@obversa/memory-markdown

@obversa/memory-markdown finds matching passages in a local Markdown directory. It returns the file path and line range for each hit, so a caller can hand only the selected files to ground and then bound them with curate.

Install

pnpm add @obversa/memory-markdown @obversa/runtime

Requirements

• Node.js 22.12 or later

Read-only corpus

openMarkdownCorpus exposes a memory property because ground reads sources through the public Memory contract. That view is read-only: view reads the corpus, while create, str_replace, insert, delete, and rename return an error. Search must not edit the source material it is selecting.

Search and the Memory view include only directories and .md files whose path segments start with a letter or number and then use letters, numbers, dots, underscores, or hyphens, up to 128 characters per segment. Dot-prefixed names, names outside that rule, non-Markdown files, and symbolic links inside the corpus are skipped without failing the search or a directory view.

One class is not skipped. A file whose full path passes 1024 bytes stops the search, which names the path it cannot use. A directory view still lists that file, and the over-long path it returns then makes ground refuse the whole directory. Keep the corpus shallow enough to stay below the limit.

The corpus root itself can be a symbolic link.

Runnable example

From the workspace root, run:

pnpm example:memory-markdown

The complete source is on the memory-markdown package page.

License

MIT