ENG-2217 Rewrite asset links in Obsidian-origin markdown for Roam

[maparent]

Reviewer brief

Given a markdown and assets, rewrite the links to use the firebase links.
We have to look up the file type to make the right kind of Roam link; since Obsidian does not have distinct link shapes, we have to infer from mimetype or file extension.

Verification

Unit tests. Real tests in the next PR.

Loom video

https://www.loom.com/share/ac85dc0a4df9411c84858e965dadb86f

Note: I did push a refactor from 2219 to 2217 after the loom, but no behaviour change.
I also renamed Token, which was ambiguous, to Locator.

Scope check

• Ran $scope-check against ENG-2217 and the final diff.
• Scope beyond Done When: Adds mime-db (+ @types/mime-db) as an apps/roam dependency.
• Required now: Done When requires correct Roam syntax per media type, and the type set is not enumerable by hand: the extension table is derived from all ~1000 mime-db types (223 extensions) so an imported asset renders as the same file would if uploaded natively. Cost: db.json is not tree-shakeable, so ~204 KB lands in extension.js (23KB compressed); a build-time codegen of the table would remove it later.
• Anyone affected or consulted: No.
• Decision: Rationale recorded in the EXTENSION_KINDS docstring; no Linear decision.
  • I thought that maintaining mimetype information ourselves was inviting drift, and wanted to rely on a library.
  • Obsidian relies on mime-types already, so there is symmetry. mime-types imports path, which would not work in a browser build.

Local delegated full review

• Ran a comprehensive review of the entire final diff in a subagent with a fresh context. Use $dg-delegated-full-review when no other full-review workflow is available.

https://linear.app/discourse-graphs/issue/ENG-2217/rewrite-asset-links-in-obsidian-origin-markdown-for-roam

[linear-code]

ENG-2217

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
discourse-graph	Skipped		Sep 10, 2026 10:06pm UTC

[supabase]

This pull request has been ignored for the connected project zytfjzqyijgagqxrzbmz because there are no changes detected in packages/database/supabase directory. You can change this behaviour in Project Integrations Settings ↗︎.

---

Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.

[graphite-app]

PR size/scope check

This PR is over our review-size guideline.

• Recommended: ~200 lines changed
• Acceptable limit: up to 400 lines when well-scoped/self-contained
• Preferred file count: fewer than 5 files

Please split this into smaller PRs unless there is a clear reason the changes need to land together.

If keeping it as one PR, please add a brief justification covering:

• What single problem this PR solves
• Why the files/changes are coupled

[devin-ai-integration]

Note

This report is out of date. Scroll down for Devin Review's latest report on this PR.

Devin Review found 4 potential issues.

2 flags not posted on this PR by your GitHub settings — view them in Devin Review. (Configure)

[devin-ai-integration]

🟡 Code examples become live assets

When markdown demonstrates a recorded asset inside code, replace rewrites that example too. The imported page no longer preserves its source text.

Learn more

Make rewriteAssetLinks syntax-aware so replacements occur only in markdown link and embed nodes, not inside inline code, fenced code blocks, or other literal markdown regions. A markdown parser is preferable to expanding the global regular expression. Add cases where the same sourceRef appears once as an actual embed and once inside inline and fenced code.

---

Was this helpful? React with 👍 or 👎 to provide feedback.

[maparent]

This is accurate, but solving it expands the scope considerably. Proposing to punt.
Written as an issue candidate in Roam.

[maparent]

Added two commits, to handle two more cases.
First two angle bracket forms:
// bracketed destination
https://storage.test/k.png // autolink
Second, make sure to capture labels of non-images:
![[vault/i.docx|a label]] // the new alias group

Those all fall into the original scope of "Find asset references in the markdown", and were just missed in first implementation.

[devin-ai-integration]

Devin Review found 2 new potential issues.

2 flags not posted on this PR by your GitHub settings — view them in Devin Review. (Configure)

[devin-ai-integration]

🟡 Unaliased wikilinks become media embeds

parseMatch gives [[image.png]] no link text, so render treats the link like an embed. The imported page embeds media instead of preserving the source link.

Learn more

An Obsidian wikilink without an alias still has visible link text derived from its target. The parser currently represents that text as empty. The early link-preservation branch then does not run, and the resolved extension selects an image, PDF, audio, or video embed.

Example: [[attachments/diagram.png]] is a normal link to an image file. After resolution, it becomes ![](firebase-url) rather than a link such as diagram.png.

Recommended fix: Distinguish unaliased wikilinks from genuinely empty Markdown links. Supply their implicit display text to render, or add an explicit wikilink form whose media rendering always remains a link.

---

Was this helpful? React with 👍 or 👎 to provide feedback.

[devin-ai-integration]

🟡 Code examples silently change

rewriteAssetLinks applies LINK_PATTERN inside fenced and inline code when a matching asset exists. Literal examples then contain destination URLs instead of the published text.

Learn more

Markdown code spans and fenced code blocks treat link syntax as literal text. A regex over the complete document cannot distinguish those regions from rendered links. If the same locator appears as a real asset elsewhere, the map contains a matching row and the literal occurrence changes too.

Example: A note embeds ![[diagram.png]] and later documents that syntax inside a fenced block. Both occurrences become Firebase syntax, although the fenced example must remain ![[diagram.png]].

Recommended fix: Parse the markdown into syntax nodes and rewrite only link, image, and supported embed nodes. If the existing regex remains, first segment the document with a Markdown-aware tokenizer and exclude code spans and fenced code blocks from replacement.

---

Was this helpful? React with 👍 or 👎 to provide feedback.