CLI Reference
The Python package includes the llm-forge command-line utility for shell scripting and quick validations without writing code.
Note: The CLI is currently only available via the Python package, not npm.
Usage
llm-forge [command] [options]
Commands
count
Counts exact tokens for a given string or file.
# Count direct string
llm-forge count gpt-4o "This is a test prompt."
# Count from file
llm-forge count claude-3-5-sonnet --file ./prompt.txt
chunk
Splits a document and outputs the chunks, either to stdout or a directory.
llm-forge chunk gpt-4o \
--file ./long-document.md \
--strategy sentence \
--max-tokens 1000 \
--overlap 100 \
--out-dir ./chunks
Supported strategies: sentence, paragraph, semantic, code, fixed
estimate
Calculates the API cost for processing a document.
llm-forge estimate gpt-4o --file ./data.txt
models
Lists all available models in the registry with their configurations.
llm-forge models
# Filter by provider
llm-forge models --provider openai
Environment Flags
You can force color output in CI/CD environments by setting FORCE_COLOR=1.