Python Engine & API

Last Updated March 22, 2026

Omni-MDX provides a native Python bridge to the high-performance Rust core. It parses MDX (Markdown + JSX) into a deeply manipulable Abstract Syntax Tree (AST) consisting of pure Python objects.

Because the Rust binary is bundled directly into the Python package as a “Fat Wheel”, no Rust toolchain is required for end-users to install and run the engine.

Installation

pip install omni-mdx

1. Parsing MDX to AST

The core feature of omni-mdx in Python is transforming raw text into a structured, easily searchable AST.

You can initialize the OmniMDX engine and call the parse_to_ast method:

from omni_mdx.engine import OmniMDX

mdx_content = """
# System Configuration
The core threshold is defined as:
$$T_c = \\mu \\times \\nabla$$

Calibration required!
"""

# Initialize the engine
engine = OmniMDX()

# Parse the text into a list of AstNode objects
nodes = engine.parse_to_ast(mdx_content)

2. Understanding the AST Structure

The parser generates a typed AstNode tree. Each node in the MDX AST is represented by an AstNode dataclass, which contains:

• node_type: A string representing the tag name (e.g., "p", "h1", "text", "BlockMath", or custom components like "Alert").
• content: Raw text content (for leaf text nodes).
• attributes: A dictionary mapping attribute names to AttrValue instances.
• children: A list of nested AstNode child nodes.
• self_closing: A boolean indicating if the tag was self-closing (<Comp />).

The AttrValue Object

JSX attributes can be complex. The AttrValue class neatly categorizes them by kind:

• text: Standard string values (prop="hello").
• expression: Raw JavaScript/Python expressions (prop={someExpr}).
• boolean: Implicitly true attributes (disabled).
• ast: Sub-trees passed as props (prop={<Component/>}).

3. Advanced AST Manipulation

The Python AST objects come with built-in helper methods to make traversing and extracting data incredibly easy for large-scale text analysis or metadata ingestion pipelines.

Finding Nodes
Use find(node_type) to get the first matching descendant using BFS, or find_all(node_type) to retrieve all matches depth-first.

Extracting Text
The text_content() method recursively collects all text content inside a node as a single string, stripping away all MDX tags. The attr_text(name) method retrieves the plain-text value of a specific attribute.

Example: Extracting Component Data

If you need to programmatically extract specific nodes while ignoring the rest of the document formatting, you can easily iterate over the AST:

from omni_mdx.engine import OmniMDX
import json

document = """
# System Logs


Database backup completed successfully.



High memory usage detected.

"""

engine = OmniMDX()
ast = engine.parse_to_ast(document)

# Extract logs for structured monitoring
structured_logs = []
for node in ast:
    if node.node_type == "LogEntry":
        structured_logs.append({
            "user": node.attr_text("user"),
            "status": node.attr_text("status"),
            "message": node.text_content().strip()
        })

print(json.dumps(structured_logs, indent=2))