View Error Logs (Knowbler)

The Error Logs feature helps you stay informed about disruptions in third-party tools, such as OpenAI. These tools interact with Knowbler, but there are instances when they might fail to respond. Common reasons include system overloads or reaching the request limit of your LLM plan. By reviewing error logs, Knowbler admins can quickly identify issues, take corrective actions, and minimize the impact on knowledge generation.

Accessing Error Logs

To view the error logs:

  1. Navigate to Large Language Model and select Error Logs.

  2. The error log report includes the following columns:

    • Agent Email: The email of the agent encountering the error.

    • Time Stamp: The date and time of the error.

    • Case ID: The associated case identifier.

    • Error: The error description.

  3. Use the search function to filter logs by case ID.

  4. Apply the Date Range filter to retrieve logs for a specific period.

  5. Click the Download button to export the error logs as a CSV file.

Fig. A snapshot of the Error Logs in Large Language Model.

Error Types (OpenAI and Azure)

The errors that can fill the fourth column are of three kinds: The ones that agents can troubleshoot, the ones that the admin has to fix, and the ones that are from the LLM.

Errors that Agents can Troubleshoot

  • API Error code: 429: Too many requests

  • Python Library Error code: Timeout

The message displayed to the agent in both errors is: OpenAI (or Azure) was not able to fetch content for auto-generation. Please retry.

The agent can click "Retry" to fix the problem.

Errors that the Knowbler Admin can Fix

  • Python Library Error code: AuthenticationError

There messaged displayed to the admin is "OpenAI (or Azure) has expired."

The admins receive an email notifying them of the Authentication Error. To fix it, they can authenticate the connection with the LLM again.

Agents cannot use auto-generation after this error until the Knowbler admin has reauthenticated the key.

Errors in OpenAI or Gemini

  • API Error code: 500: Server error from OpenAI or Gemini

  • API Error code: 503: OpenAI or Gemini is overloaded

  • Python Library Error code: APIError

  • Python Library Error code: RateLimitError

  • Python Library Error code: APIConnectionError

  • Python Library Error code: InvalidRequestError

  • Python Library Error code: ServiceUnavilableError

The message displayed to the agent is "OpenAI or Gemini is down currently. Auto-generation cannot be supported." In this case, the problem lies in the LLM and there is nothing the agent or the Knowbler admin can do. If you have to write an article, when OpenAI or Gemini is down, then the only way to do that is without the auto-generation feature.

Error Types (Claude)

The errors that can fill the fourth column are of two types: The ones that agents can troubleshoot and the ones that are from the LLM.

Errors that Agents can Troubleshoot

  • 400 - Invalid request.

  • 403 - Forbidden: your API key does not have permission to use the specified resource.

  • 404 - Not found: the requested resource was not found.

Agent can click "Retry" to make a second attempt.

Errors in Claude

  • 429 - Rate limit error.

  • 500 - An unexpected error has occurred internal to Anthropic's systems.

  • 529 - Anthropic's API is temporarily overloaded.

The message displayed to the agent is "Unable to auto generate content."

Error Types (Gemini)

The errors that can fill the fourth column are of two kinds: The ones that agents can troubleshoot and the ones that and the ones that are from the LLM.

Errors that Agents can Troubleshoot

  • 400 - Invalid request.

  • 403 - Forbidden: your API key does not have permission to use the specified resource.

  • 404 - Not found: the requested resource was not found.

Agent can click "Retry" to make a second attempt.

  • Gemini integration is down

Agent can click "Retry" to make a second attempt.

Errors in Gemini

  • 429 - Rate limit error.

  • 500 - An unexpected error has occurred internal to Gemini systems.

  • 529 - The API is temporarily overloaded.