CLI Reference
Commands
axm-smelt compact
Compact input and print the result.
| Flag | Default | Description |
|---|---|---|
--file PATH |
stdin | Read from file instead of stdin |
--strategies LIST |
— | Comma-separated strategy names |
--preset NAME |
safe |
Named preset (safe, moderate, aggressive) |
--output PATH |
stdout | Write compacted text to file |
Compacted text goes to stdout (or --output). Savings summary goes to stderr.
Exits with code 1 on unknown preset, unknown strategy, or missing file.
axm-smelt check
Analyze input without transforming it. Shows per-strategy savings estimates (only strategies with positive savings are listed).
| Flag | Default | Description |
|---|---|---|
--file PATH |
stdin | Read from file instead of stdin |
axm-smelt count
Count tokens in input.
| Flag | Default | Description |
|---|---|---|
--file PATH |
stdin | Read from file instead of stdin |
--model MODEL |
o200k_base |
tiktoken encoding name (e.g. o200k_base, cl100k_base) or OpenAI model name (e.g. gpt-4o, gpt-4) |
A Claude model name (e.g. claude-opus-4-8) or an otherwise unknown name routes to the o200k_base proxy encoding (an approximation, no network call) rather than failing — never a len // 4 heuristic. For an exact Claude count, read usage.input_tokens from the run instead.
axm-smelt version
Print the version string.
Python API
Auto-generated API reference is available under Python API.