Translation Depth - protospatial/NodeToCode GitHub Wiki

Best Practices → Translation Depth

Setting Appropriate Depth

The Max Translation Depth setting affects how Node to Code translates nested user-created graphs like functions, macros, collapsed graphs, and math expressions into code:

Depth Use Case Considerations
0 Single graph translation Fastest, lowest cost
1 Include direct function calls Good for most cases
2 Include nested functions Balance of depth/cost
3+ Deep system translation Higher cost, longer processing

Depth Configuration Guidelines

  1. Start Conservative

    • Begin with depth 1 for most translations
    • Evaluate results before increasing depth
    • Consider token usage and costs

  2. When to Increase Depth

    • When you know there are multiple nested graphs within your top-level graph that you will need to translate anyways and are comfortable with the increase in compute time

[!IMPORTANT] Higher translation depths can exponentially increase token usage and LLM processing time and are more prone to LLM context window issues.

  1. When to Decrease Depth

    • Token limit concerns
    • Quick reference translations
    • Simple graph conversion
    • Cost optimization needed