readseek
readseek is a structural source reader for scripts, editors, and coding agents.
It emits pretty-printed JSON with stable LINE:HASH anchors, structural symbol
maps, parse diagnostics, AST search matches, references, and rename plans.
Current documented CLI API: 0.4.x.
Install
Or build the native binary from source:
Common commands
Use --stdin <path> with detect, read, map, check, symbol, and
identify to analyze unsaved editor buffers while still providing a path for
language detection:
|
Use def --from-identify to pipe identify JSON into definition lookup:
|
Images
detect reports format, dimensions, and animation status for images. Add a vision
flag to analyze image contents with the Qwen2.5-VL vision model:
The flags can be combined; the model loads once per invocation. The model files (~1.8 GB GGUF + ~1.3 GB multimodal projection) are downloaded lazily into the user cache directory on first vision use and reused on subsequent runs; a progress bar is shown while downloading when stdout is an interactive TTY. Inference is CPU-only and takes a few to tens of seconds per image.
Cache
readseek init [path] creates a .readseek/ directory containing map cache files
under maps/ and definition-index shards under def-index/. Commands discover
that directory by walking up from the target path, or use the directory passed by
--readseek-dir.
Documentation
The manual page is the authoritative CLI reference:
Pass --help to any command for command-specific usage.
Licensing
The JavaScript npm wrapper is licensed under Apache-2.0. The Rust source and
native binaries are licensed under LGPL-2.1-or-later. Corresponding source for
each published native binary is available from the GitHub repository tag that
matches the package version.
readseek downloads the Qwen2.5-VL vision model (unsloth/Qwen2.5-VL-3B-Instruct-GGUF,
a re-export of Qwen/Qwen2.5-VL-3B-Instruct) into the user cache directory on
first use. It is licensed under Apache-2.0.