Token Consumption & Best Practices

When using AI-assisted programming (especially with large codebases), token consumption can be very fast.

Warning:

In extreme cases, a single complex question can consume millions of tokens. Please monitor your balance in the Wallet.

Tips to reduce costs:

1. Control context: Regularly clear conversation history to avoid carrying large amounts of irrelevant code in every request.
2. Ask precisely: Avoid vague instructions (e.g. "help me optimize the code"); target specific files or functions.
3. Use cost-effective models: Pricing varies greatly between models. Use high-value models like deepseek/deepseek-v3.2 for non-critical tasks.