Skip to main content

Content Management

Sources of information​

Assistants utilize various sources of information to provide accurate and relevant responses.

  • Searches: Utilizes advanced search algorithms to retrieve information from preloaded documents, transcriptions, etc.
  • FAQ: FAQs serve as a reference for the assistant when generating answers. The assistant will take this content into account when answering similar questions.
  • Fixed:Fixed responses are used to define topics for which the assistant should not have the freedom to generate a response but instead return a predefined one. This can be useful for sensitive topics or when a generic response is desired for any related question.

For use cases where the task involves generating a response based on internal documentation, there are different levels of freedom for both information retrieval and response generation. They are detailed below:

  • The model is allowed to choose between different searches. A semantic search is then performed to select the extracts relevant to the query. With these excerpts, the model generates a response.
  • A single search is used, where a semantic search is performed to select the extracts relevant to the query. With these excerpts, the model generates a response.
  • When frequently asked questions are used, the model has less possibility of modifying the response.
  • When fixed responses are configured and used, the response is not generated by the model, but previously configured by an administrator.

Context​

Additional context can be added to enhance understanding. This context can be added in three diferent levels.

  • Blocks: Metadata is added to chunks of information.
  • Documents: Metadata is added to the document and applies to all the chunks in it.
  • Lessons: Lessons are context that are configured globally but will apply only when que query is related to the topic of the lesson.

These information sources and configurations ensure that the assistant delivers precise and contextually relevant information to users.

Documentation organization​

Documentation can be organized in a tree-like structure, allowing the end user to specify smaller sets of information for certain conversations. The tree definition is configurable through the implementation portal.