ashlr__read

Token-efficient file read. Snip-compacts files over 2 KB, returning head + tail with the middle summarized.

Server: ashlr-efficiency — servers/efficiency-server.ts

Replaces Claude Code's built-in Read tool. Files under 2 KB are returned verbatim. Files over 2 KB are snip-compacted: head and tail are preserved, the middle is replaced with a one-line summary. For very large files, an optional LLM summarizer (Ollama or Pro cloud) can produce a semantic digest.

Input schema

Field	Type	Required	Description
`path`	`string`	yes	Absolute or repo-relative file path
`limit`	`number`	no	Max lines to return (default: full snip)
`offset`	`number`	no	Start reading from this line number

Output format

Returns the file content as a string with a confidence badge prepended:

[A] src/server.ts (4.2 KB → 1.1 KB · −74%)
<head lines>
... [snipped 312 lines — summary: defines HTTP handler, route map, middleware chain] ...
<tail lines>

The badge tiers are [A] (high confidence), [B] (medium), [C] (low), [?] (unknown). See Confidence badges.

Example

Use ashlr__read to read servers/efficiency-server.ts

Typical result: a 1,800-line file returns ~200 lines at ~89 % savings.

Savings baseline

Benchmark median: −79.5 % on files 2 KB+. Break-even point is ~800 bytes — files smaller than that are returned verbatim.

ashlr__grep — search before reading to find the right section
ashlr__edit — make changes without re-reading the full file
ashlr__orient — structural overview before diving into a file

Input schema

Output format

Example

Savings baseline

Related tools

On this page