BRain Language Vocab README

Brain Language Vocabulary — Setup & Operating Instructions

  • Purpose: Prefill brain_language large token IDs (e.g. 10000, 30000, etc.) per environment and load them at startup to ensure decoder stability.

  • Placement: Place the sample file in config/brain_language_vocab.{environment}.json.

  • Examples: config/brain_language_vocab.development.json, config/brain_language_vocab.staging.json, config/brain_language_vocab.production.json

  • Reference from settings (settings.yaml):

  • Merge config/settings.yaml and config/settings.{env}.yaml and check the following keys:
brain_language:
  vocab_file: brain_language_vocab.development.json
  • vocab_file can be a relative path (under config/) or an absolute path.

  • If not specified, config/brain_language_vocab.json will be loaded by default.

  • Startup behavior:

  • The module evospikenet.eeg_integration.brain_language_decoder checks the above settings during import, reads the specified JSON, and registers the vocabulary in the shared decoder _decoder_instance.

  • JSON keys can be strings or numbers, but internally they are converted to integer token IDs.

  • Operating procedure (example):

  • Create environment JSON: config/brain_language_vocab.production.json
  • Set brain_language.vocab_file in config/settings.production.yaml

  • Automatically loaded when starting the app (or running a test)

  • Test:

  • Unit test: Run pytest tests/unit/eeg_integration (recommended in a container).

  • Note:

  • Loading at startup is a best effort and does not throw an exception even if it fails, so check the log to make sure the intended file is being loaded.
  • If you want to update the decoder vocabulary, you can add it at runtime using BrainLanguageDecoder.update_vocab().

Please let me know if you have any questions about file placement or operation. The added README can be found in docs/BRain_Language_Vocab_README.md.