axm-smelt

Deterministic token compaction for LLM inputs.

---

What it does

axm-smelt reduces token consumption for LLM inputs by applying deterministic compaction strategies. It detects the input format, runs the selected strategies in pipeline order, and reports exact token savings via tiktoken.

Strategy	Category	Effect
minify	whitespace	Remove whitespace from JSON, YAML, and XML
drop_nulls	structural	Remove None, "", [], {} values
flatten	structural	Collapse single-child wrapper dicts
tabular	structural	Convert list[dict] to pipe-separated tables
dedup_values_with_refs	structural	Replace repeated long strings with aliases
round_numbers	cosmetic	Round floats to N decimal places
strip_quotes	cosmetic	Remove quotes on simple JSON keys (JSON only)
collapse_whitespace	whitespace	Collapse blank lines / trailing spaces in prose (skips code fences)
compact_tables	whitespace	Strip padding from Markdown table cells
strip_html_comments	cosmetic	Remove <!-- ... --> from prose / Markdown

Quick Example

Bash

# CLI
echo '{"name": "Alice", "age": 30}' | axm-smelt compact

# Or use a preset
axm-smelt compact --file data.json --preset aggressive

Python

# Python API
from axm_smelt import smelt, check, count

report = smelt('{\n  "name": "Alice",\n  "age": 30\n}')
print(f"{report.savings_pct:.1f}% saved")
# 35.7% saved

Features

• Format detection — auto-detect JSON, YAML, XML, TOML, CSV, Markdown, and plain text
• Token counting — always via tiktoken; Claude and unknown models route to the o200k_base proxy (approximate, no network)
• Composable pipeline — chain strategies or use presets (safe, moderate, aggressive)
• CLI — axm-smelt compact|check|count|version commands
• MCP tool — available to AI agents via axm-mcp
• Modern Python — 3.12+ with strict typing

Learn More

• Getting Started Tutorial
• Compact Data
• Use Strategies
• Use Presets
• Analyze Token Waste
• Use via MCP
• Strategy Catalog
• Format Detection