Monorepo for Tangled
tangled.org
1---
2title: Tangled docs
3author: The Tangled Contributors
4date: 21 Sun, Dec 2025
5abstract: |
6 Tangled is a decentralized code hosting and collaboration
7 platform. Every component of Tangled is open-source and
8 self-hostable. [tangled.org](https://tangled.org) also
9 provides hosting and CI services that are free to use.
10
11 There are several models for decentralized code
12 collaboration platforms, ranging from ActivityPub’s
13 (Forgejo) federated model, to Radicle’s entirely P2P model.
14 Our approach attempts to be the best of both worlds by
15 adopting the AT Protocol—a protocol for building decentralized
16 social applications with a central identity
17
18 Our approach to this is the idea of “knots”. Knots are
19 lightweight, headless servers that enable users to host Git
20 repositories with ease. Knots are designed for either single
21 or multi-tenant use which is perfect for self-hosting on a
22 Raspberry Pi at home, or larger “community” servers. By
23 default, Tangled provides managed knots where you can host
24 your repositories for free.
25
26 The appview at tangled.org acts as a consolidated "view"
27 into the whole network, allowing users to access, clone and
28 contribute to repositories hosted across different knots
29 seamlessly.
30---
31
32# Quick start guide
33
34## Login or sign up
35
36You can [login](https://tangled.org) by using your AT Protocol
37account. If you are unclear on what that means, simply head
38to the [signup](https://tangled.org/signup) page and create
39an account. By doing so, you will be choosing Tangled as
40your account provider (you will be granted a handle of the
41form `user.tngl.sh`).
42
43In the AT Protocol network, users are free to choose their account
44provider (known as a "Personal Data Service", or PDS), and
45login to applications that support AT accounts.
46
47You can think of it as "one account for all of the atmosphere"!
48
49If you already have an AT account (you may have one if you
50signed up to Bluesky, for example), you can login with the
51same handle on Tangled (so just use `user.bsky.social` on
52the login page).
53
54## Add an SSH key
55
56Once you are logged in, you can start creating repositories
57and pushing code. Tangled supports pushing git repositories
58over SSH.
59
60First, you'll need to generate an SSH key if you don't
61already have one:
62
63```bash
64ssh-keygen -t ed25519 -C "foo@bar.com"
65```
66
67When prompted, save the key to the default location
68(`~/.ssh/id_ed25519`) and optionally set a passphrase.
69
70Copy your public key to your clipboard:
71
72```bash
73# on X11
74cat ~/.ssh/id_ed25519.pub | xclip -sel c
75
76# on wayland
77cat ~/.ssh/id_ed25519.pub | wl-copy
78
79# on macos
80cat ~/.ssh/id_ed25519.pub | pbcopy
81```
82
83Now, navigate to 'Settings' -> 'Keys' and hit 'Add Key',
84paste your public key, give it a descriptive name, and hit
85save.
86
87## Create a repository
88
89Once your SSH key is added, create your first repository:
90
911. Hit the green `+` icon on the topbar, and select
92 repository
932. Enter a repository name
943. Add a description
954. Choose a knotserver to host this repository on
965. Hit create
97
98Knots are self-hostable, lightweight Git servers that can
99host your repository. Unlike traditional code forges, your
100code can live on any server. Read the [Knots](TODO) section
101for more.
102
103## Configure SSH
104
105To ensure Git uses the correct SSH key and connects smoothly
106to Tangled, add this configuration to your `~/.ssh/config`
107file:
108
109```
110Host tangled.org
111 Hostname tangled.org
112 User git
113 IdentityFile ~/.ssh/id_ed25519
114 AddressFamily inet
115```
116
117This tells SSH to use your specific key when connecting to
118Tangled and prevents authentication issues if you have
119multiple SSH keys.
120
121Note that this configuration only works for knotservers that
122are hosted by tangled.org. If you use a custom knot, refer
123to the [Knots](TODO) section.
124
125## Push your first repository
126
127Initialize a new Git repository:
128
129```bash
130mkdir my-project
131cd my-project
132
133git init
134echo "# My Project" > README.md
135```
136
137Add some content and push!
138
139```bash
140git add README.md
141git commit -m "Initial commit"
142git remote add origin git@tangled.org:user.tngl.sh/my-project
143git push -u origin main
144```
145
146That's it! Your code is now hosted on Tangled.
147
148## Migrating an existing repository
149
150Moving your repositories from GitHub, GitLab, Bitbucket, or
151any other Git forge to Tangled is straightforward. You'll
152simply change your repository's remote URL. At the moment,
153Tangled does not have any tooling to migrate data such as
154GitHub issues or pull requests.
155
156First, create a new repository on tangled.org as described
157in the [Quick Start Guide](#create-a-repository).
158
159Navigate to your existing local repository:
160
161```bash
162cd /path/to/your/existing/repo
163```
164
165You can inspect your existing Git remote like so:
166
167```bash
168git remote -v
169```
170
171You'll see something like:
172
173```bash
174origin git@github.com:username/my-project.git (fetch)
175origin git@github.com:username/my-project.git (push)
176```
177
178Update the remote URL to point to tangled:
179
180```bash
181git remote set-url origin git@tangled.org:user.tngl.sh/my-project
182```
183
184Verify the change:
185
186```bash
187git remote -v
188```
189
190You should now see:
191
192```bash
193origin git@tangled.org:user.tngl.sh/my-project (fetch)
194origin git@tangled.org:user.tngl.sh/my-project (push)
195```
196
197Push all your branches and tags to Tangled:
198
199```bash
200git push -u origin --all
201git push -u origin --tags
202```
203
204Your repository is now migrated to Tangled! All commit
205history, branches, and tags have been preserved.
206
207## Mirroring a repository to Tangled
208
209If you want to maintain your repository on multiple forges
210simultaneously, for example, keeping your primary repository
211on GitHub while mirroring to Tangled for backup or
212redundancy, you can do so by adding [multiple remotes](https://git-scm.com/docs/git-push#_remotes).
213
214You can configure your local repository to push to both
215Tangled and, say, GitHub. You may already have the following
216setup:
217
218```bash
219$ git remote -v
220origin git@github.com:username/my-project.git (fetch)
221origin git@github.com:username/my-project.git (push)
222```
223
224Now add Tangled as an additional push URL to the same
225remote:
226
227```bash
228git remote set-url --add --push origin git@tangled.org:user.tngl.sh/my-project
229```
230
231You also need to re-add the original URL as a push
232destination (Git will now use the original URL to fetch only):
233
234```bash
235git remote set-url --add --push origin git@github.com:username/my-project.git
236```
237
238Verify your configuration:
239
240```bash
241$ git remote -v
242origin git@github.com:username/my-project.git (fetch)
243origin git@tangled.org:user.tngl.sh/my-project (push)
244origin git@github.com:username/my-project.git (push)
245```
246
247Notice that there's one fetch URL (the primary remote) and
248two push URLs. Now, whenever you push, Git will
249automatically push to both remotes:
250
251```bash
252git push origin main
253```
254
255This single command pushes your `main` branch to both GitHub
256and Tangled simultaneously.
257
258To push all branches and tags:
259
260```bash
261git push origin --all
262git push origin --tags
263```
264
265If you prefer more control over which remote you push to,
266you can maintain separate remotes:
267
268```bash
269git remote add github git@github.com:username/my-project.git
270git remote add tangled git@tangled.org:user.tngl.sh/my-project
271```
272
273Then push to each explicitly:
274
275```bash
276git push github main
277git push tangled main
278```
279
280# Hosting websites on Tangled
281
282You can serve static websites directly from your git repositories on
283Tangled. If you've used GitHub Pages or Codeberg Pages, this should feel
284familiar.
285
286## Overview
287
288Every user gets a sites domain. If you signed up through Tangled's own
289PDS (`tngl.sh`), your sites domain is automatically
290`<your-handle>.tngl.sh` no setup needed. Otherwise, you can claim a
291`<subdomain>.tngl.io` domain from your settings.
292
293You can serve multiple sites per domain:
294
295- One **index site** served at the root of your domain (e.g.
296 `alice.tngl.sh`)
297- Any number of **sub-path sites** served under the repository name
298 (e.g. `alice.tngl.sh/my-project`)
299
300## Claiming a domain
301
302If you don't have a `tngl.sh` handle, you need to claim a domain before
303publishing sites:
304
3051. Go to **Settings → Sites**
3062. Enter a subdomain (e.g. `alice` to claim `alice.tngl.io`)
3073. Click **claim**
308
309You can only hold one domain at a time. Releasing a domain puts it in a
31030-day cooldown before anyone else can claim it.
311
312## Configuring a site for a repository
313
3141. Navigate to your repository
3152. Go to **Settings → Sites**
3163. Choose a **branch** to deploy from
3174. Set the **deploy directory** — the path within the repository
318 containing your `index.html`. Use `/` for the root, or a subdirectory
319 like `/docs` or `/public`
3205. Choose the **site type**:
321 - **Index site** — served at the root of your domain (e.g.
322 `alice.tngl.sh`)
323 - **Sub-path site** — served under the repository name (e.g.
324 `alice.tngl.sh/my-project`)
3256. Click **save**
326
327The site will be deployed automatically. You can see the status of your
328previous deploys in the **Recent Deploys** section at the bottom of the
329page.
330
331Sites are redeployed automatically on every push to the configured
332branch.
333
334## Custom domains
335
336Tangled currently doesn't support custom domains for sites. This will be
337added in a future update.
338
339## Deploy directory
340
341The deploy directory is the path within your repository that Tangled
342serves as the site root. It must contain an `index.html`.
343
344| Deploy directory | Result |
345|---|---|
346| `/` | Serves the repository root |
347| `/docs` | Serves the `docs/` subdirectory |
348| `/public` | Serves the `public/` subdirectory |
349
350Directories are served with automatic `index.html` resolution -- a
351request to `/about` will serve `/about/index.html` if it exists.
352
353## Site types
354
355| Type | URL |
356|---|---|
357| Index site | `alice.tngl.sh` |
358| Sub-path site | `alice.tngl.sh/my-project` |
359
360Only one repository can be the index site for a given domain at a time.
361If another repository already holds the index site, you will see a
362notice in the settings and only the sub-path option will be available.
363
364## Deploy triggers
365
366A deployment is triggered automatically when:
367
368- You push to the configured branch
369- You change the site configuration (branch, deploy directory, or site
370 type)
371
372## Disabling a site
373
374To stop serving a site, go to **Settings → Sites** in your repository
375and click **Disable**. This removes the site configuration and stops
376serving the site. The deployed files are also deleted from storage.
377
378Releasing your domain from **Settings → Sites** at the account level
379will disable all sites associated with it and delete their files.
380
381
382# Knot self-hosting guide
383
384So you want to run your own knot server? Great! Here are a few prerequisites:
385
3861. A server of some kind (a VPS, a Raspberry Pi, etc.). Preferably running a Linux distribution of some kind.
3872. A (sub)domain name. People generally use `knot.example.com`.
3883. A valid SSL certificate for your domain.
389
390## NixOS
391
392Refer to the [knot
393module](https://tangled.org/tangled.org/core/blob/master/nix/modules/knot.nix)
394for a full list of options. Sample configurations:
395
396- [The test VM](https://tangled.org/tangled.org/core/blob/master/nix/vm.nix#L85)
397- [@pyrox.dev/nix](https://tangled.org/pyrox.dev/nix/blob/d19571cc1b5fe01035e1e6951ec8cf8a476b4dee/hosts/marvin/services/tangled.nix#L15-25)
398
399## Docker
400
401Refer to
402[@tangled.org/knot-docker](https://tangled.org/@tangled.org/knot-docker).
403Note that this is community maintained.
