Module extract 

Shared extraction strategies for LLM output parsing.

This is the load-bearing module — every parser calls into these functions for preprocessing, code block extraction, and bracket matching.

Functions

extract_code_block

Extract content from the first matching markdown code block.

extract_code_block_for

Extract content from a code block matching a specific language.

find_bracketed

Find a bracketed substring by matching open/close delimiters.

preprocess

Full preprocessing pipeline applied to every LLM response.

strip_think_tags

Strip all <think>...</think> and <thinking>...</thinking> blocks from text.