derive-budget-three-strategies
IN premise
`_build_beliefs_section` supports three budget strategies — alphabetical truncation (default), random sampling (`sample=True`), and semantic clustering (`cluster=True`) — all controlled by a single `max_beliefs` parameter.
Summary
The belief section builder has three different ways to fit beliefs into a limited budget: it can cut off alphabetically, pick a random sample, or group them by meaning. All three modes share the same size cap, so callers just choose a strategy and set the limit without worrying about separate configuration for each approach.
Dependents
These beliefs depend on this one:
- derive-budget-is-flexible-and-efficient — The derive pipeline's budget allocation is both strategically flexible (three selection strategies: alphabetical truncation, random sampling, semantic clustering) and computationally efficient (linear accumulation with guaranteed floor of 5 per agent group), enabling callers to trade off reproducibility, diversity, and semantic coherence without performance penalty.
Details
| Source | entries/2026/05/08/reasons_lib-derive.md |