404
405## Manual setup
406
407First, clone this repository:
408
409```
410git clone https://tangled.org/@tangled.org/core
411```
412
413Then, build the `knot` CLI. This is the knot administration
414and operation tool. For the purpose of this guide, we're
415only concerned with these subcommands:
416
417- `knot server`: the main knot server process, typically
418 run as a supervised service
419- `knot guard`: handles role-based access control for git
420 over SSH (you'll never have to run this yourself)
421- `knot keys`: fetches SSH keys associated with your knot;
422 we'll use this to generate the SSH
423 `AuthorizedKeysCommand`
424
425```
426cd core
427export CGO_ENABLED=1
428go build -o knot ./cmd/knot
429```
430
431Next, move the `knot` binary to a location owned by `root` --
432`/usr/local/bin/` is a good choice. Make sure the binary itself is also owned by `root`:
433
434```
435sudo mv knot /usr/local/bin/knot
436sudo chown root:root /usr/local/bin/knot
437```
438
439This is necessary because SSH `AuthorizedKeysCommand` requires [really
440specific permissions](https://stackoverflow.com/a/27638306). The
441`AuthorizedKeysCommand` specifies a command that is run by `sshd` to
442retrieve a user's public SSH keys dynamically for authentication. Let's
443set that up.
444
445```
446sudo tee /etc/ssh/sshd_config.d/authorized_keys_command.conf <<EOF
447Match User git
448 AuthorizedKeysCommand /usr/local/bin/knot keys -o authorized-keys
449 AuthorizedKeysCommandUser nobody
450EOF
451```
452
453Then, reload `sshd`:
454
455```
456sudo systemctl reload ssh
457```
458
459Next, create the `git` user. We'll use the `git` user's home directory
460to store repositories:
461
462```
463sudo adduser git
464```
465
466Create `/home/git/.knot.env` with the following, updating the values as
467necessary. The `KNOT_SERVER_OWNER` should be set to your
468DID, you can find your DID in the [Settings](https://tangled.sh/settings) page.
469
470```
471KNOT_REPO_SCAN_PATH=/home/git
472KNOT_SERVER_HOSTNAME=knot.example.com
473APPVIEW_ENDPOINT=https://tangled.org
474KNOT_SERVER_OWNER=did:plc:foobar
475KNOT_SERVER_INTERNAL_LISTEN_ADDR=127.0.0.1:5444
476KNOT_SERVER_LISTEN_ADDR=127.0.0.1:5555
477```
478
479If you run a Linux distribution that uses systemd, you can
480use the provided service file to run the server. Copy
481[`knotserver.service`](https://tangled.org/tangled.org/core/blob/master/systemd/knotserver.service)
482to `/etc/systemd/system/`. Then, run:
483
484```
485systemctl enable knotserver
486systemctl start knotserver
487```
488
489The last step is to configure a reverse proxy like Nginx or Caddy to front your
490knot. Here's an example configuration for Nginx:
491
492```
493server {
494 listen 80;
495 listen [::]:80;
496 server_name knot.example.com;
497
498 location / {
499 proxy_pass http://localhost:5555;
500 proxy_set_header Host $host;
501 proxy_set_header X-Real-IP $remote_addr;
502 proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
503 proxy_set_header X-Forwarded-Proto $scheme;
504 }
505
506 # wss endpoint for git events
507 location /events {
508 proxy_set_header X-Forwarded-For $remote_addr;
509 proxy_set_header Host $http_host;
510 proxy_set_header Upgrade websocket;
511 proxy_set_header Connection Upgrade;
512 proxy_pass http://localhost:5555;
513 }
514 # additional config for SSL/TLS go here.
515}
516
517```
518
519Remember to use Let's Encrypt or similar to procure a certificate for your
520knot domain.
521
522You should now have a running knot server! You can finalize
523your registration by hitting the `verify` button on the
524[/settings/knots](https://tangled.org/settings/knots) page. This simply creates
525a record on your PDS to announce the existence of the knot.
526
527### Custom paths
528
529(This section applies to manual setup only. Docker users should edit the mounts
530in `docker-compose.yml` instead.)
531
532Right now, the database and repositories of your knot lives in `/home/git`. You
533can move these paths if you'd like to store them in another folder. Be careful
534when adjusting these paths:
535
536- Stop your knot when moving data (e.g. `systemctl stop knotserver`) to prevent
537 any possible side effects. Remember to restart it once you're done.
538- Make backups before moving in case something goes wrong.
539- Make sure the `git` user can read and write from the new paths.
540
541#### Database
542
543As an example, let's say the current database is at `/home/git/knotserver.db`,
544and we want to move it to `/home/git/database/knotserver.db`.
545
546Copy the current database to the new location. Make sure to copy the `.db-shm`
547and `.db-wal` files if they exist.
548
549```
550mkdir /home/git/database
551cp /home/git/knotserver.db* /home/git/database
552```
553
554In the environment (e.g. `/home/git/.knot.env`), set `KNOT_SERVER_DB_PATH` to
555the new file path (_not_ the directory):
556
557```
558KNOT_SERVER_DB_PATH=/home/git/database/knotserver.db
559```
560
561#### Repositories
562
563As an example, let's say the repositories are currently in `/home/git`, and we
564want to move them into `/home/git/repositories`.
565
566Create the new folder, then move the existing repositories (if there are any):
567
568```
569mkdir /home/git/repositories
570# move all DIDs into the new folder; these will vary for you!
571mv /home/git/did:plc:wshs7t2adsemcrrd4snkeqli /home/git/repositories
572```
573
574In the environment (e.g. `/home/git/.knot.env`), update `KNOT_REPO_SCAN_PATH`
575to the new directory:
576
577```
578KNOT_REPO_SCAN_PATH=/home/git/repositories
579```
580
581Similarly, update your `sshd` `AuthorizedKeysCommand` to use the updated
582repository path:
583
584```
585sudo tee /etc/ssh/sshd_config.d/authorized_keys_command.conf <<EOF
586Match User git
587 AuthorizedKeysCommand /usr/local/bin/knot keys -o authorized-keys -git-dir /home/git/repositories
588 AuthorizedKeysCommandUser nobody
589EOF
590```
591
592Make sure to restart your SSH server!
593
594#### MOTD (message of the day)
595
596To configure the MOTD used ("Welcome to this knot!" by default), edit the
597`/home/git/motd` file:
598
599```
600printf "Hi from this knot!\n" > /home/git/motd
601```
602
603Note that you should add a newline at the end if setting a non-empty message
604since the knot won't do this for you.
605
606## Secure Mode
607
608Secure Mode isolates each `git` subprocess to the repository it is
609operating on, using two mechanisms:
610
611- **Linux Landlock** restricts the filesystem paths the subprocess
612 can access -- it can only read/write its own repository and the
613 system directories it needs to run.
614- **UID isolation** runs each subprocess as a virtual UID assigned
615 to the repository owner, so that repositories belonging to
616 different owners are isolated from each other at the OS level
617 even if Landlock were somehow bypassed.
618
619Secure Mode requires:
620
621- Linux kernel >= 5.19 (Landlock V2). This is the minimum needed
622 for `git push` to work, because receive-pack's quarantine
623 migration uses cross-directory rename which requires the
624 Landlock `REFER` access right (added in V2). Kernels 5.13-5.18
625 support Landlock V1 and clones will work, but pushes will fail
626 with cross-device link errors. On kernels without any Landlock
627 support (< 5.13), the sandbox call is a no-op: UID isolation
628 still applies but no filesystem restriction is enforced.
629- `CAP_SETUID`, `CAP_SETGID`, and `CAP_CHOWN` available to the
630 knot process. The NixOS module grants these automatically; for
631 manual setups see the `setcap` step below.
632
633### NixOS
634
635Add `server.secureMode = true;` to your knot module configuration:
636
637```nix
638services.tangled.knot = {
639 server.secureMode = true;
640 # ... other options
641};
642```
643
644The NixOS module handles everything else automatically:
645
646- Grants the required capabilities to the knot service via
647 `AmbientCapabilities` in the systemd unit.
648- Installs a capability-bearing wrapper at
649 `/run/wrappers/bin/knot` via `security.wrappers`, so that
650 SSH-invoked git operations (pushes) also run under the correct
651 UID without requiring the service to run as root.
652- Runs `knot migrate-isolation` at service start to chown
653 existing repositories to their virtual UIDs.
654
655### Manual setup
656
657**Step 1.** Grant the required capabilities to the knot binary.
658This allows the knot process to switch to virtual UIDs at runtime
659without running as root. You will need to repeat this step
660whenever the binary is updated.
661
662```
663sudo setcap cap_setuid,cap_setgid,cap_chown+eip /usr/local/bin/knot
664```
665
666**Step 2.** Run the migration tool to assign virtual UIDs to all
667existing repositories and set their filesystem permissions. This
668must be run as root:
669
670```
671sudo knot migrate-isolation \
672 --git-dir /home/git \
673 --db /home/git/knotserver.db \
674 --internal-api 127.0.0.1:5444
675```
676
677You can re-run this at any time with `--force` to reapply
678permissions (e.g. after a manual repair or after updating the
679binary).
680
681**Step 2a.** Ensure the home directory is traversable by
682non-group users. Git subprocesses run as virtual UIDs that are
683not in the git group, and they need to resolve
684`$HOME/.config/git/config` to load the global config:
685
686```
687sudo chmod o+x /home/git
688```
689
690This adds only the execute bit, not read -- the virtual UIDs can
691traverse to known paths but cannot list directory contents.
692
693**Step 3.** Enable Secure Mode in your environment file:
694
695```
696KNOT_SERVER_SECURE_MODE=true
697```
698
699Or pass it as a flag:
700
701```
702knot server --secure-mode
703```
704
705**Step 4.** Regenerate the `AuthorizedKeysCommand` with the
706`-secure-mode` flag. This causes `knot keys` to emit guard
707command lines that include `-secure-mode`, so SSH pushes also
708get UID isolation:
709
710```
711sudo tee /etc/ssh/sshd_config.d/authorized_keys_command.conf <<EOF
712Match User git
713 AuthorizedKeysCommand /usr/local/bin/knot keys \
714 -o authorized-keys -secure-mode
715 AuthorizedKeysCommandUser nobody
716EOF
717```
718
719Reload `sshd` after making this change.
720
721> **Note:** the server will refuse to start in Secure Mode if any
722> repositories have not yet been isolation-migrated. Re-run
723> `migrate-isolation` if you see this error.
724
725## Troubleshooting
726
727If you run your own knot, you may run into some of these
728common issues. You can always join the
729[IRC](https://web.libera.chat/#tangled) or
730[Discord](https://chat.tangled.org/) if this section does
731not help.
732
733### Unable to push
734
735If you are unable to push to your knot or repository:
736
7371. First, ensure that you have added your SSH public key to
738 your account
7392. Check to see that your knot has synced the key by running
740 `knot keys`
7413. Check to see if git is supplying the correct private key
742 when pushing: `GIT_SSH_COMMAND="ssh -v" git push ...`
7434. Check to see if `sshd` on the knot is rejecting the push
744 for some reason: `journalctl -xeu ssh` (or `sshd`,
745 depending on your machine). These logs are unavailable if
746 using docker.
7475. Check to see if the knot itself is rejecting the push,
748 depending on your setup, the logs might be in one of the
749 following paths:
750 - `/tmp/knotguard.log`
751 - `/home/git/log`
752 - `/home/git/guard.log`
753
754# Spindles
755
756## Pipelines
757
758Spindle workflows allow you to write CI/CD pipelines in a
759simple format. They're located in the `.tangled/workflows`
760directory at the root of your repository, and are defined
761using YAML.
762
763A workflow has a set of common fields that apply no matter
764which engine you pick:
765
766- [Trigger](#trigger): A **required** field that defines
767 when a workflow should be triggered.
768- [Engine](#engine): A **required** field that defines which
769 engine a workflow should run on.
