Skip to content

Overview

mish answers one question: where did the disk space go? It scans a filesystem into an immutable snapshot, serves an interactive web client from a headless daemon, and exposes the same index through a JSON-first CLI.

A snapshot is a directory of flat files, not a database. Queries read it; they never re-crawl the filesystem.

It contains four flat files:

  • manifest.json: metadata, stats, the extension table and scan errors.
  • nodes.bin: fixed-size node records, with the pre-order index as the id.
  • names.bin: concatenated names.
  • subtree.bin: one u32 subtree size per node.

Nodes are emitted in pre-order, so every subtree is a contiguous id range. subtree_size is computed on the way back up during the scan. That one column makes a folder total a single read.

  • Architecture: the process boundary, the index format, and the data flow.
  • CLI: every command, the JSON contract, and the terminal view.
  • API: REST and SSE endpoints for the web client.
  • Docker: what is collected and how to keep the socket read-only.
  • Configuration: environment variables and defaults.
  • Security: read-only guarantees, permissions, and network exposure.
  • Troubleshooting: the failure modes and what to check first.