Patitas

Python Markdown Parser for Modern Content

CommonMark. Typed AST. Secure parsing.

Patitas is a pure-Python Markdown parser for HTML rendering, content pipelines, and Markdown tooling. It uses a state-machine lexer with O(n) guaranteed parsing, exposes a typed AST, and supports directives, frontmatter, notebook parsing, and LLM-safe output.

from patitas import Markdown

md = Markdown()
html = md("# Hello **World**")
# Output: <h1>Hello <strong>World</strong></h1>

Why Use Patitas

Common Use Cases

• Parsing Markdown to HTML in Python applications
• Extracting typed structure from Markdown for analysis or transforms
• Handling YAML frontmatter and notebook content in publishing pipelines
• Replacing regex-heavy Markdown parsers in security-sensitive systems
• Powering docs stacks with directives, syntax highlighting, and incremental parsing

Performance

• 652 CommonMark examples — ~26ms single-threaded
• Incremental parsing — For a 1-char edit in a ~100KB doc,parse_incrementalis ~200x faster than full re-parse (~160µs vs ~32ms)
• Parallel scaling — ~2.5x speedup with 4 threads under Python 3.14t free-threading

Patitas prioritizes safety over raw speed: O(n) guaranteed parsing, typed AST, and full thread-safety.

Typed AST

Every node is a frozen dataclass with full type information:

from patitas import parse

doc = parse("# Hello **World**")
heading = doc[0]

# Type-safe access
print(heading.level)      # 1
print(heading.children)   # [Text, Strong]

# IDE autocompletion works
heading.level  # int
heading.children[1].children  # Sequence[Inline]

The Bengal Ecosystem

A structured reactive stack — every layer written in pure Python for 3.14t free-threading.

Python-native. Free-threading ready. No npm required.