770- [Clone options](#clone-options): An **optional** field
771 that defines how the repository should be cloned.
772- [Environment](#environment): An **optional** field that
773 allows you to define environment variables.
774- [Steps](#steps): An **optional** field that allows you to
775 define what steps should run in the workflow.
776
777On top of these, each engine has its own options for things
778like dependencies and images. See [Engines](#engines) for
779the per-engine fields.
780
781### Trigger
782
783The first thing to add to a workflow is the trigger, which
784defines when a workflow runs. This is defined using a `when`
785field, which takes in a list of conditions. Each condition
786has the following fields:
787
788- `event`: This is a **required** field that defines when
789 your workflow should run. It's a list that can take one or
790 more of the following values:
791 - `push`: The workflow should run every time a commit is
792 pushed to the repository.
793 - `pull_request`: The workflow should run every time a
794 pull request is made or updated.
795 - `manual`: The workflow can be triggered manually.
796- `branch`: Defines which branches the workflow should run
797 for. If used with the `push` event, commits to the
798 branch(es) listed here will trigger the workflow. If used
799 with the `pull_request` event, updates to pull requests
800 targeting the branch(es) listed here will trigger the
801 workflow. This field has no effect with the `manual`
802 event. Supports glob patterns using `*` and `**` (e.g.,
803 `main`, `develop`, `release-*`). Either `branch` or `tag`
804 (or both) must be specified for `push` events.
805- `tag`: Defines which tags the workflow should run for.
806 Only used with the `push` event - when tags matching the
807 pattern(s) listed here are pushed, the workflow will
808 trigger. This field has no effect with `pull_request` or
809 `manual` events. Supports glob patterns using `*` and `**`
810 (e.g., `v*`, `v1.*`, `release-**`). Either `branch` or
811 `tag` (or both) must be specified for `push` events.
812
813For example, if you'd like to define a workflow that runs
814when commits are pushed to the `main` and `develop`
815branches, or when pull requests that target the `main`
816branch are updated, or manually, you can do so with:
817
818```yaml
819when:
820 - event: ["push", "manual"]
821 branch: ["main", "develop"]
822 - event: ["pull_request"]
823 branch: ["main"]
824```
825
826You can also trigger workflows on tag pushes. For instance,
827to run a deployment workflow when tags matching `v*` are
828pushed:
829
830```yaml
831when:
832 - event: ["push"]
833 tag: ["v*"]
834```
835
836You can even combine branch and tag patterns in a single
837constraint (the workflow triggers if either matches):
838
839```yaml
840when:
841 - event: ["push"]
842 branch: ["main", "release-*"]
843 tag: ["v*", "stable"]
844```
845
846### Engine
847
848Next is the engine on which the workflow should run, defined
849using the **required** `engine` field. The currently
850supported engines are:
851
852- `nixery`: This uses an instance of
853 [Nixery](https://nixery.dev) to run steps, which allows
854 you to add [dependencies](#dependencies) from
855 Nixpkgs (https://github.com/NixOS/nixpkgs). You can
856 search for packages on https://search.nixos.org, and
857 there's a pretty good chance the package(s) you're looking
858 for will be there.
859 See [Nixery engine](#nixery-engine).
860- `microvm`: Runs the whole workflow inside its own
861 microVM. Has configuration features for NixOS images
862 that will let you enable services, do Docker-in-VM, etc.
863 See [microVM engine](#microvm-engine).
864
865Example:
866
867```yaml
868engine: "nixery"
869```
870
871Each engine also adds its own workflow fields (dependencies,
872images, services, and so on). These are documented under
873[Engines](#engines).
874
875### Clone options
876
877When a workflow starts, the first step is to clone the
878repository. You can customize this behavior using the
879**optional** `clone` field. It has the following fields:
880
881- `skip`: Setting this to `true` will skip cloning the
882 repository. This can be useful if your workflow is doing
883 something that doesn't require anything from the
884 repository itself. This is `false` by default.
885- `depth`: This sets the number of commits, or the "clone
886 depth", to fetch from the repository. For example, if you
887 set this to 2, the last 2 commits will be fetched. By
888 default, the depth is set to 1, meaning only the most
889 recent commit will be fetched, which is the commit that
890 triggered the workflow.
891- `submodules`: If you use Git submodules
892 (https://git-scm.com/book/en/v2/Git-Tools-Submodules)
893 in your repository, setting this field to `true` will
894 recursively fetch all submodules. This is `false` by
895 default.
896
897The default settings are:
898
899```yaml
900clone:
901 skip: false
902 depth: 1
903 submodules: false
904```
905
906### Environment
907
908The `environment` field allows you define environment
909variables that will be available throughout the entire
910workflow. **Do not put secrets here, these environment
911variables are visible to anyone viewing the repository. You
912can add secrets for pipelines in your repository's
913settings.**
914
915Example:
916
917```yaml
918environment:
919 GOOS: "linux"
920 GOARCH: "arm64"
921 NODE_ENV: "production"
922 MY_ENV_VAR: "MY_ENV_VALUE"
923```
924
925By default, the following environment variables are set:
926
927- `CI` - Always set to `true` to indicate a CI environment
928- `TANGLED_PIPELINE_ID` - The AT URI of the current pipeline
929- `TANGLED_PIPELINE_KIND` - One of `push`, `pull_request` or
930 `manual`
931- `TANGLED_REPO_KNOT` - The repository's knot hostname
932- `TANGLED_REPO_DID` - The DID of the repository owner
933- `TANGLED_REPO_NAME` - The name of the repository
934- `TANGLED_REPO_DEFAULT_BRANCH` - The default branch of the
935 repository
936- `TANGLED_REPO_URL` - The full URL to the repository
937
938These variables are only available when the pipeline is
939triggered by a push:
940
941- `TANGLED_REF` - The full git reference (e.g.,
942 `refs/heads/main` or `refs/tags/v1.0.0`)
943- `TANGLED_REF_NAME` - The short name of the reference
944 (e.g., `main` or `v1.0.0`)
945- `TANGLED_REF_TYPE` - The type of reference, either
946 `branch` or `tag`
947- `TANGLED_SHA` - The commit SHA that triggered the pipeline
948- `TANGLED_COMMIT_SHA` - Alias for `TANGLED_SHA`
949
950These variables are only available when the pipeline is
951triggered by a pull request:
952
953- `TANGLED_PR_SOURCE_BRANCH` - The source branch of the pull
954 request
955- `TANGLED_PR_TARGET_BRANCH` - The target branch of the pull
956 request
957- `TANGLED_PR_SOURCE_SHA` - The commit SHA of the source
958 branch
959
960### Steps
961
962The `steps` field allows you to define what steps should run
963in the workflow. It's a list of step objects, each with the
964following fields:
965
966- `name`: This field allows you to give your step a name.
967 This name is visible in your workflow runs, and is used to
968 describe what the step is doing.
969- `command`: This field allows you to define a command to
970 run in that step. The step is run in a Bash shell, and the
971 logs from the command will be visible in the pipelines
972 page on the Tangled website. Any dependencies you added in
973 your engine's section (see [Engines](#engines)) will be
974 available to use here.
975- `environment`: Similar to the global
976 [environment](#environment) config, this **optional**
977 field is a key-value map that allows you to set
978 environment variables for the step. **Do not put secrets
979 here, these environment variables are visible to anyone
980 viewing the repository. You can add secrets for pipelines
981 in your repository's settings.**
982
983Example:
984
985```yaml
986steps:
987 - name: "Build backend"
988 command: "go build"
989 environment:
990 GOOS: "darwin"
991 GOARCH: "arm64"
992 - name: "Build frontend"
993 command: "npm run build"
994 environment:
995 NODE_ENV: "production"
996```
997
998## Engines
999
1000The common fields above apply to every workflow. Each engine
1001then adds its own fields on top. Pick an engine with the
1002[`engine`](#engine) field and use the matching section below.
1003
1004### Nixery engine
1005
1006#### Dependencies
1007
1008When you're running a workflow you'll usually need additional
1009dependencies. The `dependencies` field lets you define which
1010dependencies to get, and from where. It's a key-value map,
1011with the key being the registry to fetch dependencies from,
1012and the value being the list of dependencies to fetch.
1013
1014The registry URL syntax can be found [on the nix
1015manual](https://nix.dev/manual/nix/2.18/command-ref/new-cli/nix3-registry-add).
1016
1017Say you want to fetch Node.js and Go from `nixpkgs`, and a
1018package called `my_pkg` you've made from your own registry
1019at your repository at
1020`https://tangled.org/@example.com/my_pkg`. You can define
1021those dependencies like so:
1022
1023```yaml
1024dependencies:
1025 # nixpkgs
1026 nixpkgs:
1027 - nodejs
1028 - go
1029 # unstable
1030 nixpkgs/nixpkgs-unstable:
1031 - bun
1032 # custom registry
1033 git+https://tangled.org/@example.com/my_pkg:
1034 - my_pkg
1035```
1036
1037Now these dependencies are available to use in your
1038workflow!
1039
1040#### Complete nixery workflow
1041
1042```yaml
1043# .tangled/workflows/build.yml
1044
1045when:
1046 - event: ["push", "manual"]
1047 branch: ["main", "develop"]
1048 - event: ["pull_request"]
1049 branch: ["main"]
1050
1051engine: "nixery"
1052
1053# using the default values
1054clone:
1055 skip: false
1056 depth: 1
1057 submodules: false
1058
1059dependencies:
1060 # nixpkgs
1061 nixpkgs:
1062 - nodejs
1063 - go
1064 # custom registry
1065 git+https://tangled.org/@example.com/my_pkg:
1066 - my_pkg
1067
1068environment:
1069 GOOS: "linux"
1070 GOARCH: "arm64"
1071 NODE_ENV: "production"
1072 MY_ENV_VAR: "MY_ENV_VALUE"
1073
1074steps:
1075 - name: "Build backend"
1076 command: "go build"
1077 environment:
1078 GOOS: "darwin"
1079 GOARCH: "arm64"
1080 - name: "Build frontend"
1081 command: "npm run build"
1082 environment:
1083 NODE_ENV: "production"
1084```
1085
1086If you want another example of a workflow, you can look at
1087the one [Tangled uses to build the
1088project](https://tangled.org/@tangled.org/core/blob/master/.tangled/workflows/build.yml).
1089
1090### microVM engine
1091
1092#### Image
1093
1094A workflow picks the image to boot with the top-level `image`
1095field:
1096
1097```yaml
1098engine: microvm
1099image: nixos
1100```
1101
1102There are two flavours of images:
1103
1104- **NixOS images** (e.g. `nixos`): the whole guest is built
1105 with Nix, so you can configure it from the workflow file
1106 itself. The `dependencies`, `services`, `virtualisation`,
1107 `registry` and `caches` fields below are all understood
1108 here, and the guest builds and activates that configuration
1109 before any of your steps run.
1110- **Non-NixOS images** (e.g. `alpine`): there's no NixOS to
1111 configure, so the workflow-level config fields above have
1112 no effect. You still get a full machine to run steps in.
1113
1114The available image names depend on what the spindle operator
1115has installed. `nixos` and `alpine` are examples. If `image`
1116is omitted, the spindle's configured default image is used.
1117
1118#### Dependencies
1119
1120On the microVM engine, `dependencies` is a flat list of
1121packages that are made available to every step. This field
1122only applies to **NixOS images**; for other images you can
1123use the package manager included in a step.
1124
1125The guest builds a [`nix develop`](https://nix.dev/manual/nix/2.18/command-ref/new-cli/nix3-develop)-style
1126devshell from your dependencies and uses it for each step,
1127so you can, for example, add `pkg-config` and `openssl` and
1128have the `openssl-sys` crate while compiling a Rust project
1129just work.
