feat: aggregations in structured views

[mhordynski]

Feature description

db-ally's structured views should support a new type of method exposed to the LLM called aggregation, in addition to the currently supported filters. Aggregations would allow the resulting query to be modified with a wrapper function that, for example, groups the results by dimension or calculates metrics.

Motivation

Often, users may ask questions such as:

• What is the average number of years of experience for candidates in Europe?
• What is the average amount spent on purchases by clients in the loyalty program, grouped by city?

Additional context

Based on our initial brainstorming session, we decided on the following:

• Only one aggregation may be chosen for a given question.

• Aggregations should be able to consume extra arguments (similar to filters). For example:

  @decorators.aggregation()
  def avg_amount_spent(self, grouped_by: Literal['city', 'country', 'gender']):
      ...

• To ensure that the output schema is well-known, additional context should be added to the ExecutionResult class:

  ExecutionResult(
       view_name="CandidatesView",
       used_aggregation={
            "name": "avg_amount_spent",
            "args": {
                "group_by": "city"
            }
       }
  )