simple-workspace-contextualizer 0.1.0

Serialize a repository into XML-like workspace context for LLM input

simple-workspace-contextualizer

A small CLI for serializing a source repository into a structured, LLM-friendly context format.

It prints:

• a filtered file tree for the workspace
• the full contents of selected files
• stable XML-like wrapper tags so the output can be pasted directly into an LLM prompt

The intended use case is: "take a repo and turn it into a single prompt payload that gives an LLM useful project context."

Why this exists

LLMs usually perform better when they can see:

• the repository structure
• the important source files
• toolchain and config files
• README / markdown documentation
• the exact source text, without reformatting or lossy transformations

This tool packages that information into one stream.

Features

• Recursively prints a workspace file tree
• Includes common source files by default
• Includes common toolchain / config files by default
• Supports additional user-supplied glob patterns
• Supports --override to disable the built-in file selection globs
• Hides ignored paths from the file tree
• Allows user-supplied globs to explicitly include ignored files in the emitted file contents
• Produces a simple structured output format that is easy to paste into an LLM

Installation

From PyPI

pip install simple-workspace-contextualizer

Local development

git clone <your-repo-url>
cd simple-workspace-contextualizer
python -m pip install -e .

This installs the CLI command:

swc

Or the full command name:

simple_workspace_contextualizer

Requirements

• Python 3.13+

Usage

swc PROJECT_ROOT [--override] [GLOB ...]

Arguments

ROOT

Root project directory to scan.

--override

Disable the built-in default include globs.

When --override is present, only the user-supplied globs are used for <file_contents>.

GLOB ...

Zero or more additional glob patterns to include in <file_contents>.

These user globs are additive by default, and they can explicitly include files that would otherwise be ignored.

Examples

Serialize a repo with the default built-in file patterns:

swc .

Add a few extra file patterns:

swc . "docs/**/*.txt" "proto/**/*.proto"

Only include files matching your own globs:

swc . --override "src/**/*.py" "tests/**/*.py"

Explicitly include something under an otherwise ignored path:

swc . ".venv/lib/python3.13/site-packages/pip-25.3.dist-info/LICENSE.txt"

Output format

The tool emits XML-like structured text:

<workspace_context>
  <file_tree>
    ...
  </file_tree>
  <file_contents>
    <file path="src/example.py">
      ...
    </file>
    <file path="pyproject.toml">
      ...
    </file>
  </file_contents>
</workspace_context>

Semantics

• <file_tree> contains the filtered directory tree
• <file_contents> contains the full contents of selected files
• each <file> element has a path attribute relative to the workspace root

File selection model

The set of files included in <file_contents> is:

1. the built-in default glob set
2. plus any user-supplied globs

If --override is used, the built-in default glob set becomes empty.

So the effective rule is:

• default mode: default globs ∪ user globs
• override mode: user globs

Ignore behavior

Ignored paths are hidden from the file tree and excluded from files selected by the built-in default glob set.

However, user-supplied globs are allowed to explicitly include ignored files.

That means:

• ignored paths do not appear in <file_tree>
• ignored files do not get included by default
• ignored files can still appear in <file_contents> if the user explicitly requests them with a glob argument

This is intentional: it keeps the default output clean while still allowing targeted overrides.

Default included files

The built-in include globs are aimed at "files that are useful to show an LLM."

They include:

• source files such as Python, JavaScript, TypeScript, Go, Rust, Java, C/C++, shell, SQL, and others
• markdown files
• common toolchain and config files such as:
  • Makefile
  • Dockerfile
  • pyproject.toml
  • requirements.txt
  • package.json
  • tsconfig.json
  • go.mod
  • Cargo.toml
  • pom.xml
  • Gradle files
  • common lint / formatter config files

Typical ignored paths

The ignore list is intended to filter out noisy or generated content such as:

• VCS metadata
• editor / IDE folders
• caches
• virtual environments
• node_modules
• build outputs
• coverage artifacts
• generated packaging metadata

Examples include:

• .git/
• .venv/
• node_modules/
• dist/
• build/
• __pycache__/
• *.egg-info/
• *.dist-info/

Example workflow with an LLM

Generate workspace context:

swc . > workspace_context.xml

Then paste that output into your LLM prompt along with a task, for example:

Here is the current repository context. Please explain the architecture, identify likely entry points, and suggest how to add feature X.

Design goals

• Minimal dependencies
• Easy to inspect
• Easy to pipe to stdout
• Easy to compose with shell tooling
• Optimized for LLM context generation rather than archival fidelity

Non-goals

• Full XML document modeling
• Perfect reproduction of every file in a repository
• Preserving large binary assets
• Replacing dedicated indexing or retrieval systems

Limitations

• Large repositories can produce very large outputs
• The default globs are heuristic, not exhaustive
• Ignored paths are policy-driven and may need adjustment for your codebase
• Prompt quality still depends on what files are included

Development

Run locally:

python -m simple_workspace_contextualizer.cli .

Or, if installed in editable mode:

swc .

Naming

Package name:

simple-workspace-contextualizer

Default CLI:

swc

License

MIT