1130
1131A bare name like `go` is looked up in nixpkgs. You can also
1132point at any flake with the `flakeref#attr` syntax, so
1133`github:nixos/nixpkgs#hello` pulls `hello` straight out of
1134that flake.
1135
1136```yaml
1137dependencies:
1138 - go
1139 - github:nixos/nixpkgs#hello
1140```
1141
1142#### Registry
1143
1144The `registry` field remaps flake references, the same way
1145`nix registry` does. This lets you pin or alias the flakes
1146used by `dependencies`.
1147
1148For example, pin `nixpkgs` to `nixos-unstable` so that the
1149bare `go` above resolves from unstable, and alias your own
1150flake so you can use `myflake#tool` in `dependencies`:
1151
1152```yaml
1153registry:
1154 nixpkgs: github:nixos/nixpkgs/nixos-unstable
1155 myflake: github:me/x
1156```
1157
1158#### Caches
1159
1160The `caches` field is a map of Nix binary cache URL to its
1161trusted public key. These are fed into the spindle's read
1162proxy, so the guest can substitute prebuilt paths from them
1163instead of building everything from scratch.
1164
1165```yaml
1166caches:
1167 https://nix-community.cachix.org: "nix-community.cachix.org-1:mB9FSh9qf2dCimDSUo8Zy7bkq5CX+/rkCWyvRCYg3Fs="
1168```
1169
1170#### Services and virtualisation
1171
1172The `services` and `virtualisation` fields are passed straight
1173through to NixOS. Anything you could write under
1174`services.*` or `virtualisation.*` in a NixOS configuration,
1175you can write here, and it's brought up before any of your
1176steps run.
1177
1178As a convenience, `true` works as shorthand for
1179`.enable = true` anywhere an `enable` option exists (e.g.
1180`virtualisation.docker: true`).
1181
1182```yaml
1183services:
1184 postgresql:
1185 enable: true
1186 ensureDatabases: ["spindle-workflow"]
1187 ensureUsers:
1188 - name: spindle-workflow
1189 ensureDBOwnership: true
1190
1191virtualisation:
1192 docker: true
1193```
1194
1195#### Recipes
1196
1197##### Lint, test and build a Node project
1198
1199```yaml
1200when:
1201 - event: ["push", "pull_request"]
1202 branch: ["main"]
1203
1204engine: microvm
1205image: nixos
1206
1207dependencies:
1208 - pnpm
1209
1210steps:
1211 - name: "Install dependencies"
1212 command: pnpm install --frozen-lockfile
1213 - name: "Lint and test"
1214 command: |
1215 pnpm run lint
1216 pnpm test
1217 - name: "Build"
1218 command: pnpm run build
1219```
1220
1221##### Build a Rust project that links OpenSSL
1222
1223```yaml
1224when:
1225 - event: ["push", "pull_request"]
1226 branch: ["main"]
1227
1228engine: microvm
1229image: nixos
1230
1231dependencies:
1232 - gcc
1233 - cargo
1234 - rustc
1235 - clippy
1236 - rustfmt
1237 - pkg-config # exports PKG_CONFIG_PATH for the libraries below
1238 - openssl # the C library + headers openssl-sys links against
1239
1240steps:
1241 - name: "Check formatting"
1242 command: cargo fmt --check
1243 - name: "Clippy"
1244 command: cargo clippy --all-targets -- -D warnings
1245 - name: "Test"
1246 command: cargo test --all
1247 - name: "Release build"
1248 command: cargo build --release
1249```
1250
1251##### Run migrations and integration tests against PostgreSQL
1252
1253```yaml
1254when:
1255 - event: ["push", "pull_request"]
1256 branch: ["main"]
1257
1258engine: microvm
1259image: nixos
1260
1261environment:
1262 DATABASE_URL: "postgresql:///spindle-workflow?host=/run/postgresql"
1263
1264dependencies:
1265 - gcc
1266 - cargo
1267 - rustc
1268 - pkg-config
1269 - openssl
1270 - sqlx-cli
1271
1272services:
1273 postgresql:
1274 enable: true
1275 # has to be same name as the user for peer auth to work automatically
1276 ensureDatabases: ["spindle-workflow"]
1277 ensureUsers:
1278 - name: spindle-workflow
1279 ensureDBOwnership: true
1280
1281steps:
1282 - name: "Run migrations"
1283 command: sqlx migrate run
1284 - name: "Integration tests"
1285 command: cargo test --all
1286```
1287
1288##### Build and push a Docker image on tag
1289
1290```yaml
1291when:
1292 - event: ["push"]
1293 tag: ["v*"]
1294
1295engine: microvm
1296image: nixos
1297
1298virtualisation:
1299 docker: true
1300
1301steps:
1302 - name: "Build and push to ghcr.io"
1303 command: |
1304 set -euo pipefail
1305
1306 echo "$REGISTRY_TOKEN" | docker login ghcr.io -u "$REGISTRY_USER" --password-stdin
1307 image="ghcr.io/$REGISTRY_USER/myapp:$TANGLED_REF_NAME"
1308
1309 docker build -t "$image" -t "ghcr.io/$REGISTRY_USER/myapp:latest" .
1310 docker push "$image"
1311 docker push "ghcr.io/$REGISTRY_USER/myapp:latest"
1312```
1313
1314##### Deploy to Cloudflare Workers on tag
1315
1316```yaml
1317# .tangled/workflows/deploy.yml
1318when:
1319 - event: ["push"]
1320 tag: ["v*"]
1321
1322engine: microvm
1323image: nixos
1324
1325dependencies:
1326 - pnpm
1327
1328steps:
1329 - name: "Install dependencies"
1330 command: pnpm install --frozen-lockfile
1331 - name: "Deploy worker"
1332 # `wrangler` picks up `CLOUDFLARE_API_TOKEN` from the env.
1333 # set it under **Settings → Secrets**.
1334 command: pnpm exec wrangler deploy
1335```
1336
1337##### Publish a release artifact
1338
1339```yaml
1340when:
1341 - event: ["push"]
1342 tag: ["v*"] # trigger on versions
1343
1344engine: microvm
1345image: nixos
1346
1347dependencies:
1348 - go
1349
1350steps:
1351 - name: "Build release binary"
1352 command: |
1353 mkdir -p dist
1354 CGO_ENABLED=0 go build -trimpath -ldflags "-s -w" -o dist/myapp ./cmd/myapp
1355
1356 - name: "Publish artifact record"
1357 command: |
1358 set -euo pipefail
1359 # change this if you're not on `tngl.sh`
1360 PDS="https://tngl.sh"
1361 # also update this to your handle or did
1362 ATP_IDENTIFIER="user.tngl.sh"
1363 ARTIFACT_PATH="dist/myapp"
1364 ARTIFACT_NAME="myapp"
1365
1366 # set `ATP_APP_PASSWORD` under **Settings → Secrets**
1367 session=$(curl -fsS -X POST "$PDS/xrpc/com.atproto.server.createSession" \
1368 -H "Content-Type: application/json" \
1369 -d "{\"identifier\":\"$ATP_IDENTIFIER\",\"password\":\"$ATP_APP_PASSWORD\"}")
1370 jwt=$(echo "$session" | jq -r .accessJwt)
1371 did=$(echo "$session" | jq -r .did)
1372
1373 # upload the binary as a blob
1374 blob=$(curl -fsS -X POST "$PDS/xrpc/com.atproto.repo.uploadBlob" \
1375 -H "Authorization: Bearer $jwt" \
1376 -H "Content-Type: application/octet-stream" \
1377 --data-binary @"$ARTIFACT_PATH")
1378
1379 # note that this requires an annotated tag (`git tag -a v1.0.0 -m ...`)
1380 tag_hash=$(git rev-parse "$TANGLED_REF_NAME^{tag}")
1381 tag_bytes=$(printf '%s' "$tag_hash" | xxd -r -p | base64 | tr -d '=')
1382
1383 # the sh.tangled.repo.artifact record for your artifact
1384 record=$(jq -n \
1385 --arg did "$did" \
1386 --arg tag "$tag_bytes" \
1387 --arg name "$ARTIFACT_NAME" \
1388 --arg repo "$TANGLED_REPO_URL" \
1389 --arg created "$(date -Iseconds)" \
1390 --argjson blob "$(echo "$blob" | jq .blob)" '{
1391 repo: $did,
1392 collection: "sh.tangled.repo.artifact",
1393 validate: false,
1394 record: {
1395 "$type": "sh.tangled.repo.artifact",
1396 tag: {"$bytes": $tag},
1397 name: $name,
1398 repo: $repo,
1399 artifact: $blob,
1400 createdAt: $created
1401 }
1402 }')
1403
1404 # create the record on the PDS
1405 curl -fsS -X POST "$PDS/xrpc/com.atproto.repo.createRecord" \
1406 -H "Authorization: Bearer $jwt" \
1407 -H "Content-Type: application/json" \
1408 -d "$record"
1409```
1410
1411## Self-hosting guide
1412
1413### Prerequisites
1414
1415- Go
1416- For the **nixery** engine: Docker (or Podman with Docker
1417 compatibility enabled).
1418- For the **microVM** engine: a Linux host with KVM, plus the
1419 microVM host dependencies described in [Running microVM
1420 workflows](#running-microvm-workflows).
1421
1422### Configuration
1423
1424Spindle is configured using environment variables. The following environment variables are available:
1425
1426- `SPINDLE_SERVER_LISTEN_ADDR`: The address the server listens on (default: `"0.0.0.0:6555"`).
1427- `SPINDLE_SERVER_DB_PATH`: The path to the SQLite database file (default: `"spindle.db"`).
1428- `SPINDLE_SERVER_HOSTNAME`: The hostname of the server (required).
1429- `SPINDLE_SERVER_JETSTREAM_ENDPOINT`: The endpoint of the Jetstream server (default: `"wss://jetstream1.us-west.bsky.network/subscribe"`).
1430- `SPINDLE_SERVER_DEV`: A boolean indicating whether the server is running in development mode (default: `false`).
1431- `SPINDLE_SERVER_OWNER`: The DID of the owner (required).
1432- `SPINDLE_SERVER_LOG_DIR`: The directory to store workflow logs (default: `"/var/log/spindle"`).
1433- `SPINDLE_SERVER_DOCKER_SOCKET`: Path to Docker socket to expose to invoked Spindle containers (default: `""`).
1434- `SPINDLE_PIPELINES_NIXERY`: The Nixery URL (default: `"nixery.tangled.sh"`).
1435- `SPINDLE_PIPELINES_WORKFLOW_TIMEOUT`: The default workflow timeout (default: `"5m"`).
1436
1437For the microVM engine, the following are also available
1438(prefix `SPINDLE_MICROVM_PIPELINES_`):
1439
1440- `SPINDLE_MICROVM_PIPELINES_IMAGE_DIR`: Directory containing
1441 microVM images (**required** to use the engine). See
1442 [Running microVM workflows](#running-microvm-workflows).
1443- `SPINDLE_MICROVM_PIPELINES_DEFAULT_IMAGE`: Image used when a
1444 workflow doesn't set `image` (default: `"nixos-x86_64"`).
1445- `SPINDLE_MICROVM_PIPELINES_OVERLAY_DIR`: Where per-workflow
1446 temporary disks are created (default: the system temp dir).
1447- `SPINDLE_MICROVM_PIPELINES_ENABLE_KVM`: Use KVM hardware
1448 acceleration (default: `true`). Without KVM, guests fall
1449 back to slow software emulation.
1450- `SPINDLE_MICROVM_PIPELINES_WORKFLOW_TIMEOUT`: Default
1451 workflow timeout (default: `"5m"`).
