A tool for providing intelligent codebase context to Large Language Models (LLMs).
Built on files-to-prompt.
READMEs are pulled out into a separate section of the output at the top of the context.
If there is a README in a containing directory for any paths provided, it will be included.
For example, if you provide codebase/env/data, codebase/env/README.md will be included.
- All paths are relative to $CODE_CONTEXT_ROOT (default: ~/src)
- The first directory in any path is considered the codebase name
- We handle cases where codebase names are repeated:
- Direct access:
codebase/env->$CODE_CONTEXT_ROOT/codebase/env - Auto-prefixed:
codebase/env->$CODE_CONTEXT_ROOT/codebase/codebase/env(if direct path doesn't exist)
- Direct access:
- Special cases:
- Tests directory is always at
codebase/tests - Root files (like README.md) are accessed directly
- Tests directory is always at
from root:
pip install -e .Copy manabot codebase to clipboard (OS X only):
code-context -p manabotCopy multiple codebases and directories:
code-context manabot,managym/agent,managym/testsFilter to python and use raw (non-XML) output
code-context -e .py -r manabot - Projects live in $CODE_CONTEXT_ROOT (default: ~/src)
- Automatically traverse repeated directory names (e.g.
manabot/manabot/env->manabot/env) - READMEs at any level are automatically included