Codebase Search

This skill appears to do what it says—build a local ChromaDB-based semantic index of Python code—but there are practical inconsistencies you should be aware of before installing or copying files: - Import/path mismatch: code_index.py uses imports like from prsm.compute.nwtn.corpus.code_chunker import CodeChunker and the SKILL.md shows convenience imports under prsm.compute.nwtn.corpus. The provided files live under scripts/. If you copy the scripts into a project as-is, those imports will likely fail. Either run the scripts from a package layout matching prsm.* or edit imports to use relative/local module names (e.g., import code_chunker). - Dependencies: you must install chromadb (and its runtime dependencies such as onnxruntime/tokenizers) as noted. Verify those packages are acceptable in your environment (they can be heavy and may require native wheels). - Local indexing and privacy: the index is persisted to {repo_root}/.codebase_index and will contain snippets and docstrings from your code. Add that directory to .gitignore (as SKILL.md suggests) and consider whether you want it included in backups or shared storage. - Behavior is local-only: there are no network calls or secret exfiltration in the code, but the indexer will read all .py files not excluded by default. Review the exclude list and adjust if you need to skip sensitive directories. Recommended next steps: test in a disposable/sandbox repo first, fix the import paths or package layout before using in production, and confirm chromadb works in your environment. If the package is to be used by an agent, ensure the agent's runtime has chromadb and the correct module path available.