docs: simplify lifecycle sections for the converge model

The apply/converge + gateway refactors let the docs collapse a lot:

- design.md §Component Lifecycle: the build→install→deploy three-step
  ("enables and starts services") becomes two moves — develop, then
  `castle apply` (activation polymorphic over the manager). Filesystem
  layout de-duped and corrected (program source is /data/repos/<name>,
  not the long-removed ~/.castle/code).
- registry.md §Lifecycle: three near-identical per-kind flows (service/
  tool/job) collapse into one flow + a "what apply does per manager"
  table. Manager table's "start/stop" column → "how apply activates".
  Fixed stale paths-table rows (code/<name>) and a stale `service run`.
- developing-castle.md endpoint list: `POST /apply` is the one lifecycle
  endpoint; dropped retired `/deploy`; service actions are restart-only;
  added `/config/deployments/{name}/enabled`.

Net −37 lines; no retired-verb or stale-path references remain in docs.
This commit is contained in:
2026-07-02 12:20:33 -07:00
parent 7410680f57
commit 43a6ad76bb
3 changed files with 66 additions and 103 deletions

View File

@@ -416,71 +416,53 @@ Same contract as tools, plus:
## Component Lifecycle
The path from source to managed process:
The path from source to managed process is two moves — **develop** (against the
program) and **converge** (`castle apply`):
```
source → [build] → artifact → [install] → available → [deploy] → managed
source → [dev verbs: build/test/…] → config (programs/ + deployments/) → [castle apply] → running
```
Each step is distinct:
- **Develop** — language-specific dev verbs (`build`, `test`, …) that Castle
records but never runs implicitly. Frontends build in place under the repo
(`<source>/<dist>/`), served from there — no copy step.
- **Converge** — `castle apply` reads the config, generates systemd units +
Caddyfile entries, then reconciles reality to the desired state: activate what's
enabled, restart what changed, deactivate what's disabled. Activation is
*polymorphic over the manager* — systemd `enable --now`, `uv tool install` for a
tool on PATH, a gateway route for a static — so there is no separate install or
start step, and no per-kind verb. `castle apply --plan` shows the diff first.
1. **Build** — Language-specific. Produces an artifact (binary, venv
entry point, static bundle). Castle records the commands but doesn't
execute them implicitly.
2. **Install** — Makes the artifact available on the system. For tools:
`uv tool install` or compiled binary placed in `~/.local/bin/`. For
services: same — the binary or entry point is on PATH or in a known
location. For frontends: assets built in place under the repo
(`<source>/<dist>/`), served from there — no copy.
3. **Deploy** — Materializes the runtime configuration. Reads the
program spec, merges with node config, generates systemd units
and Caddyfile entries that reference *installed* artifacts — never
the source tree. Enables and starts services.
For compiled languages (Rust, Go), build produces a standalone binary
and install is just placing it in `~/.local/bin/`. For interpreted
languages (Python, Node), the runtime wrapper (uv, node) handles
finding the installed artifact.
Desired on/off is `enabled` on the deployment; the only way to durably stop
something is `enabled: false` + apply. `castle restart` is the one imperative
bounce that re-actualizes current state without changing it.
## Runtime Filesystem Layout
Two roots, each overridable by an env var: `$CASTLE_HOME` (config, code,
artifacts, secrets; default `~/.castle`) and `$CASTLE_DATA_DIR` (bulk program
data; default `/data/castle`, kept on a dedicated volume):
Two roots, each overridable by an env var: `$CASTLE_HOME` (config, artifacts,
secrets; default `~/.castle`) and `$CASTLE_DATA_DIR` (bulk program data; default
`/data/castle`, on a dedicated volume). Program source lives separately under
`/data/repos/<name>/` (`$CASTLE_REPOS_DIR`).
```
$CASTLE_HOME/ ← Config & artifacts (default ~/.castle)
├── castle.yaml ← Global settings; resources under programs/ + deployments/
├── castle.yaml ← Global settings (gateway, repo, agents)
├── programs/ deployments/ ← One YAML file per program / deployment
├── infra.conf ← Infrastructure install choices
├── code/ ← Program source (your programs)
── <name>/
├── artifacts/
│ ├── specs/ ← Generated by `castle apply`
│ │ ├── Caddyfile
│ │ └── registry.yaml ← Node config (what's deployed here)
│ └── content/ ← Built frontend assets
│ └── castle/ ← (index.html + assets, served at root)
├── artifacts/specs/ ← Generated by `castle apply`
── Caddyfile
│ └── registry.yaml ← Node config (what's deployed here)
└── secrets/ ← Secret files (NAME → value)
└── PROTONMAIL_API_KEY
$CASTLE_DATA_DIR/ ← Persistent service data (default /data/castle)
└── <name>/
~/.config/systemd/user/ ← Systemd units (standard location)
├── castle-central-context.service
├── castle-protonmail.service
├── castle-protonmail.timer
└── ...
/data/repos/<name>/ ← Program source (your programs; absolute source:)
$CASTLE_DATA_DIR/<name>/ ← Persistent service data (default /data/castle)
~/.config/systemd/user/ ← Systemd units + timers (castle-*.service/.timer)
```
Compiled-language tools (Rust, Go — planned) install their binaries to the
standard `~/.local/bin/`, not under `$CASTLE_HOME`.
Source (the repo) is referenced only during build and install. Everything
the runtime touches lives under `$CASTLE_HOME`, `$CASTLE_DATA_DIR`, or standard
systemd paths.
standard `~/.local/bin/`. Source is referenced only by dev verbs and while a
deployment is materialized; everything the runtime touches lives under
`$CASTLE_HOME`, `$CASTLE_DATA_DIR`, or standard systemd paths.
## OTP as Design Guide

