axm-smelt 0.3.0

axm-smelt — Deterministic token compaction for LLM inputs

axm-smelt reduces token consumption for LLM inputs by applying deterministic compaction strategies — whitespace collapsing, structural transforms, and optional lossy simplifications. It works through the Python API and three registered AXMTools, which provide MCP, AXM CLI, and DAG-node access from one declaration.

📖 Full documentation

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)
• 10 strategies — minify, drop_nulls, flatten, tabular, round_numbers, strip_quotes, dedup_values_with_refs, collapse_whitespace, compact_tables, strip_html_comments
• Composable pipeline — chain strategies explicitly or use presets (safe, moderate, aggressive)
• AXMTools — smelt, smelt_check, and smelt_count, available through MCP, axm, and DAG nodes
• MCP tool — SmeltTool for use by AI agents via axm-mcp
• Modern Python — 3.12+ with strict typing

Installation

uv add axm-smelt

Quick Start

AXM CLI

The CLI surface is derived from the registered AXMTools. Each command accepts redirected text or a UTF-8 file.

Compact data with smelt:

printf '{"name": "Alice", "notes": null}\n' | axm smelt
# smelt | json | 11->9 tok (-18.18%) | minify | tiktoken
# {"name":"Alice","notes":null}
axm smelt --input-path ./payload.json

Analyze possible savings with smelt_check:

printf '{"name": "Alice", "notes": null}\n' | axm smelt_check
# smelt_check | json | 11 tok
#   drop_nulls: -54.55%
#   minify: -18.18%
#   flatten: -18.18%
#   round_numbers: -18.18%
#   strip_quotes: -18.18%
axm smelt_check --input-path ./payload.json

Count tokens with smelt_count:

printf 'alpha beta gamma delta epsilon\n' | axm smelt_count
# smelt_count | 6 tokens | 31 chars | o200k_base | tiktoken
axm smelt_count --input-path ./payload.txt

Explicit data takes precedence over --input-path, which takes precedence over non-interactive stdin. If the designated path does not exist or its contents are not valid UTF-8, the command exits with a non-zero status and a diagnostic that names that path.

There is no standalone axm-smelt executable and python -m axm_smelt is intentionally unsupported.

Python API

from axm_smelt import smelt, check, count

# Compact using the safe preset (default)
report = smelt('{\n  "name": "Alice",\n  "age": 30\n}')
print(f"{report.savings_pct:.1f}% saved")
# Tokens: 14 -> 9

# Compact with explicit strategies
report = smelt(data, strategies=["minify", "drop_nulls"])

# Compact with a preset
report = smelt(data, preset="aggressive")

# Analyze without transforming
report = check('{"data": [1, 2, 3]}')
for strat, pct in report.strategy_estimates.items():
    print(f"  {strat}: {pct:.1f}%")

# Count tokens
tokens = count("hello world")

MCP (AI Agent)

axm-smelt is available through axm-mcp. AI agents can call smelt(data, preset="moderate"), smelt_check(...), or smelt_count(...) directly.

See the MCP how-to guide for details.

AXMTool Commands

These commands come from the axm.tools registry; the same definitions power MCP and DAG nodes. Use --help for their generated CLI signatures. The removed standalone façade has no compatibility alias.

Strategies

Presets

Development

This package is part of the axm-forge workspace.

git clone https://github.com/axm-protocols/axm-forge.git
cd axm-forge
uv sync --all-groups
uv run --package axm-smelt --directory packages/axm-smelt pytest -x -q

License

Apache-2.0 — © 2026 axm-protocols