1452
1453Optional resource limits (a value of `0` disables that
1454limit). The limits cap usage across all running microVM
1455workflows:
1456
1457- `SPINDLE_MICROVM_PIPELINES_MAX_TOTAL_MEMORY_MIB`
1458- `SPINDLE_MICROVM_PIPELINES_MAX_TOTAL_VCPUS`
1459- `SPINDLE_MICROVM_PIPELINES_MAX_TOTAL_DISK_MIB`
1460
1461Optional cgroup enforcement:
1462
1463- `SPINDLE_MICROVM_PIPELINES_ENABLE_CGROUPS`: Place each
1464 workflow's QEMU and slirp4netns in a per-workflow cgroup=
1465 (default: `false`).
1466- `SPINDLE_MICROVM_PIPELINES_CGROUP_PARENT`: Parent cgroup;
1467 `self` resolves the spindle service's own cgroup (default:
1468 `"self"`).
1469- `SPINDLE_MICROVM_PIPELINES_CGROUP_PIDS_MAX`: Max processes
1470 per workflow cgroup (default: `4096`).
1471- `SPINDLE_MICROVM_PIPELINES_CGROUP_SWAP_MAX_MIB`: Max swap
1472 per workflow cgroup (default: `0`, no swap).
1473- `SPINDLE_MICROVM_PIPELINES_CGROUP_SUPERVISOR_MEMORY_MIN_MIB`:
1474 Memory protected for spindle itself so it isn't OOM-killed
1475 before the workflows (default: `512`).
1476
1477To push paths built inside microVMs back to a shared Nix
1478cache (and read from it), configure the cache (prefix
1479`SPINDLE_NIX_CACHE_`):
1480
1481- `SPINDLE_NIX_CACHE_READ_URLS`: Comma-separated binary cache
1482 URLs the guest reads from.
1483- `SPINDLE_NIX_CACHE_TRUSTED_PUBLIC_KEYS`: Comma-separated
1484 trusted public keys for those caches.
1485- `SPINDLE_NIX_CACHE_UPLOAD_URL`: Cache URL that paths built
1486 in the guest are uploaded to.
1487
1488### Running spindle
1489
14901. **Set the environment variables.** For example:
1491
1492 ```shell
1493 export SPINDLE_SERVER_HOSTNAME="your-hostname"
1494 export SPINDLE_SERVER_OWNER="your-did"
1495 ```
1496
14972. **Build the Spindle binary.**
1498
1499 ```shell
1500 cd core
1501 go mod download
1502 go build -o cmd/spindle/spindle cmd/spindle/main.go
1503 ```
1504
15053. **Create the log directory.**
1506
1507 ```shell
1508 sudo mkdir -p /var/log/spindle
1509 sudo chown $USER:$USER -R /var/log/spindle
1510 ```
1511
15124. **Run the Spindle binary.**
1513
1514 ```shell
1515 ./cmd/spindle/spindle
1516 ```
1517
1518Spindle will now start, connect to the Jetstream server, and begin processing pipelines.
1519
1520### Running microVM workflows
1521
1522The microVM engine needs a few extra things on the host, and
1523it needs images to boot.
1524
1525#### Host dependencies
1526
1527microVM workflows depend on a handful of host tools and
1528devices. spindle checks for the ones an image needs right
1529before it launches, so a missing dependency surfaces as a
1530clear error. You'll need:
1531
1532- `qemu`: the runner. The QEMU binary for the image's arch
1533 must be present (e.g. `qemu-system-x86_64`).
1534- `mkfs.ext4` (from `e2fsprogs`): to format the per-workflow
1535 writable volumes.
1536- [`slirp4netns`](https://github.com/rootless-containers/slirp4netns#install),
1537 `ip` (from `iproute2`), `mount` and `unshare` (from `util-linux`):
1538 used to sandbox guest networking.
1539- `/dev/kvm`: for hardware acceleration (unless you disable
1540 KVM with `SPINDLE_MICROVM_PIPELINES_ENABLE_KVM=false`).
1541- `/dev/vhost-vsock`: the guest agent talks to spindle over
1542 vsock.
1543
1544On NixOS, the [spindle
1545module](https://tangled.org/tangled.org/core/blob/master/nix/modules/spindle.nix)
1546puts `qemu`, `e2fsprogs`, `slirp4netns`, `iproute2` and
1547`util-linux` on the service's `PATH` for you.
1548
1549#### Building images
1550
1551Images are built with Nix. The flake exposes packages for the
1552two stock images (use the `-tarball` prefixed ones for a gzipped
1553tarball you can copy to another host):
1554
1555```shell
1556# a NixOS image
1557nix build .#spindle-nixos-image
1558# an Alpine image
1559nix build .#spindle-alpine-image
1560```
1561
1562#### Installing images
1563
1564Spindle looks for images in
1565`SPINDLE_MICROVM_PIPELINES_IMAGE_DIR`. An image is resolved by
1566the name a workflow puts in its `image` field, matched
1567literally against what's on disk:
1568
15691. a directory `<name>/` containing a `spec.json` (next to the
1570 kernel/initrd/store-disk), or
15712. a flat `<name>.json` self-contained spec.
1572
1573Resolution depends only on the name and what's on disk, never
1574on the host doing the resolving, so the same workflow resolves
1575to the same image on every spindle. If you keep multiple
1576arches side by side, you can name them `<name>-<arch>` (e.g.
1577`nixos-x86_64`, `alpine-aarch64`); the suffix is just part of
1578the name. To make a name like `nixos` work if you are hosting
1579multiple arches, you can use symlinks.
1580
1581On NixOS, you'll most likely want to use `systemd.tmpfiles.rules`
1582to set these up declaratively.
1583
1584## Architecture
1585
1586Spindle is a small CI runner service. Here's a high-level overview of how it operates:
1587
1588- Listens for [`sh.tangled.spindle.member`](/lexicons/spindle/member.json) and
1589 [`sh.tangled.repo`](/lexicons/repo.json) records on the Jetstream.
1590- When a new repo record comes through (typically when you add a spindle to a
1591 repo from the settings), spindle then resolves the underlying knot and
1592 subscribes to repo events (see:
1593 [`sh.tangled.pipeline`](/lexicons/pipeline.json)).
1594- The spindle engine then handles execution of the pipeline, with results and
1595 logs beamed on the spindle event stream over WebSocket
1596
1597### The engines
1598
1599Spindle has two execution backends, picked per-workflow with
1600the [`engine`](#engine) field:
1601
1602- **nixery**: executes each step in a fresh Docker container
1603 (Podman works too, if Docker compatibility is enabled so
1604 that `/run/docker.sock` is created), with state persisted
1605 across steps within the `/tangled/workspace` directory. The
1606 base image for the container is constructed on the fly using
1607 [Nixery](https://nixery.dev), which is/rhandy for caching
1608 layers for frequently used packages.
1609- **microvm**: runs the whole workflow inside its own
1610 microVM, supporting different images, with extra
1611 configuration for NixOS images (e.g. services in workflow file)
1612 See the [engine
1613 README](https://tangled.org/tangled.org/core/blob/master/spindle/engines/microvm/README.md)
1614 for the architecture in depth.
1615
1616The pipeline manifest is [specified here](https://docs.tangled.org/spindles.html#pipelines).
1617
1618## Secrets with openbao
1619
1620This document covers setting up spindle to use OpenBao for secrets
1621management via OpenBao Proxy instead of the default SQLite backend.
1622
1623### Overview
1624
1625Spindle now uses OpenBao Proxy for secrets management. The proxy handles
1626authentication automatically using AppRole credentials, while spindle
1627connects to the local proxy instead of directly to the OpenBao server.
1628
1629This approach provides better security, automatic token renewal, and
1630simplified application code.
1631
1632### Installation
1633
1634Install OpenBao from Nixpkgs:
1635
1636```bash
1637nix shell nixpkgs#openbao # for a local server
1638```
1639
1640### Setup
1641
1642The setup process can is documented for both local development and production.
1643
1644#### Local development
1645
1646Start OpenBao in dev mode:
1647
1648```bash
1649bao server -dev -dev-root-token-id="root" -dev-listen-address=127.0.0.1:8201
1650```
1651
1652This starts OpenBao on `http://localhost:8201` with a root token.
1653
1654Set up environment for bao CLI:
1655
1656```bash
1657export BAO_ADDR=http://localhost:8200
1658export BAO_TOKEN=root
1659```
1660
1661#### Production
1662
1663You would typically use a systemd service with a
1664configuration file. Refer to
1665[@tangled.org/infra](https://tangled.org/@tangled.org/infra)
1666for how this can be achieved using Nix.
1667
1668Then, initialize the bao server:
1669
1670```bash
1671bao operator init -key-shares=1 -key-threshold=1
1672```
1673
1674This will print out an unseal key and a root key. Save them
1675somewhere (like a password manager). Then unseal the vault
1676to begin setting it up:
1677
1678```bash
1679bao operator unseal <unseal_key>
1680```
1681
1682All steps below remain the same across both dev and
1683production setups.
1684
1685#### Configure openbao server
1686
1687Create the spindle KV mount:
1688
1689```bash
1690bao secrets enable -path=spindle -version=2 kv
1691```
1692
1693Set up AppRole authentication and policy:
1694
1695Create a policy file `spindle-policy.hcl`:
1696
1697```hcl
1698# Full access to spindle KV v2 data
1699path "spindle/data/*" {
1700 capabilities = ["create", "read", "update", "delete"]
1701}
1702
1703# Access to metadata for listing and management
1704path "spindle/metadata/*" {
1705 capabilities = ["list", "read", "delete", "update"]
1706}
1707
1708# Allow listing at root level
1709path "spindle/" {
1710 capabilities = ["list"]
1711}
1712
1713# Required for connection testing and health checks
1714path "auth/token/lookup-self" {
1715 capabilities = ["read"]
1716}
1717```
1718
1719Apply the policy and create an AppRole:
1720
1721```bash
1722bao policy write spindle-policy spindle-policy.hcl
1723bao auth enable approle
1724bao write auth/approle/role/spindle \
1725 token_policies="spindle-policy" \
1726 token_ttl=1h \
1727 token_max_ttl=4h \
1728 bind_secret_id=true \
1729 secret_id_ttl=0 \
1730 secret_id_num_uses=0
1731```
1732
1733Get the credentials:
1734
1735```bash
1736# Get role ID (static)
1737ROLE_ID=$(bao read -field=role_id auth/approle/role/spindle/role-id)
1738
1739# Generate secret ID
1740SECRET_ID=$(bao write -f -field=secret_id auth/approle/role/spindle/secret-id)
1741
1742echo "Role ID: $ROLE_ID"
1743echo "Secret ID: $SECRET_ID"
1744```
1745
1746#### Create proxy configuration
1747
1748Create the credential files:
1749
1750```bash
1751# Create directory for OpenBao files
1752mkdir -p /tmp/openbao
1753
1754# Save credentials
1755echo "$ROLE_ID" > /tmp/openbao/role-id
1756echo "$SECRET_ID" > /tmp/openbao/secret-id
1757chmod 600 /tmp/openbao/role-id /tmp/openbao/secret-id
1758```
1759
1760Create a proxy configuration file `/tmp/openbao/proxy.hcl`:
1761
1762```hcl
1763# OpenBao server connection
1764vault {
1765 address = "http://localhost:8200"
1766}
1767
1768# Auto-Auth using AppRole
1769auto_auth {
1770 method "approle" {
1771 mount_path = "auth/approle"
1772 config = {
1773 role_id_file_path = "/tmp/openbao/role-id"
1774 secret_id_file_path = "/tmp/openbao/secret-id"
1775 }
1776 }
1777
1778 # Optional: write token to file for debugging
1779 sink "file" {
1780 config = {
1781 path = "/tmp/openbao/token"
1782 mode = 0640
1783 }
1784 }
1785}
1786
1787# Proxy listener for spindle
1788listener "tcp" {
1789 address = "127.0.0.1:8201"
1790 tls_disable = true
1791}
1792
1793# Enable API proxy with auto-auth token
1794api_proxy {
1795 use_auto_auth_token = true
1796}
1797
1798# Enable response caching
1799cache {
1800 use_auto_auth_token = true
1801}
1802
1803# Logging
1804log_level = "info"
1805```
1806
1807#### Start the proxy
1808
1809Start OpenBao Proxy:
1810
1811```bash
1812bao proxy -config=/tmp/openbao/proxy.hcl
1813```
1814
1815The proxy will authenticate with OpenBao and start listening on
1816`127.0.0.1:8201`.
