Module llm 

DX Serializer LLM Format Module

This module provides support for multiple interconvertible formats:

DX Serializer Format (Token-Efficient LLM Format)

A token-efficient serialization format optimized for LLMs:

• key=value for simple key-value pairs
• name(key=val key2=val2) for objects (parentheses, space-separated)
• name[col1 col2 col3](rows) for tables (wrapped dataframe format)
• key=[item1 item2 item3] for arrays (square brackets, space-separated)
• true/false for booleans, null for null values
• Use quotes "..." for multi-word strings

Human Format (Clean TOML-like)

Clean, hand-editable format:

• = for single scalar values
• : + - lines for arrays
• # for comments
• Nested sections like [stack.js]

Machine Format

Binary format for runtime using RKYV’s zero-copy architecture.

The architecture follows a “hub and spoke” model where all formats convert through a common internal representation (DxDocument), ensuring consistent round-trip behavior.

Re-exports

pub use abbrev::AbbrevDict;

pub use cache_generator::CacheConfig;

pub use cache_generator::CacheError;

pub use cache_generator::CacheGenerator;

pub use cache_generator::CachePaths;

pub use cache_generator::CacheResult;

pub use convert::ConvertError;

pub use convert::MachineFormat;

pub use convert::document_to_human;

pub use convert::document_to_llm;

pub use convert::document_to_machine;

pub use convert::human_to_document;

pub use convert::human_to_llm;

pub use convert::human_to_machine;

pub use convert::human_to_machine_uncompressed;

pub use convert::is_llm_format;

pub use convert::llm_to_document;

pub use convert::llm_to_human;

pub use convert::llm_to_machine;

pub use convert::machine_bytes_to_document;

pub use convert::machine_to_document;

pub use convert::machine_to_human;

pub use convert::machine_to_llm;

pub use convert::try_document_to_machine_with_compression;

pub use human_formatter::HumanFormatConfig;

pub use human_formatter::HumanFormatter;

pub use human_parser::HumanParseError;

pub use human_parser::HumanParser;

pub use machine_zerocopy::ZeroCopyDocument;

pub use machine_zerocopy::ZeroCopyError;

pub use machine_zerocopy::ZeroCopyMachine;

pub use parser::LlmParser;

pub use parser::ParseError;

pub use pretty_printer::PrettyPrintError;

pub use pretty_printer::PrettyPrinter;

pub use pretty_printer::PrettyPrinterConfig;

pub use section_names::SectionNameDict;

pub use serializer::LlmSerializer;

pub use serializer::SerializerConfig;

pub use serializer_output::SerializerOutput;

pub use serializer_output::SerializerOutputConfig;

pub use serializer_output::SerializerOutputError;

pub use serializer_output::SerializerPaths;

pub use serializer_output::SerializerResult;

pub use table_wrapper::TableWrapper;

pub use table_wrapper::TableWrapperConfig;

pub use tokens::ModelType;

pub use tokens::TokenCounter;

pub use tokens::TokenInfo;

pub use types::DxDocument;

pub use types::DxLlmValue;

pub use types::DxSection;

Modules

abbrev

Key abbreviation dictionary for LLM and Human format conversion

cache_generator

Cache Generator for DX Serializer

convert

Format conversion functions

human_formatter

Human Format Formatter

human_parser

Human format parser

machine_zerocopy

Zero-copy machine format implementation

parser

DX Serializer LLM Format Parser

pretty_printer

Pretty Printer for Human Format

section_names

Section name dictionary for Human V3 format conversion

serializer

DX Serializer LLM Format

serializer_output

Serializer Output Module for DX Serializer

table_wrapper

Table Wrapper for Wide Tables

tokens

Token counting for LLM serializer output

types

Core data types for LLM and Human format serialization