clatcli 0.1.2

Command line assistance tool. Describe what you want in plain English; clat generates a shell script and runs it.

clat

Command line assistance tool. Describe what you want in plain English; clat generates a shell script and runs it.

clat open a port, docker pull void-base, close port
clat compress all jpegs in this directory to 80% quality
clat show disk usage sorted by size for the current directory

Works with any OpenAI-compatible inference API — LM Studio, Ollama, or a remote API. Supports reasoning models (DeepSeek-R1, QwQ, etc.) — <think> blocks are stripped automatically.

Install

pip install clatcli

Configuration

Config lives at ~/.clat/config.toml. Create it with:

clat --init

api_url           = "http://localhost:1234/v1"   # OpenAI-compatible endpoint
model             = "local-model"                # model name (see: clat -l)
api_key           = ""                           # optional bearer token
auto_run          = false                        # true = always skip confirmation
use_tools         = true                         # false for models without tool-call support
auto_run_patterns = []                           # command names that skip confirmation

Usage

clat [OPTIONS] <prompt>...

Flag	Description
-y, --yes	Skip confirmation, run immediately
-n, --dry-run	Show generated script, don't execute
-l, --list	List models available from the API
-L, --load <ID>	Load a model in LM Studio
--model <MODEL>	Override model for this invocation
--api <URL>	Override API URL for this invocation
-v, --verbose	Print prompt, API URL, model, and tool status
--config	Show current config and its path
--init	Write default config file

License

MIT