openclaw - 💡(How to fix) Fix [Feature]: Add model parameter prompts (context window, max_tokens, modalities) during OpenAI-compatible provider onboarding CLI flow [1 participants]

When running openclaw onboard and selecting an OpenAI-compatible provider, users can pick a model but cannot configure model-specific parameters. Setting context window, max output tokens, and supported modalities requires manually editing ~/.openclaw/config.yaml, which is error-prone, discouraged for non-technical users, and easy to break YAML formatting.

• Manual config editing: Works but is error-prone and not user-friendly.

TL;DR

Add interactive prompts for model-specific parameters during the openclaw onboard flow for OpenAI-compatible providers to improve user experience and reduce configuration errors.

Guidance

• Identify the current implementation of the openclaw onboard flow and locate the model selection step to determine where to insert the interactive prompts.
• Design and implement optional prompts for context window, max output tokens, and modalities, allowing users to skip with sensible defaults.
• Update the configuration handling to store the user-inputted values in the ~/.openclaw/config.yaml file, ensuring correct YAML formatting.
• Test the updated onboarding flow with various OpenAI-compatible providers to verify that the interactive prompts are working as expected and that the configuration is correctly applied.

Example

# Example of updated config.yaml with user-inputted values
providers:
  openai-compatible:
    models:
      my-model:
        contextWindow: 128000
        maxOutputTokens: 4096
        modalities: [text, image]

Notes

The implementation details of the interactive prompts and configuration handling may vary depending on the specific requirements and constraints of the openclaw project. Additionally, consideration should be given to handling invalid or inconsistent user input to ensure a robust and user-friendly experience.

Recommendation

Apply a workaround by adding interactive prompts for model-specific parameters during the openclaw onboard flow, as this will improve the user experience and reduce configuration errors without requiring significant changes to the underlying architecture.