1817
1818#### Configure spindle
1819
1820Set these environment variables for spindle:
1821
1822```bash
1823export SPINDLE_SERVER_SECRETS_PROVIDER=openbao
1824export SPINDLE_SERVER_SECRETS_OPENBAO_PROXY_ADDR=http://127.0.0.1:8201
1825export SPINDLE_SERVER_SECRETS_OPENBAO_MOUNT=spindle
1826```
1827
1828On startup, spindle will now connect to the local proxy,
1829which handles all authentication automatically.
1830
1831### Production setup for proxy
1832
1833For production, you'll want to run the proxy as a service:
1834
1835Place your production configuration in
1836`/etc/openbao/proxy.hcl` with proper TLS settings for the
1837vault connection.
1838
1839### Verifying setup
1840
1841Test the proxy directly:
1842
1843```bash
1844# Check proxy health
1845curl -H "X-Vault-Request: true" http://127.0.0.1:8201/v1/sys/health
1846
1847# Test token lookup through proxy
1848curl -H "X-Vault-Request: true" http://127.0.0.1:8201/v1/auth/token/lookup-self
1849```
1850
1851Test OpenBao operations through the server:
1852
1853```bash
1854# List all secrets
1855bao kv list spindle/
1856
1857# Add a test secret via the spindle API, then check it exists
1858bao kv list spindle/repos/
1859
1860# Get a specific secret
1861bao kv get spindle/repos/your_repo_path/SECRET_NAME
1862```
1863
1864### How it works
1865
1866- Spindle connects to OpenBao Proxy on localhost (typically
1867 port 8200 or 8201)
1868- The proxy authenticates with OpenBao using AppRole
1869 credentials
1870- All spindle requests go through the proxy, which injects
1871 authentication tokens
1872- Secrets are stored at
1873 `spindle/repos/{sanitized_repo_path}/{secret_key}`
1874- Repository paths like `did:plc:alice/myrepo` become
1875 `did_plc_alice_myrepo`
1876- The proxy handles all token renewal automatically
1877- Spindle no longer manages tokens or authentication
1878 directly
1879
1880### Troubleshooting
1881
1882**Connection refused**: Check that the OpenBao Proxy is
1883running and listening on the configured address.
1884
1885**403 errors**: Verify the AppRole credentials are correct
1886and the policy has the necessary permissions.
1887
1888**404 route errors**: The spindle KV mount probably doesn't
1889exist—run the mount creation step again.
1890
1891**Proxy authentication failures**: Check the proxy logs and
1892verify the role-id and secret-id files are readable and
1893contain valid credentials.
1894
1895**Secret not found after writing**: This can indicate policy
1896permission issues. Verify the policy includes both
1897`spindle/data/*` and `spindle/metadata/*` paths with
1898appropriate capabilities.
1899
1900Check proxy logs:
1901
1902```bash
1903# If running as systemd service
1904journalctl -u openbao-proxy -f
1905
1906# If running directly, check the console output
1907```
1908
1909Test AppRole authentication manually:
1910
1911```bash
1912bao write auth/approle/login \
1913 role_id="$(cat /tmp/openbao/role-id)" \
1914 secret_id="$(cat /tmp/openbao/secret-id)"
1915```
1916
1917# Webhooks
1918
1919Webhooks allow you to receive HTTP POST notifications when events occur in your repositories. This enables you to integrate Tangled with external services, trigger CI/CD pipelines, send notifications, or automate workflows.
1920
1921## Overview
1922
1923Webhooks send HTTP POST requests to URLs you configure whenever specific events happen. Currently, Tangled supports push events, with more event types coming soon.
1924
1925## Configuring webhooks
1926
1927To set up a webhook for your repository:
1928
19291. Navigate to your repository
19302. Go to **Settings → Hooks**
19313. Click **new webhook**
19324. Configure your webhook:
1933 - **Payload URL**: The endpoint that will receive the webhook POST requests
1934 - **Secret**: An optional secret key for verifying webhook authenticity (leave blank to send unsigned webhooks)
1935 - **Events**: Select which events trigger the webhook (currently only push events)
1936 - **Active**: Toggle whether the webhook is enabled
1937
1938## Webhook payload
1939
1940### Push
1941
1942When a push event occurs, Tangled sends a POST request with a JSON payload of the format:
1943
1944```json
1945{
1946 "after": "7b320e5cbee2734071e4310c1d9ae401d8f6cab5",
1947 "before": "c04ddf64eddc90e4e2a9846ba3b43e67a0e2865e",
1948 "pusher": {
1949 "did": "did:plc:hwevmowznbiukdf6uk5dwrrq"
1950 },
1951 "ref": "refs/heads/main",
1952 "repository": {
1953 "clone_url": "https://tangled.org/did:plc:hwevmowznbiukdf6uk5dwrrq/some-repo",
1954 "created_at": "2025-09-15T08:57:23Z",
1955 "description": "an example repository",
1956 "fork": false,
1957 "full_name": "did:plc:hwevmowznbiukdf6uk5dwrrq/some-repo",
1958 "html_url": "https://tangled.org/did:plc:hwevmowznbiukdf6uk5dwrrq/some-repo",
1959 "name": "some-repo",
1960 "open_issues_count": 5,
1961 "owner": {
1962 "did": "did:plc:hwevmowznbiukdf6uk5dwrrq"
1963 },
1964 "ssh_url": "ssh://git@tangled.org/did:plc:hwevmowznbiukdf6uk5dwrrq/some-repo",
1965 "stars_count": 1,
1966 "updated_at": "2025-09-15T08:57:23Z"
1967 }
1968}
1969```
1970
1971## HTTP headers
1972
1973Each webhook request includes the following headers:
1974
1975- `Content-Type: application/json`
1976- `User-Agent: Tangled-Hook/<short-sha>` — User agent with short SHA of the commit
1977- `X-Tangled-Event: push` — The event type
1978- `X-Tangled-Hook-ID: <webhook-id>` — The webhook ID
1979- `X-Tangled-Delivery: <uuid>` — Unique delivery ID
1980- `X-Tangled-Signature-256: sha256=<hmac>` — HMAC-SHA256 signature (if secret configured)
1981
1982## Verifying webhook signatures
1983
1984If you configured a secret, you should verify the webhook signature to ensure requests are authentic. For example, in Go:
1985
1986```go
1987package main
1988
1989import (
1990 "crypto/hmac"
1991 "crypto/sha256"
1992 "encoding/hex"
1993 "io"
1994 "net/http"
1995 "strings"
1996)
1997
1998func verifySignature(payload []byte, signatureHeader, secret string) bool {
1999 // Remove 'sha256=' prefix from signature header
2000 signature := strings.TrimPrefix(signatureHeader, "sha256=")
2001
2002 // Compute expected signature
2003 mac := hmac.New(sha256.New, []byte(secret))
2004 mac.Write(payload)
2005 expected := hex.EncodeToString(mac.Sum(nil))
2006
2007 // Use constant-time comparison to prevent timing attacks
2008 return hmac.Equal([]byte(signature), []byte(expected))
2009}
2010
2011func webhookHandler(w http.ResponseWriter, r *http.Request) {
2012 // Read the request body
2013 payload, err := io.ReadAll(r.Body)
2014 if err != nil {
2015 http.Error(w, "Bad request", http.StatusBadRequest)
2016 return
2017 }
2018
2019 // Get signature from header
2020 signatureHeader := r.Header.Get("X-Tangled-Signature-256")
2021
2022 // Verify signature
2023 if signatureHeader != "" && verifySignature(payload, signatureHeader, yourSecret) {
2024 // Webhook is authentic, process it
2025 processWebhook(payload)
2026 w.WriteHeader(http.StatusOK)
2027 } else {
2028 http.Error(w, "Invalid signature", http.StatusUnauthorized)
2029 }
2030}
2031```
2032
2033## Delivery retries
2034
2035Webhooks are automatically retried on failure:
2036
2037- **3 total attempts** (1 initial + 2 retries)
2038- **Exponential backoff** starting at 1 second, max 10 seconds
2039- **Retried on**:
2040 - Network errors
2041 - HTTP 5xx server errors
2042- **Not retried on**:
2043 - HTTP 4xx client errors (bad request, unauthorized, etc.)
2044
2045### Timeouts
2046
2047Webhook requests timeout after 30 seconds. If your endpoint needs more time:
2048
20491. Respond with 200 OK immediately
20502. Process the webhook asynchronously in the background
2051
2052## Example integrations
2053
2054### Discord notifications
2055
2056```javascript
2057app.post("/webhook", (req, res) => {
2058 const payload = req.body;
2059
2060 fetch("https://discord.com/api/webhooks/...", {
2061 method: "POST",
2062 headers: { "Content-Type": "application/json" },
2063 body: JSON.stringify({
2064 content: `New push to ${payload.repository.full_name}`,
2065 embeds: [
2066 {
2067 title: `${payload.pusher.did} pushed to ${payload.ref}`,
2068 url: payload.repository.html_url,
2069 color: 0x00ff00,
2070 },
2071 ],
2072 }),
2073 });
2074
2075 res.status(200).send("OK");
2076});
2077```
2078
2079# Migrating knots and spindles
2080
2081Sometimes, non-backwards compatible changes are made to the
2082knot/spindle XRPC APIs. If you host a knot or a spindle, you
2083will need to follow this guide to upgrade. Typically, this
2084only requires you to deploy the newest version.
2085
2086This document is laid out in reverse-chronological order.
2087Newer migration guides are listed first, and older guides
2088are further down the page.
2089
2090## Upgrading to v1.15.0-alpha
2091
2092With v1.15.0-alpha, a knot itself owns its members and
2093per-repo collaborators directly. Previously this data was sourced from
2094PDS records (`sh.tangled.knot.member` and `sh.tangled.repo.collaborator`)
2095that the appview and the knot both read off the firehose.
2096The knot is now the source of truth and serves them over XRPC instead:
2097
2098- `sh.tangled.knot.addMember`, `sh.tangled.knot.removeMember`, `sh.tangled.knot.listMembers`
2099- `sh.tangled.repo.addCollaborator`, `sh.tangled.repo.removeCollaborator`, `sh.tangled.repo.listCollaborators`
2100
2101Until your knot is upgraded, the appview keeps reading its
2102members and collaborators from the old firehose-sourced records.
2103Upgrade to move your knot onto knot-owned access control.
2104
2105- Upgrade to the latest tag (v1.15.0 or above)
2106- Head to the [knot dashboard](https://tangled.org/settings/knots) and
2107 hit the "retry" button to verify your knot
2108
2109## Upgrading to v1.14.0-alpha
2110
2111Starting with v1.14.0-alpha, the fully knot uses the repoDID as its
2112canonical handle for repositories. This unlocks repository
2113renames from the appview UI and changes the wire format for
2114the following lexicons (`sh.tangled.repo.pull`, `sh.tangled.repo.collaborator`,
2115`sh.tangled.repo.issue`, `sh.tangled.git.refUpdate`).
2116
2117Knots that have not been upgraded may silently drop new push
2118events, pull requests, issues, and collaborator invites for
2119repositories they host until upgraded. So upgrade please!!!
