PropositionAbstractor

Generates higher-level abstract propositions from a group of related propositions.

This is an "active abstraction" operation - given N propositions about an entity, topic, or context, it synthesizes higher-level insights that capture the essence of the group.

Example:

Input propositions about Bob:
  - "Bob prefers morning meetings"
  - "Bob likes detailed documentation"
  - "Bob reviews PRs thoroughly"

Output abstraction:
  - "Bob values thoroughness and clarity in work processes"

Abstractions are stored as regular propositions with:

• level = max(source levels) + 1

• sourceIds = IDs of source propositions

• decay = average decay of sources

• confidence = LLM-assessed confidence in the abstraction

Inheritors