Configuration

The application may be configured by editing the settings in Config.toml. The main settings are:

• embeddings: the model to create embeddings.

  • dense_model: the model to create dense embeddings. For example, OpenAI embedding are set by "OpenAI/text-embedding-3-small"
  • sparse_model: the model for sparse embeddings. The sparse embeddings supported by Qdrant are used, i.e. "Qdrant/bm25"

• major: the main language model used, for example, for chatting.

  • model: the provider and the model name, e.g. "OpenAI/gpt-4.1-mini"
  • temperature: temperature (defaults to 0.1)
  • max_retries: number of connection attempts when using model (default to 2)

• major.provider_params: other parameters set for the model (model-dependent)

• minor: the accessory language model used, for example, for creating questions and for creating summaries.

  • model: the provider and the model name, e.g. "OpenAI/gpt-4.1-nano"
  • temperature: temperature (defaults to 0.1)
  • max_retries: number of connection attempts when using model (default to 2)

• major.provider_params: other parameters set for the model (model-dependent)

• aux: the auxiliary language model used, for example to classify text. This model should have low latency.

  • model: the provier and the model name, e.g. "Mistral/mistral-small-latest"
  • temperature: temperature (defaults to 0.7)
  • max_retries: number of connection attempts when using model (default to 2)

• aux.provider_params: other parameters set for the model (model-dependent)