Skip to content

FAQ

This page addresses frequently asked questions and common issues.

Why is the agent's output truncated?

If you observe that the agent's output is being cut off unexpectedly, the issue may be related to the max_output_token limit set by the LLM service you are using.

For example, the DeepSeek API sets a default limit of 4,096 output tokens for the deepseek-chat model, but this can be manually extended up to 8,192 tokens.

To resolve this, you can explicitly set the max_tokens parameter in your model's configuration (ModelSettingsConfig).

model:
  # ... other provider settings
  model_settings:
    temperature: 0.3
    top_p: 0.95
    max_tokens: 8000

For more context, see this issue.

How can I resolve LLM request timeouts?

If you are encountering request timeouts, first ensure that the LLM service is operational. If the service is running correctly, you may need to increase the request timeout period.

The default timeout for the openai Python package is 600 seconds. You can override this by setting the timeout value within extra_args in your ModelSettingsConfig.

model:
  # ... other provider settings
  model_settings:
    extra_args:
      timeout: 1200 # Sets the timeout to 1200 seconds

How to use LiteLLM (or Azure) model?

Method 1: If the LiteLLM service is compatible with the openai chat.completions API, you can simply set basic environment variables in your .env file:

UTU_LLM_TYPE=chat.completions  # use the default llm calling method
# basic openai configs, see `.env.full` if you're not familiar with these configs
UTU_LLM_MODEL=
UTU_LLM_BASE_URL=
UTU_LLM_API_KEY=

Method 2: If the service need to be used by the litellm package, you should install the additional package and config the following environment variables:

UTU_LLM_TYPE=litellm  # set the llm type as litellm
# set the litellm model name. e.g. azure/gpt-5
UTU_LLM_MODEL=
# add other necessary litellm configs bellow, see https://docs.litellm.ai/docs/providers/

e.g. for Azure support, you need to set:

AZURE_API_BASE=https://<YOUR-RESOURCE-NAME>.azure.com/
AZURE_API_KEY=<AZURE_OPENAI_API_KEY>

For more context, ref this issue