Skip to content

Troubleshooting Guide

This guide is intended to help you diagnose and resolve common issues you may encounter when using AgentKit. If you cannot find a solution here, feel free to contact us via GitHub Issues.

Installation Issues

1. agentkit command not found

  • Symptom: When running the agentkit command in the terminal, the system reports command not found.
  • Cause: This is usually because the executable path installed by pip has not been added to the system PATH environment variable.
  • Solution:
    • Confirm install location: Run pip show agentkit-sdk-python or uv pip show agentkit-sdk-python and find the install path indicated by the Location field.
    • Locate the executable: Under the install path’s bin directory (for example .../site-packages/bin), you should be able to find the agentkit executable.
    • Add to PATH: Add the absolute path of this bin directory to your ~/.bashrc, ~/.zshrc, or other shell configuration files.
      bash
      export PATH="/path/to/your/python/bin:$PATH"
    • Reload configuration: Run source ~/.bashrc or restart the terminal to apply the changes.

2. Dependency conflicts

  • Symptom: When installing agentkit-sdk-python, pip reports dependency version conflicts.
  • Cause: Your Python environment may already contain library versions that are incompatible with AgentKit.
  • Solution:
    • Use a virtual environment (recommended): We strongly recommend using uv or venv to create a clean virtual environment in your project to avoid conflicts with system-level Python packages.
      bash
      uv venv
      source .venv/bin/activate
      uv pip install agentkit-sdk-python
    • Clean and reinstall: If you must use the existing environment, uninstall and reinstall:
      bash
      pip uninstall agentkit-sdk-python
      pip install agentkit-sdk-python

Configuration Issues

1. Environment variables not taking effect

  • Symptom: Prompts indicate it cannot connect to AgentKit Platform or authentication fails.
  • Cause: Required environment variables may not have been passed in, or incorrect environment variable names were used.
  • Solution:
    • Confirm required environment variables are set
    bash
    echo $VOLCENGINE_ACCESS_KEY
    echo $VOLCENGINE_SECRET_KEY
    • Ensure there are no extra spaces or quotes
    • Re-export the environment variables in the current shell session

2. Configuration file format error

  • Symptom: When running agentkit launch, it reports configuration file parsing failure.
  • Cause: File format issues
  • Solution:
    • Check whether agentkit.yaml is formatted correctly (pay attention to indentation)
    • Re-run agentkit config to generate configuration

Deployment Issues

1. CR creation failed

  • Symptom:
CreateRegistry: QuotaExceeded.Registry The quota of Registry is exceeded.
  • Cause: Insufficient account quota
  • Solution:
    • Number of CR instances exceeds the quota limit
    • In agentkit config, configure it to use an existing CR instance name
    • Or contact the administrator to increase quota

2. Image build failed

  • Symptom: Pipeline build fails and reports dependency installation errors.
  • Cause: Dependency installation failure
  • Solution:
    • Check whether dependencies in requirements.txt are correct
    • Confirm dependency versions are compatible with Python 3.12
    • Check Pipeline logs for detailed error information (logs will be automatically downloaded to your local root directory; pay attention to prompts in the interactive UI and find a file named like pipeline_failed_****ff20ce223.log in the root directory)

3. Runtime deployment timeout

  • Symptom: Runtime stays in Releasing state for a long time.
  • Cause: The first deployment takes longer, or there may be insufficient resources.
  • Solution:
    • The first deployment usually takes 2–3 minutes; please be patient
    • Use agentkit status to check status
    • If it is still not ready after 5 minutes, resources may be insufficient; try agentkit destroy and then redeploy

4. Runtime status abnormal

  • Symptom: Runtime status is Failed or Error.
  • Cause: Internal Runtime exception
  • Solution:
    • Check whether environment variable configuration is correct (especially the model API Key)
    • Check logs in the AgentKit Platform console
    • Confirm the application code has no startup errors
    • Try cleaning up with agentkit destroy and redeploying

Invocation Issues

1. invoke failed

  • Symptom: When running agentkit invoke, it reports connection failure or timeout.
  • Cause: Network issues or incorrect Endpoint configuration
  • Solution:
    • Use agentkit status to confirm Runtime status is Ready
    • Check network connectivity
    • Confirm the Endpoint address is correct
    • Check firewall or proxy settings

2. Model invocation failed

  • Symptom: The Agent returns an error indicating model access failure.
  • Cause: Incorrect model API Key configuration or exhausted model quota
  • Solution:
    • Confirm the Ark model API Key configuration is correct
    • Check whether the endpoint ID is valid
    • Confirm whether model quota has been exhausted
    • Log in to the Ark platform to check API Key permissions

Permission Issues

1. AK/SK authentication failed

  • Symptom: Reports InvalidAccessKeyId or SignatureDoesNotMatch.
  • Cause: Incorrect AK/SK configuration or insufficient permissions
  • Solution:
    • Confirm AK/SK is correct and valid
    • Check whether you have access permissions for the AgentKit service
    • Confirm AK/SK has not been disabled or expired
    • Contact the administrator to assign the required permissions

2. Resource access denied

  • Symptom: Reports AccessDenied or insufficient permissions.
  • Cause: The account does not have enough permissions to perform the operation
  • Solution:
    • Confirm the account has the required permissions for the relevant resources
    • Check IAM role configuration
    • Contact the administrator to grant necessary permissions

Getting Help

If the above solutions cannot resolve your issue, please:

  1. Check logs:

    • Local logs: Check command line output
    • Platform logs: Visit the AgentKit console to view detailed logs
  2. Contact support:

    • Provide detailed error information and logs
    • Describe reproduction steps
    • Attach relevant configuration files (be sure to redact sensitive data)
    • Contact Volcengine technical support
  3. Read the docs:

Released under the Apache-2.0 License.