bug: Treesitter language pack is not installed automatically

[Wizdore]

The code is from docs

import tree_sitter_language_pack as tslp

# Parsers download automatically on first use
result = tslp.process(
    "def hello():\n    print('world')\n",
    tslp.ProcessConfig(language="python", structure=True, imports=True),
)

print(f"Language: {result['language']}")
print(f"Functions: {len(result['structure'])}")

after adding the tree-sitter-language-pack package the code should just work and download the treesitter grammar from python. but it doesnt do that. throws the following error instead

Traceback (most recent call last):
  File "/Users/shaon.rahman2/Work/kreuzberg/main.py", line 4, in <module>
    result = tslp.process(
        "def hello():\n    print('world')\n",
        tslp.ProcessConfig(language="python", structure=True, imports=True),
    )
tree_sitter_language_pack.ParseError: Language 'python' not found

[yurivict]

I have the same problem on FreeBSD.
tslp.available_languages() returns an empty list.
Where do I go from here?

Could you please make a new release with the fix?

[yurivict]

@Goldziher

This is not resolved in 1.6.2

tslp.available_languages() still returns an empty list.

Could you please reopen this issue?

[Goldziher]

Thanks for the report @yurivict. The fix for available_languages() returning an empty list is on main (commit e10684f) but hasn't been released yet — it will be in v1.6.3.

The root cause: available_languages(), has_language(), and language_count() weren't registering the download cache directory before querying the registry, so dynamically downloaded parsers were invisible to those functions.

Note: get_language() and process() should auto-download and work correctly on v1.6.2 — the issue was specific to the listing/querying functions.

We'll close this once v1.6.3 is published.