extract_symbols() broken with tree-sitter-language-pack>=1.0

[chang-andrew]

Summary

Fresh install of cased-kit==3.5.1 declares tree-sitter-language-pack>=0.7.2 with no upper bound. As of tree-sitter-language-pack==1.0.0, get_parser() was changed to return a bundled native (Rust) parser instead of a tree_sitter.Parser. Its node objects are incompatible with the tree_sitter.QueryCursor API that TreeSitterSymbolExtractor.extract_symbols() uses, so symbol extraction fails on a fresh install (which resolves to the latest, 1.10.x).

Environment

• cased-kit==3.5.1
• tree-sitter-language-pack==1.10.1
• tree-sitter==0.25.2
• Python 3.14, macOS

Reproduction

from kit import Repository
repo = Repository("./")
repo.extract_symbols()

(same results using CLI)

Observed

First, every file errors out of extract_symbols():

[EXTRACT] Error parsing or processing file with ext .js: 'bytes' object is not an instance of 'str'

Traceback (most recent call last):
  File ".../kit/tree_sitter_symbol_extractor.py", line 371, in extract_symbols
    tree = parser.parse(bytes(source_code, "utf8"))
TypeError: 'bytes' object is not an instance of 'str'

The 1.x native parser expects str, not bytes. I tried changing that line to pass str, but extraction then fails at the query stage:

[EXTRACT] No compatible tree-sitter API found for extension .js

…and extract_symbols() silently returns []. The 1.x parser's Tree/Node come from the bundled native module (builtins-level types, root_node is a method), whereas get_language()/get_query() and tree_sitter.QueryCursor come from the tree-sitter PyPI package. The two ABIs don't interoperate, so QueryCursor.matches(root) raises TypeError: ... must be tree_sitter.Node.

Workaround

Pin below the breaking change:

"tree-sitter-language-pack<1.0" # resolves to 0.13.0

With 0.13.0, extract_symbols() works unmodified (verified: 2089 symbols extracted from my project).