2120
2121- Upgrade to the latest tag (v1.14.0 or above)
2122- Head to the [knot dashboard](https://tangled.org/settings/knots) and
2123 hit the "retry" button to verify your knot
2124
2125## Upgrading to v1.13.0-alpha
2126
2127Starting with v1.13.0-alpha, every repository on a knot is
2128assigned a DID. This makes repositories stable across
2129renames and transfers.
2130
2131When you upgrade your knot to this version, the server will
2132automatically mint DIDs for all existing repositories on
2133startup. This is a one-time process and you may see
2134additional log output during the first boot as DIDs are
2135assigned.
2136
2137- Upgrade to the latest tag (v1.13.0 or above)
2138- Head to the [knot dashboard](https://tangled.org/settings/knots) and
2139 hit the "retry" button to verify your knot
2140
2141## Upgrading from v1.8.x
2142
2143After v1.8.2, the HTTP API for knots and spindles has been
2144deprecated and replaced with XRPC. Repositories on outdated
2145knots will not be viewable from the appview. Upgrading is
2146straightforward however.
2147
2148For knots:
2149
2150- Upgrade to the latest tag (v1.9.0 or above)
2151- Head to the [knot dashboard](https://tangled.org/settings/knots) and
2152 hit the "retry" button to verify your knot
2153
2154For spindles:
2155
2156- Upgrade to the latest tag (v1.9.0 or above)
2157- Head to the [spindle
2158 dashboard](https://tangled.org/settings/spindles) and hit the
2159 "retry" button to verify your spindle
2160
2161## Upgrading from v1.7.x
2162
2163After v1.7.0, knot secrets have been deprecated. You no
2164longer need a secret from the appview to run a knot. All
2165authorized commands to knots are managed via [Inter-Service
2166Authentication](https://atproto.com/specs/xrpc#inter-service-authentication-jwt).
2167Knots will be read-only until upgraded.
2168
2169Upgrading is quite easy, in essence:
2170
2171- `KNOT_SERVER_SECRET` is no more, you can remove this
2172 environment variable entirely
2173- `KNOT_SERVER_OWNER` is now required on boot, set this to
2174 your DID. You can find your DID in the
2175 [settings](https://tangled.org/settings) page.
2176- Restart your knot once you have replaced the environment
2177 variable
2178- Head to the [knot dashboard](https://tangled.org/settings/knots) and
2179 hit the "retry" button to verify your knot. This simply
2180 writes a `sh.tangled.knot` record to your PDS.
2181
2182If you use the nix module, simply bump the flake to the
2183latest revision, and change your config block like so:
2184
2185```diff
2186 services.tangled.knot = {
2187 enable = true;
2188 server = {
2189- secretFile = /path/to/secret;
2190+ owner = "did:plc:foo";
2191 };
2192 };
2193```
2194
2195# Bobbin
2196
2197Bobbin is an API appview for Tangled records. It serves XRPC
2198endpoints for `sh.tangled.*`, with it you can get repos,
2199issues, pulls, comments, follows, stars, labels, pipelines,
2200and profiles. It is read-only, there is no auth, since that
2201should all be handled direct-to-PDS and knot respectively.
2202
2203**Bobbin has no permanent storage**.
2204
2205It is only a glorified edge index, in the graph theory
2206sense. Additionally it has a record cache, re-filled on
2207demand. All other data that Bobbin serves comes live from
2208PDSes & knots.
2209
2210## What Bobbin needs
2211
2212The way that Bobbin is able to pull off being
2213so stateless is by moving state upstream.
2214Primarily it depends on an instance of
2215[Hydrant](https://tangled.org/did:plc:6v3ul2ptnqctyxwkz5ti4amn)
2216, which is the service that gives an event stream
2217for Bobbin to quickly backfill from on every restart.
2218Backfilling ought to take less than a couple of minutes
2219maximum. If the upstream instance of Hydrant fails
2220while Bobbin is live, its list/count endpoints stop
2221advancing and report a stale cursor. Single-lookups
2222will continue working, due to the second dependency:
2223[Slingshot](https://tangled.org/did:plc:c7mc2fn47ihdihul4vjwsuy3/tree/main/slingshot).
2224Slingshot fetches individual records & resolves identities.
2225If the upstream instance of Slingshot fails, single-lookups
2226will fail with a `502` error. There are some aggregation
2227endpoints that use Slingshot for hydrating, which will also
2228fail.
2229
2230A soft dependency that ought to exist for Bobbin to operate
2231correctly is simply the plethora of knots that are out
2232there, that Bobbin talks to directly for git data and, for
2233knots at v1.15+, members & collaborators.
2234
2235## Building Bobbin
2236
2237Bobbin is under [Tangled's core monorepo, under bobbin/](https://tangled.org/did:plc:j5hmlfdrwkvtxm7cjmu7j2is/tree/master/bobbin).
2238Here's an easy local debug-build:
2239
2240```sh
2241cargo build -p bobbin
2242```
2243
2244Bobbin loves being in a container. When using
2245`bobbin/containerfiles/bobbin.Containerfile`, it runs `cargo
2246build --release --bin bobbin --package bobbin` within a
2247little Debian runtime, exposing port 8090.
2248
2249## Configuration
2250
2251The best way to configure Bobbin is via a toml config file.
2252There's an `example.toml` in [Bobbin's subdir](https://tangled.org/did:plc:j5hmlfdrwkvtxm7cjmu7j2is/blob/master/bobbin/example.toml).
2253Every value is overridable by a `BOBBIN_*` env var.
2254The load order is env, then `--config <path>`, then
2255`/etc/bobbin/config.toml`, then built-in defaults.
2256
2257Load and check a config without starting the server:
2258
2259```sh
2260bobbin --config config.toml validate
2261```
2262
2263Minimal config is the two upstream URLs. The hydrant URL
2264takes `ws://` or `wss://`. An `http://` or `https://`
2265URL is rewritten to the matching websocket scheme at
2266connection-time.
2267
2268```toml
2269[server]
2270binds = ["127.0.0.1:8090"]
2271
2272# Loopback-only & can leave empty to disable debug introspection.
2273debug_bind = "127.0.0.1:8091"
2274
2275[hydrant]
2276url = "https://hydrant.example.com"
2277
2278[slingshot]
2279url = "https://slingshot.example.com"
2280```
2281
2282> 🦪 Lewis
2283>
2284> At time of writing, we (Tangled) don't host public
2285> instances of Hydrant or Slingshot. You will have to
2286> find public instances or spin these up yourself! :P
2287
2288Take a gander in the project's example.toml for an
2289exhaustive list of things to configure.
2290
2291You will discover fun things such as a configurable adaptive
2292loop that watches the cgroup memory limit & throttles heavy
2293requests under pressure. It only works if it detects a
2294cgroup limit is present. The config for that is in the
2295`[backpressure]` block of the config template.
2296
2297## Running Bobbin
2298
2299Start the server using a config toml:
2300
2301```bash
2302bobbin --config config.toml
2303```
2304Bobbin wakes up in a cold sweat and immediately gets to
2305work:
23061. It binds its listeners, connects to the Hydrant stream
2307 in the background.
23082. It serves requests from the first
2309 moment it's alive, even before the Hydrant stream connects
2310 or finishes catching up. Having a cold Hydrant itself
2311 costs only latency and approximate counts.
2312
2313## The API
2314
2315**Single lookups** take a record's AT-URI.
2316
2317- `getRepo` takes the repo URI:
2318
2319```sh
2320curl "$BOBBIN/xrpc/sh.tangled.repo.getRepo?repo=at://did:plc:boltless/sh.tangled.repo/squid"
2321```
2322```json
2323{
2324 "uri": "at://did:plc:boltless/sh.tangled.repo/squid",
2325 "cid": "bafyrei...",
2326 "value": { "$type": "sh.tangled.repo", "knot": "knot1.tangled.sh", "description": "...", "createdAt": "..." }
2327}
2328```
2329
2330- `getProfile` takes the full profile record URI, so a bare
2331 handle or DID will not resolve:
2332
2333```sh
2334curl "$BOBBIN/xrpc/sh.tangled.actor.getProfile?actor=at://did:plc:boltless/sh.tangled.actor.profile/self"
2335```
2336
2337- If Slingshot cannot serve the record, the response is `502`:
2338
2339```json
2340{ "error": "UpstreamFailed", "message": "upstream unavailable: ..." }
2341```
2342
2343**Aggregation** endpoints come in `list*` and `count*` pairs,
2344each with a `*By` sibling, and require a `subject` query param.
2345
2346- `listRepos` and `countRepos` key on the owner DID:
2347
2348```sh
2349curl "$BOBBIN/xrpc/sh.tangled.repo.countRepos?subject=did:plc:boltless"
2350```
2351```json
2352{ "count": 7, "distinctAuthors": 1 }
2353```
2354
2355```sh
2356curl "$BOBBIN/xrpc/sh.tangled.repo.listRepos?subject=did:plc:boltless&limit=3"
2357```
2358```json
2359{ "items": [ { "uri": "at://did:plc:boltless/sh.tangled.repo/squid", "cid": "bafyrei...", "value": { } } ], "cursor": null }
2360```
2361
2362- Bobbin validates the subject per collection. Here a repo URI
2363 is passed where a bare DID is required, so the call returns a
2364 `400`:
2365
2366```sh
2367curl "$BOBBIN/xrpc/sh.tangled.graph.listFollows?subject=at://did:plc:boltless/sh.tangled.repo/squid"
2368```
2369```json
2370{ "error": "InvalidRequest", "message": "invalid request: subject must be a bare did, got at-uri with collection sh.tangled.repo" }
2371```
2372
2373**Search** is a single endpoint over an in-mem full-text
2374index:
2375
2376```sh
2377curl "$BOBBIN/xrpc/sh.tangled.search.query?q=tangled&limit=2"
2378```
2379```json
2380{ "hits": [ { "uri": "at://...", "cid": "...", "nsid": "sh.tangled.repo", "score": 27.1, "value": { } } ], "cursor": null }
2381```
2382
2383**Git data** such as blob, tree, diff, log, and archive proxies
2384straight to the repo's knot, streamed back without caching.
2385
2386## Coverage and warm-up
2387
2388- While the edge index is catching up from Hydrant,
2389 the aggregation count is a lower bound & may still climb.
2390- One endpoint reports how far along the backfill it is:
2391
2392```sh
2393curl "$BOBBIN/xrpc/sh.tangled.bobbin.getCoverage"
2394```
2395
2396While warming up:
2397
2398```json
2399{ "ready": false, "eventsProcessed": 45588, "lastCursor": 51658 }
2400```
2401
2402Once caught up, Bobbin flips to ready:
2403
2404```json
2405{ "ready": true, "eventsProcessed": 106085, "lastCursor": 116527 }
2406```
2407
2408If starting up Hydrant for the first time, Hydrant itself
2409will take a decent while (a couple of hours) to backfill
2410from PDSes. Hydrant stores its backfill on disk. Bobbin
2411restart reaches `ready` in minutes by replaying event from
2412an already-populated Hydrant. If your Hydrant is new, expect
2413Bobbin to backfill in that same couple of hours that Hydrant
2414takes.
2415
2416## Loose ends and not-gonna-impl
2417
2418- **No coverage signal for per-knot rosters yet.**
2419 Coverage tracks the hydrant stream only. A v1.15 knot
2420 that is unreachable serves a stale or empty member set
2421 with nothing to flag it.
2422- **Knot eventstream fan-out isn't pooled.**
2423 Bobbin opens one websocket per v1.15
2424 knot on top of the hydrant subscription. A network with
2425 thousands of knots wants pooling or a shared subscription.
2426- **No sequential issue or PR numbers.** bobbin returns rkeys,
2427 not `#42` style ids like the web appview. A client
2428 deriving a display number does it from creation order. But
2429 why bother? rkeys are the IDs.