View File

@@ -47,15 +47,17 @@ for tools/libraries.
## castle-api endpoints (port 9020)
- Core: `GET /health`, `GET /stream` (SSE: health, service-action, mesh)
- Converge: `POST /apply` (`{name?, plan?}`) — the one lifecycle endpoint
- Deployments: `GET /deployments[/{name}]`, `GET /status`
- Catalog + editing: `GET /programs[/{name}]`, `POST /programs/{name}/{action}`,
`PUT|DELETE /programs/{name}`; likewise `/services`, `/jobs`
- Config: `GET|PUT /`, `POST /apply`, `POST /deploy`
- Gateway: `GET /gateway`, `GET /gateway/caddyfile` (reload is convergence: `POST /apply`)
- Catalog + editing: `GET /programs[/{name}]`, `POST /programs/{name}/{action}`
(dev verbs only), `PUT|DELETE /programs/{name}`; likewise `/services`, `/jobs`
- Config: `GET|PUT /`, `POST /config/apply`, `PUT /config/deployments/{name}/enabled`
- Gateway: `GET /gateway`, `GET /gateway/caddyfile`, `PUT /gateway/config` (making
routes live is convergence: `POST /apply`)
- Mesh: `GET /mesh/status`, `GET /nodes[/{hostname}]`
- Agents (dashboard terminal UX): `GET /agents`, `GET /agents/sessions`,
`GET /agents/history`, `DELETE /agents/sessions/{id}`, `WS /agents/{name}/session`
- Service actions: `POST /services/{name}/{action}`, `GET /services/{name}/unit`
- Service actions: `POST /services/{name}/restart`, `GET /services/{name}/unit`
## Infrastructure internals

