obsilink

obsilink is a small, deterministic Python library for extracting and replacing links in Markdown text.

It supports:

• Obsidian wikilinks like [[Note]], [[Note|Alias]], and ![[Embedded]]
• Markdown links like [Docs](https://example.com) and embeds like ![Logo](logo.png)
• Plain URLs like https://example.com, ftp://files.example.com, and mailto:user@example.com

The parser returns structured Link objects in encounter order and preserves duplicates. Links can also be replaced in-place using structured (old, new) pairs.

Features

• Deterministic parsing with stable ordering
• Two public functions: extract_links(...) and replace_links(...)
• Accepts either raw str or file-like input with .read()
• Preserves duplicates instead of deduplicating
• Strict, typed API compatible with mypy strict mode
• Links are split into structured fields (target, alias, heading, blockid) — heading (#) and block-id (^) fragments are extracted separately and never included in target

Installation

With uv (recommended)

uv add obsilink

With pip

pip install obsilink

From source

git clone https://github.com/chgroeling/obsilink.git
cd obsilink
uv sync

If the package is not yet published to your package index, use the source install method.

Quick start

from obsilink import extract_links

text = """
See [[Project/Plan#Milestones|Roadmap]],
![Logo](assets/logo.png),
and https://example.com.
"""

links = extract_links(text)

for link in links:
    print(link.type.value, link.target, link.alias)

Replacing links

from obsilink import Link, LinkType, replace_links

text = "See [[OldNote]] for details."

old = Link(type=LinkType.WIKILINK, target="OldNote", alias=None, heading=None, blockid=None)
new = Link(type=LinkType.WIKILINK, target="NewNote", alias=None, heading=None, blockid=None)

result_text, applied = replace_links(text, [(old, new)])
print(result_text)   # See [[NewNote]] for details.
print(applied)       # [True]

More complex Obsidian wikilink example

from obsilink import extract_links

text = """
[[Project/Plan#Milestones|Roadmap]]
![[Assets/Diagrams/System Overview#v2]]
[[People/Ada Lovelace#Biography^early-life|Ada]]
[[Project/Plan#Milestones|Roadmap]]
[[Malformed
"""

links = extract_links(text)

for link in links:
    print(
        link.type.value,
        link.target,
        link.alias,
        link.heading,
        link.blockid,
    )

# Encounter order is preserved, duplicates are kept, and malformed
# wikilinks (like "[[Malformed") are ignored.

API

extract_links(source)

Extracts links from:

• str
• text file-like object with .read() returning str

Returns:

• list[Link] in encounter order

Raises:

• TypeError for unsupported source types
• TypeError if .read() returns non-string data
• Any exception raised by .read() is propagated

replace_links(source, replacements)

Replaces links in text based on a list of (old_link, new_link) pairs.

Accepts:

• source: str or text file-like object with .read() returning str
• replacements: list[tuple[Link, Link]]

Returns:

• tuple[str, list[bool]] — the modified text and a boolean per pair indicating whether the replacement was applied

Replacements are applied sequentially; each pair replaces only the first occurrence.

Raises:

• TypeError for unsupported source types
• TypeError if .read() returns non-string data
• Any exception raised by .read() is propagated

Link

Link is a frozen dataclass with these fields:

• type: LinkType
• target: str
• alias: str | None
• heading: str | None
• blockid: str | None

Convenience properties:

• link.is_url -> True for URL targets (http, https, ftp, file, mailto)
• link.as_path -> pathlib.Path for non-URL targets (raises ValueError for URLs)

Supported link types

• LinkType.WIKILINK
• LinkType.WIKILINK_EMBED
• LinkType.MARKDOWN_LINK
• LinkType.MARKDOWN_EMBED
• LinkType.PLAIN_URL

How links are split

Each link is split into its constituent parts. The target field contains only the note, path, or URL — heading (#) and block-id (^) fragments are extracted into their own fields and are not included in target.

Obsidian wikilinks

Markdown links

For Markdown links, heading and blockid are always None. The target is the URL or path inside parentheses, and alias is the link text inside brackets.

Development

This project uses uv, ruff, mypy, and pytest.

uv sync
uv run ruff format src/ tests/
uv run ruff check src/ tests/
uv run mypy src/
uv run pytest

License

This project is licensed under the MIT License. See LICENSE.