---

2026-03-24: SHORT TERM HELPER FOR LITELLM ISSUE (sh script not python)

inventory_litellm.sh is a temporary quick script for a first-pass local check related to the LiteLLM package vulerability discussed at Hacker News. It is intended only to help you quickly gauge whether that issue might not be impacting you. Its LiteLLM version inventory is a quick first pass and should not be considered definitive, authoritative, or trusted for security decisions. v1.82.7, v1.82.8 are known bad.

---

uvpowered-tools

Minimal command-line tools implemented as executable uv-powered Python scripts. Each script is a standalone Python file that uses the #!/usr/bin/env -S uv … shebang to spin up a cached environment with the correct package versions. No building, compiling, or installing packages into your system.

• Zero overhead to run (no virtualenv setup, no pip installs)

• Fully standalone and directly executable

• Easy to modify as small, readable Python scripts

List of uv powered tools (may not be complete)

LLM token counter: count-tokens.py

Loads a HuggingFace tokenizer for an LLM model and tokenizes a given file. Outputs a compact summary: token count, compression ratio, bytes per token, vocabulary size, and special-token information.

Details

❯ ./count-tokens.py --file count-tokens.py --model unsloth/Qwen3-VL-30B-A3B-Instruct
Counting tokens in count-tokens.py using the tokenizer from Huggingface model unsloth/Qwen3-VL-30B-A3B-Instruct
Tokenizer class: Qwen2TokenizerFast
Vocabulary size: 152k possible tokens
Special tokens (3): ['eos_token', 'pad_token', 'additional_special_tokens']

Input characters: 2546
Compression ratio (tokens/chars): 27.4%
UTF-8 bytes: 2546
Bytes per token: 3.6

Token count: 698

❯ ./count-tokens.py --file count-tokens.py --model unsloth/gpt-oss-120b
Counting tokens in count-tokens.py using the tokenizer from Huggingface model unsloth/gpt-oss-120b
Tokenizer class: PreTrainedTokenizerFast
Vocabulary size: 200k possible tokens
Special tokens (3): ['bos_token', 'eos_token', 'pad_token']

Input characters: 2546
Compression ratio (tokens/chars): 26.4%
UTF-8 bytes: 2546
Bytes per token: 3.8

Token count: 671

Example output for gpt-oss-120b, Qwen3-VL-30B-A3B-Instruct

Extensible minimal example: GENERIC-SCRIPT-TEMPLATE.py

A minimal script template showing how to build new uv-powered tools. Useful as a starting point for extending or adding your own utilities.

Installation

Clone this repo

git clone https://github.com/kinchahoy/uvpowered-tools
cd uvpowered-tools

Make sure you have uv installed

echo "Examine the uv install script"
curl -LsSf https://astral.sh/uv/install.sh | less
echo "Run the uv install script Reference at https://docs.astral.sh/uv/getting-started/installation/
curl -LsSf https://astral.sh/uv/install.sh | sh

Run the script (chmod +x if they are not executable)

./count-tokens.py

or

./GENERIC-SCRIPT-TEMPLATE.py

or

./inventory_litellm.sh

Notes

• Most scripts run quickly unless they require uv to do a bunch of dependency checking (e.g. if they use import 'transformers').
• If you want them available system-wide, you can copy the script into /usr/local/bin and make it executable.