Collective4 GB2 vCPURecommended production floor
Label8 GB4 vCPUHeavy production + browser/tool work
Network16 GB8 vCPUHigh-throughput production
If you’re hitting memory limits, check for memory leaks in custom skills or upgrade your plan. The Solo tier (1 vCPU / 2 GB) is the trial/light-use floor — for production workloads, we recommend at least Collective (2 vCPU / 4 GB).
# Open the user-facing skills surface
GET /dashboard/skills

# Marketplace/API install flow
POST /api/skills
{ "skill": "weather" }
Skills are loaded at agent startup. If a skill fails to load, the agent will start without it and log the error.
If you see “Agent offline. Install your agent first, then retry installing skills.” in the dashboard, the skill was saved to your account but the agent’s gateway is not reachable.This happens when the API returns "deployed": false with a deployWarning field. To resolve:
  1. Verify your agent is running: GET /api/agents/{id}
  2. Start or restart the agent from the dashboard
  3. The skill will sync automatically once the agent comes back online
  4. You can also trigger a manual sync using POST /api/agents/{id}/sync
See the Skills API reference for full response details.

​
Billing Issues

The legacy /api/checkout endpoint is deprecated. Use:
GET /api/stripe/checkout?plan=solo
  1. Verify your API key is valid (test it with curl)
  2. Check the provider name matches exactly: openrouter, anthropic, openai, google, groq
  3. Ensure the key has credits/quota remaining
  4. Disable and re-enable BYOK from the billing dashboard

​
Getting Help