Dwell is a simple CMS for creating your own
indieweb website, including support
for micropub,
indieauth, and
webmention. Dwell
was created to power the personal site of
cleverdevil, which currently
runs the known content
management system. Under the hood, Dwell is built in the
python programming language, and uses
duckdb for its query engine, though all
content is stored on-disk as JSON files contained within a
hive
partitioned directory structure.
For many years, I have had my own site. In fact, my current website includes a massive archive all of my content starting from 2002! I have even migrated all of the content I posted on Instagram, Twitter, and Facebook before I left those silos entirely after they became cesspools of misinformation and surveillance capitalism owned and operated by billionaires with values that do not align with my own.
My current site runs on known, which has served me well. While I am able to write PHP, it is absolutely not my language of choice or expertise. In addition, I would like to make my site more lightweight, with as few dependencies as possible, and no need for complex relational databases.
Why not use something like WordPress? Yes, WordPress has the benefit of a massive community, but is essentially a giant pile of legacy. I have a lot of experience with that community, and I think its lovely, but isn’t really focused on what I care about, and has all of the same complexity issues as Known, with the downside of being an absolute security nightmare.
Why not a static site generator? I considered it, but I like the ability to create extremely dynamic pages. With DuckDB, I get all of the benefits of a relational database with none of the operational pain.
Lastly, because its fun. I love working on my site!
Content is stored on disk as JSON files in a directory structure that is partitioned based upon date, as follows:
content/
year=2023/
month=11/
day=12/
post-uuid.json
Dwell uses the aforementioned DuckDB to make the entire content
directory queryable with SQL, without the need to run a separate,
heavyweight database server. Very large sites with tens of thousands of
posts can be completely indexed by Dwell in a matter of a few seconds.
The complete content is stored in a table called raw_content within
the content.db database, and a convenience view called content is
created which makes the data easier to interact with.
This is early. Very early. The first focus has been proving out the data storage, indexing, and querying mechanism using DuckDB. Below is a list (that is likely incomplete) of features, including what is done, and what is in-progress.
- Consume and index a hive-partitioned directory of MF2 JSON
- Storage and index support for post kinds
- Blog
- Status
- Bookmark
- Watch
- Play
- Reply
- Photo
- Checkin
- Listen
- Recipe
- Review
- RSVP
- Query for posts
- Get by post id
- Query by date range and post kind
- Ordering, limit, and offset
- Micropub support
- Create content
- MF2 JSON
- Form encoded
- Multipart form
- Update content
- Delete content
- Undelete content
- Media endpoint
- Syndication
- Mastodon
- Bluesky
- Create content
- IndieAuth support
- Token database
- Password configuration
- Micropub endpoint enforcement
- User-facing auth flow
- User-facing auth flow beautification
- Separate IndieAuth code
- Webmention
- Move all post kinds to plugins, other than Blog, Status, and Photo
- Website views
- Home stream
- Custom streams
- Monthly summaries
- On this day
- Photos page
- Static pages
- Profile
- Archive
- Now
- Overview
- Make all streams and post kinds use proper microformats
- Forms for authenticated users to create all “kinds” of content
- Fixing ingestion for specific types and content that is missed
- Recipe
- Like / Repost kind
- Interactions (likes, primarily)
At first, I will entirely focus on getting my own website fully functional on-top of Dwell, to enable me to migrate off of Known, but I am actively seeking collaborators that are interested in a shared codebase for building our own sites.