Skip to content

Concepts

CodeNib compiles one repository into several source-linked views. These pages explain what each view preserves, how the views align, and where their support boundaries differ.

  • Start here

    Context methods and support matrix

    See every retrieval, workflow, and agent method side by side: which views it reads, who controls the search, where it is exposed, and how far it has been evaluated.

    Compare the methods →

  • System design

    Repository context architecture

    Follow one repository snapshot from view compilation to manifests, query planning, and bounded evidence delivery.

    Read the architecture →

  • Static navigation

    SCIP index

    Follow language-aware definitions and references through SCIP or LSP backends and understand cache-level behavior.

    Understand SCIP indexing →

  • Structural view

    CodeGraph

    Learn the typed hierarchy model and query graph nodes by source-aligned symbol or line range.

    Read the hierarchy model →

  • Commit-aware updates

    Incremental graph

    See how a diff updates supported graph views in place, when verification admits the patch, and when CodeNib rebuilds.

    Explore incremental updates →

  • Retrieval planning

    RAG ops and planner

    Understand retrieval operators, query-aware path selection, fusion, expansion, reranking, and the boundary between policy and execution.

    Inspect the retrieval model →

  • Native acceleration

    Core C++ backend

    Understand decoder parity, libigraph-backed execution, fallback behavior, and the boundary between Python and native code.

    Inspect the native core →

Other index views