View File

@@ -225,14 +225,14 @@ declares a **`manager`** — who makes it available and supervises its lifecycle
### `manager` — Who realizes it (the discriminant)
A deployment is a *managed materialization* of a program. Its **`manager`** is
the stored discriminant — the single axis lifecycle, deploy, and status all
dispatch on:
the stored discriminant — the single axis `castle apply` and status dispatch on
(it's what makes activation polymorphic: one verb, kind-specific mechanism):
| Manager | Makes available as | Launch mechanism | start/stop | Kind |
|---------|--------------------|------------------|------------|------|
| **systemd** | a running process (or a `.timer` for jobs) | nested `run: { launcher: … }` | `systemctl` | service / job |
| **caddy** | a gateway static file_server route | *(none — files on disk; `root:`)* | add/remove route + reload | static |
| **path** | an installed CLI on `PATH` | *(none — `uv tool install`)* | `uv tool install` / `uninstall` | tool |
| Manager | Makes available as | Launch mechanism | how `apply` activates | Kind |
|---------|--------------------|------------------|-----------------------|------|
| **systemd** | a running process (or a `.timer` for jobs) | nested `run: { launcher: … }` | `systemctl enable --now` | service / job |
| **caddy** | a gateway static file_server route | *(none — files on disk; `root:`)* | wire the route + reload | static |
| **path** | an installed CLI on `PATH` | *(none — `uv tool install`)* | `uv tool install` | tool |
| **none** | an external reference | *(none; `base_url:`/`health_url:`)* | *(nothing — not ours)* | reference |
The **kind** (service/job/tool/static/reference) is **derived** from `manager` (+
@@ -630,55 +630,34 @@ manage:
## Lifecycle
### Service lifecycle
One flow for every kind — scaffold, implement, **`castle apply`**. Activation is
polymorphic over the `manager`, so the *verb* never changes; only what apply *does*
does:
```bash
castle program create my-service --stack python-fastapi # 1. Scaffold + register
cd /data/repos/my-service && uv sync # 2. Install deps
# ... implement ...
castle program test my-service # 3. Run tests
castle apply my-service # 4. Render unit + routes and reconcile (start it)
castle program create my-thing --stack python-fastapi # scaffold source + deployment
cd /data/repos/my-thing && uv sync # implement
castle program test my-thing # dev verbs (build/test/lint/…)
castle apply my-thing # converge: render + activate
```
`castle apply` renders the systemd unit + Caddy route and starts the service
(which then runs on boot and restarts on failure). Manage with:
| kind (manager) | what `castle apply` does |
|----------------|--------------------------|
| **service** (systemd) | render the `.service` unit + gateway route, `enable --now` |
| **job** (systemd + schedule) | render a `.service` (Type=oneshot) **and** a `.timer` |
| **tool** (path) | `uv tool install` — put the executable on PATH |
| **static** (caddy) | wire the gateway `file_server` route to `<source>/<root>` |
Manage a running deployment:
```bash
castle logs my-service -f # Tail logs
castle service run my-service # Run in foreground (for debugging)
castle service restart my-service # Imperative bounce
# To stop it durably: set `enabled: false` in its deployment, then `castle apply`
castle logs my-thing -f # Tail logs
castle program run my-thing # Run in foreground (for debugging)
castle restart my-thing # Imperative bounce — re-actualize current state
```
### Tool lifecycle
```bash
castle program create my-tool --stack python-cli # 1. Scaffold + register
cd /data/repos/my-tool && uv sync # 2. Install deps
# ... implement ...
castle program test my-tool # 3. Run tests
castle apply my-tool # 4. Install the path deployment on PATH
```
### Job lifecycle
Jobs are deployments with `manager: systemd` plus a `schedule` — a
`deployments/my-job.yaml` file with a nested `run:` launch block:
```yaml
# deployments/my-job.yaml (manager: systemd + schedule → kind: job)
program: my-job
manager: systemd
run:
launcher: command
argv: ["my-job"]
schedule: "0 2 * * *"
manage:
systemd: {}
```
`castle apply my-job` generates both a `.service` (Type=oneshot)
and a `.timer` file.
To durably turn something off, set `enabled: false` in its deployment and
`castle apply` — there is no start/stop/enable/disable/install verb.
## Infrastructure paths
@@ -693,8 +672,8 @@ variable (both expand `~` and resolve relative paths):
| What | Where |
|------|-------|
| Castle home | `$CASTLE_HOME` (default `~/.castle`) |
| Registry | `$CASTLE_HOME/castle.yaml` |
| Program source (yours) | `$CASTLE_HOME/code/<name>/` |
| Config | `$CASTLE_HOME/castle.yaml` + `programs/` + `deployments/` |
| Program source (yours) | `/data/repos/<name>/` (`$CASTLE_REPOS_DIR`; absolute `source:`) |
| Program source (castle's) | `<repo>/<name>` (via `source: repo:<name>`) |
| Secrets | `$CASTLE_HOME/secrets/<NAME>` |
| Generated Caddyfile | `$CASTLE_HOME/artifacts/specs/Caddyfile` |