Yore – Deterministic document indexer for large, agent-driven codebases
11 days ago
- #LLM-context
- #documentation-tools
- #search-algorithms
- Yore is a deterministic tool for indexing, analyzing, and retrieving documentation, optimized for LLM context assembly.
- It addresses 'documentation sprawl' by identifying duplicates, canonical documents, and deprecated content.
- Yore supports Architecture Decision Records (ADRs) and their chains to provide historical context for 'why' questions.
- Features include BM25 search, structural analysis, link graph inspection, duplicate detection, and extractive refinement.
- It provides canonicality scoring based on path, naming conventions, recency, and other signals.
- Yore is designed for LLM context assembly with token budgeting, cross-reference expansion, and extractive refinement.
- It includes quality checks like link validation and an evaluation harness for retrieval correctness.
- Yore is not a replacement for general-purpose search tools but builds on them for documentation-specific needs.
- It operates in four phases: indexing, retrieval and analysis, context assembly for LLMs, and evaluation.
- Yore is deterministic, ensuring the same inputs produce the same outputs, which is crucial for automation and testing.