GPT-Tokenizer icon

GPT-Tokenizer

Encode / decodes BPE Tokens or check Token Limits before working with the OpenAI GPT models.

Overview

The GPT-Tokenizer node (Resource: Default, Operation: Decode) decodes an array of BPE (Byte Pair Encoding) tokens back into a string. This is particularly useful when working with OpenAI GPT models or other systems that use BPE tokenization, allowing you to convert tokenized data back into human-readable text.

Common scenarios:

  • Decoding model outputs or intermediate results stored as token arrays.
  • Integrating with APIs or workflows that require conversion from tokens to strings.
  • Post-processing results from language models for display or further analysis.

Example:
If you have a list of tokens like [5661, 318, 1337], this node will output the corresponding decoded string.

Properties

Name Meaning
Input Tokens Array of BPE tokens to decode. Example: [5661, 318, 1337]
Destination Key The key in the output JSON where the decoded string will be stored. If left empty, defaults to data.

Output

  • The node outputs a JSON object for each input item.
  • The decoded string is placed under the key specified by Destination Key (defaults to data if not set).

Example output:

{
  "data": "Decoded string here"
}

Dependencies

  • External library: gpt-tokenizer (used for decoding tokens)
  • No API keys or external services are required for the decode operation.

Troubleshooting

Common issues:

  • "Input Tokens is not an array":
    The value provided for Input Tokens must be a valid array (e.g., [123, 456]). Ensure you are passing a proper array and not a string or other type.
  • "Input Tokens field is empty":
    You must provide at least one token in the array. An empty array is not allowed.
  • Output key missing:
    If you do not specify a Destination Key, the result will be under the default key data.

How to resolve:

  • Double-check your Input Tokens property to ensure it is a non-empty array.
  • Use the correct format for arrays in n8n expressions or static values.

Links and References

Discussion