2430
2431# Hacking on Tangled
2432
2433We highly recommend [installing
2434Nix](https://nixos.org/download/) (the package manager)
2435before working on the codebase. The Nix flake provides a lot
2436of helpers to get started and most importantly, builds and
2437dev shells are entirely deterministic.
2438
2439To set up your dev environment:
2440
2441```bash
2442nix develop
2443```
2444
2445Non-Nix users can look at the `devShell` attribute in the
2446`flake.nix` file to determine necessary dependencies.
2447
2448## Running the appview
2449
2450The appview requires Redis and OAuth JWKs. Start these
2451first, before launching the appview itself.
2452
2453```bash
2454# OAuth JWKs should already be set up by the Nix devshell:
2455echo $TANGLED_OAUTH_CLIENT_SECRET
2456z42ty4RT1ovnTopY8B8ekz9NuziF2CuMkZ7rbRFpAR9jBqMc
2457
2458echo $TANGLED_OAUTH_CLIENT_KID
24591761667908
2460
2461# if not, you can set it up yourself:
2462goat key generate -t P-256
2463Key Type: P-256 / secp256r1 / ES256 private key
2464Secret Key (Multibase Syntax): save this securely (eg, add to password manager)
2465 z42tuPDKRfM2mz2Kv953ARen2jmrPA8S9LX9tRq4RVcUMwwL
2466Public Key (DID Key Syntax): share or publish this (eg, in DID document)
2467 did:key:zDnaeUBxtG6Xuv3ATJE4GaWeyXM3jyamJsZw3bSPpxx4bNXDR
2468
2469# the secret key from above
2470export TANGLED_OAUTH_CLIENT_SECRET="z42tuP..."
2471
2472# Run Redis in a new shell to store OAuth sessions
2473redis-server
2474```
2475
2476The Nix flake exposes a few `app` attributes (run `nix
2477flake show` to see a full list of what the flake provides),
2478one of the apps runs the appview with the `air`
2479live-reloader:
2480
2481```bash
2482TANGLED_DEV=true nix run .#watch-appview
2483
2484# TANGLED_DB_PATH might be of interest to point to
2485# different sqlite DBs
2486
2487# in a separate shell, you can live-reload tailwind
2488nix run .#watch-tailwind
2489```
2490
2491## Running knots and spindles
2492
2493An end-to-end knot setup requires setting up a machine with
2494`sshd`, `AuthorizedKeysCommand`, and a Git user, which is
2495quite cumbersome. So the Nix flake provides a
2496`nixosConfiguration` to do so.
2497
2498<details>
2499 <summary><strong>macOS users will have to set up a Nix Builder first</strong></summary>
2500
2501In order to build Tangled's dev VM on macOS, you will
2502first need to set up a Linux Nix builder. The recommended
2503way to do so is to run a [`darwin.linux-builder`
2504VM](https://nixos.org/manual/nixpkgs/unstable/#sec-darwin-builder)
2505and to register it in `nix.conf` as a builder for Linux
2506with the same architecture as your Mac (`linux-aarch64` if
2507you are using Apple Silicon).
2508
2509If you're on nix-darwin, you can simply add
2510
2511```
2512nix.linux-builder.enable = true;
2513```
2514
2515to your host's `configuration.nix`.
2516
2517Alternatively, you can use any other method to set up a
2518Linux machine with Nix installed that you can `sudo ssh`
2519into (in other words, root user on your Mac has to be able
2520to ssh into the Linux machine without entering a password)
2521and that has the same architecture as your Mac. See
2522[remote builder
2523instructions](https://nix.dev/manual/nix/2.28/advanced-topics/distributed-builds.html#requirements)
2524for how to register such a builder in `nix.conf`.
2525
2526> WARNING: If you'd like to use
2527> [`nixos-lima`](https://github.com/nixos-lima/nixos-lima) or
2528> [Orbstack](https://orbstack.dev/), note that setting them up so that `sudo
2529ssh` works can be tricky. It seems to be [possible with
2530> Orbstack](https://github.com/orgs/orbstack/discussions/1669).
2531
2532</details>
2533
2534To begin, grab your DID from http://localhost:3000/settings.
2535Then, set `TANGLED_VM_KNOT_OWNER` and
2536`TANGLED_VM_SPINDLE_OWNER` to your DID. You can now start a
2537lightweight NixOS VM like so:
2538
2539```bash
2540nix run --impure .#vm
2541
2542# type `poweroff` at the shell to exit the VM
2543```
2544
2545This starts a knot on port 6444, a spindle on port 6555
2546with `ssh` exposed on port 2222.
2547
2548Once the services are running, head to
2549http://localhost:3000/settings/knots and hit "Verify". It should
2550verify the ownership of the services instantly if everything
2551went smoothly.
2552
2553You can push repositories to this VM with this ssh config
2554block on your main machine:
2555
2556```bash
2557Host nixos-shell
2558 Hostname localhost
2559 Port 2222
2560 User git
2561 IdentityFile ~/.ssh/my_tangled_key
2562```
2563
2564Set up a remote called `local-dev` on a git repo:
2565
2566```bash
2567git remote add local-dev git@nixos-shell:user/repo
2568git push local-dev main
2569```
2570
2571The above VM should already be running a spindle on
2572`localhost:6555`. Head to http://localhost:3000/settings/spindles and
2573hit "Verify". You can then configure each repository to use
2574this spindle and run CI jobs.
2575
2576Of interest when debugging spindles:
2577
2578```
2579# Service logs from journald:
2580journalctl -xeu spindle
2581
2582# CI job logs from disk:
2583ls /var/log/spindle
2584
2585# Debugging spindle database:
2586sqlite3 /var/lib/spindle/spindle.db
2587
2588# litecli has a nicer REPL interface:
2589litecli /var/lib/spindle/spindle.db
2590```
2591
2592If for any reason you wish to disable either one of the
2593services in the VM, modify [nix/vm.nix](/nix/vm.nix) and set
2594`services.tangled.spindle.enable` (or
2595`services.tangled.knot.enable`) to `false`.
2596
2597# Contribution guide
2598
2599## Commit guidelines
2600
2601We follow a commit style similar to the Go project. Please keep commits:
2602
2603- **atomic**: each commit should represent one logical change
2604- **descriptive**: the commit message should clearly describe what the
2605 change does and why it's needed
2606
2607### Message format
2608
2609```
2610<service/top-level directory>/<affected package/directory>: <short summary of change>
2611
2612Optional longer description can go here, if necessary. Explain what the
2613change does and why, especially if not obvious. Reference relevant
2614issues or PRs when applicable. These can be links for now since we don't
2615auto-link issues/PRs yet.
2616```
2617
2618Here are some examples:
2619
2620```
2621appview/state: fix token expiry check in middleware
2622
2623The previous check did not account for clock drift, leading to premature
2624token invalidation.
2625```
2626
2627```
2628knotserver/git/service: improve error checking in upload-pack
2629```
2630
2631### General notes
2632
2633- PRs get merged "as-is" (fast-forward)—like applying a patch-series
2634 using `git am`. At present, there is no squashing—so please author
2635 your commits as they would appear on `master`, following the above
2636 guidelines.
2637- If there is a lot of nesting, for example "appview:
2638 pages/templates/repo/fragments: ...", these can be truncated down to
2639 just "appview: repo/fragments: ...". If the change affects a lot of
2640 subdirectories, you may abbreviate to just the top-level names, e.g.
2641 "appview: ..." or "knotserver: ...".
2642- Keep commits lowercased with no trailing period.
2643- Use the imperative mood in the summary line (e.g., "fix bug" not
2644 "fixed bug" or "fixes bug").
2645- Try to keep the summary line under 72 characters, but we aren't too
2646 fussed about this.
2647- Follow the same formatting for PR titles if filled manually.
2648- Don't include unrelated changes in the same commit.
2649- Avoid noisy commit messages like "wip" or "final fix"—rewrite history
2650 before submitting if necessary.
2651
2652## Code formatting
2653
2654We use a variety of tools to format our code, and multiplex them with
2655[`treefmt`](https://treefmt.com). All you need to do to format your changes
2656is run `nix run .#fmt` (or just `treefmt` if you're in the devshell).
2657
2658## Proposals for bigger changes
2659
2660Small fixes like typos, minor bugs, or trivial refactors can be
2661submitted directly as PRs.
2662
2663For larger changes—especially those introducing new features, significant
2664refactoring, or altering system behavior—please open a proposal first. This
2665helps us evaluate the scope, design, and potential impact before implementation.
2666
2667Create a new issue titled:
2668
2669```
2670proposal: <affected scope>: <summary of change>
2671```
2672
2673In the description, explain:
2674
2675- What the change is
2676- Why it's needed
2677- How you plan to implement it (roughly)
2678- Any open questions or tradeoffs
2679
2680We'll use the issue thread to discuss and refine the idea before moving
2681forward.
2682
2683## Developer Certificate of Origin (DCO)
2684
2685We require all contributors to certify that they have the right to
2686submit the code they're contributing. To do this, we follow the
2687[Developer Certificate of Origin
2688(DCO)](https://developercertificate.org/).
2689
2690By signing your commits, you're stating that the contribution is your
2691own work, or that you have the right to submit it under the project's
2692license. This helps us keep things clean and legally sound.
2693
2694To sign your commit, just add the `-s` flag when committing:
2695
2696```sh
2697git commit -s -m "your commit message"
2698```
2699
2700This appends a line like:
2701
2702```
2703Signed-off-by: Your Name <your.email@example.com>
2704```
2705
2706We won't merge commits if they aren't signed off. If you forget, you can
2707amend the last commit like this:
2708
2709```sh
2710git commit --amend -s
2711```
2712
2713If you're submitting a PR with multiple commits, make sure each one is
2714signed.
2715
2716For [jj](https://jj-vcs.github.io/jj/latest/) users, you can run the following command
2717to make it sign off commits in the tangled repo:
2718
2719```shell
2720# Safety check, should say "No matching config key..."
2721jj config list templates.commit_trailers
2722# The command below may need to be adjusted if the command above returned something.
2723jj config set --repo templates.commit_trailers "format_signed_off_by_trailer(self)"
2724```
2725
2726Refer to the [jujutsu
2727documentation](https://jj-vcs.github.io/jj/latest/config/#commit-trailers)
2728for more information.
2729
2730# Troubleshooting guide
2731
2732## Login issues
2733
2734Owing to the distributed nature of OAuth on AT Protocol, you
2735may run into issues with logging in. If you run a
2736self-hosted PDS:
2737
2738- You may need to ensure that your PDS is timesynced using
2739 NTP:
2740 - Enable the `ntpd` service
2741 - Run `ntpd -qg` to synchronize your clock
2742- You may need to increase the default request timeout:
2743 `NODE_OPTIONS="--network-family-autoselection-attempt-timeout=500"`
2744
2745## Empty punchcard
2746
2747For Tangled to register commits that you make across the
2748network, you need to setup one of following:
2749
2750- The committer email should be a verified email associated
2751 to your account. You can add and verify emails on the
2752 settings page.
2753- Or, the committer email should be set to your account's
2754 DID: `git config user.email "did:plc:foobar"`. You can find
2755 your account's DID on the settings page
2756
2757## Commit is not marked as verified
2758
2759Presently, Tangled only supports SSH commit signatures.
2760
2761To sign commits using an SSH key with git:
2762
2763```
2764git config --global gpg.format ssh
2765git config --global user.signingkey ~/.ssh/tangled-key
2766```
2767
2768To sign commits using an SSH key with jj, add this to your
2769config:
2770
2771```
2772[signing]
2773behavior = "own"
2774backend = "ssh"
2775key = "~/.ssh/tangled-key"
2776```
2777
2778## Self-hosted knot issues
2779
2780If you need help troubleshooting a self-hosted knot, check
2781out the [knot troubleshooting
2782guide](/knot-self-hosting-guide.html#troubleshooting).