Crate vision_squeezer 

Structs

DimensionResult

OptimizationReport

OptimizeResult

Persistence

ProcessConfig

All tuneable knobs for the pipeline.

ProcessConfigBuilder

ProcessResult

SavingsReport

SqueezerStats

TokenEstimate

TokenSavingsTable

Full token savings report for a before/after dimension pair across all models.

Enums

ImageOp

Atomic image operations for the Sandbox mode.

OutputFormat

Output encoding format.

ProcessMode

VisionModel

Supported vision model families with their patch pricing.

Functions

binarize

calculate_optimal_dimensions

Snap W×H to tile boundaries using default tile size (512).

calculate_optimal_dimensions_with

Snap W×H to tile boundaries using a custom tile size.

crop_padding

Remove solid-color borders using corner sampling + configurable tolerance.

decode_base64_image

detect_ocr_mode

encode_image_base64

encode_to_bytes

Encode image to raw bytes using the configured output format.

estimate_tokens

Estimate LLM vision tokens for an image of given dimensions.

fit_within

Scale dimensions to fit within max_side while preserving aspect ratio.

fit_within_pixels

Scale dimensions to fit within both a max-side limit and a total-pixel limit.

optimal_send_dimensions

Compute the optimal dimensions to send to a given model to minimize tiles.

optimize_image

MCP entry point: base64 in → base64 JPEG out + savings report.

process

Full pipeline: [crop] → tile-snap resize → [OCR binarize]. Pass input_bytes = 0 if unknown (omits file-size from report).

process_with_operations

Execute a sequence of operations on an image.

token_savings_table