fp-conv-cli

Cross-platform command-line tool for floating-point conversions. Encode, decode, and convert numbers across FP32, FP16, BF16, TF32, FP8, FP6, FP4, and integer formats, with binary/hex breakdowns and precision-loss analysis.

The installed command is fp-conv. It reuses the same JavaScript conversion engine that powers the fp-conv web app and the fp-conv MCP server, so results are identical across all three.

Runs anywhere Node.js 18+ runs: Windows, macOS, and Linux.

Installation

npm install -g fp-conv-cli

Or run without installing:

npx fp-conv-cli encode 3.14 --format fp32

Usage

fp-conv <command> [options]

Commands

| Command | Description | | ----------------- | ------------------------------------------------------ | | encode <value> | Encode a decimal/keyword value into a format. | | decode <bits> | Decode a binary or hex bit-pattern in a format. | | convert <value> | Convert a value from one format to another. | | info <format> | Show range, bias, and special-value details. | | list | List all available preset formats. |

Options

| Option | Description | | ---------------------- | --------------------------------------------------------------------------- | | -f, --format <fmt> | Format preset key (e.g. fp32, int8) or a JSON object for a custom format. | | --from <fmt> | Source format for convert. | | --to <fmt> | Target format for convert. | | -r, --rounding <m> | tiesToEven (default), tiesToAway, towardZero, towardPositive, towardNegative. | | --json | Emit machine-readable JSON instead of formatted text. | | -h, --help | Show help. | | -v, --version | Show version. |

Values accept a number, hex (e.g. 0xFF), or keyword (infinity, -infinity, nan). Bit-patterns accept a binary string (e.g. 0100000) or hex (e.g. 0x40).

Examples

# Encode a decimal into FP32
fp-conv encode 3.14 --format fp32

# Negative values work directly; use `--` if a value is mistaken for an option
fp-conv encode -1.5 --format fp16
fp-conv encode --format fp16 -- -1.5

# Decode a hex bit-pattern as FP16
fp-conv decode 0x4048 --format fp16

# Convert between formats and see precision loss
fp-conv convert 3.14 --from fp32 --to fp16

# Inspect a format's range and special-value support
fp-conv info bf16

# List every supported format
fp-conv list

# Use a custom floating-point format
fp-conv encode 1.5 --format '{"signBits":1,"exponentBits":8,"mantissaBits":7}'

# Machine-readable output for scripting
fp-conv convert 3.14 --from fp32 --to fp16 --json

Development

This package lives in the fp-conv repository as an npm workspace. The published bundle (dist/index.js) inlines the shared conversion engine (lib/floating-point.js) and tool definitions (src/webmcp.js) from the repo root via esbuild, keeping a single source of truth and no runtime dependencies.

npm install                         # from the repo root
npm run build --workspace fp-conv-cli
npm test  --workspace fp-conv-cli

License

MIT © Spencer Williams