Release v0.3.2

.github/workflows/rust.yml

@@ -138,6 +138,9 @@ jobs:
         run: cargo nextest run --partition count:${{ matrix.shard }}/${{ matrix.shards }} -E 'not binary(bdd)'
       # The `bdd` test uses a custom harness (harness = false), which nextest
       # cannot run, so execute it once on shard 1 with the standard test runner.
+      - name: Prepare config files for BDD tests
+        if: ${{ matrix.shard == 1 }}
+        run: cp access_control.conf.dist access_control.conf
       - name: Run BDD suite
         if: ${{ matrix.shard == 1 }}
         run: cargo test --test bdd

.gitignore

@@ -1,7 +1,6 @@
 target
 .idea/
 files
-access_control.conf
 configuration.yaml
 configuration.yaml.backup
 configuration.yaml.orig
@@ -17,3 +16,5 @@ plan/
 .stacker
 done.txt
 post-deploy-ran.txt
+
+bake/

CHANGELOG.md

@@ -4,6 +4,102 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+## [0.3.2] — 2026-08-26
+
+### Added — Container-health monitoring & alerts (`stacker monitor`)
+
+- New `stacker monitor` command: watches the deployment's live container health
+  (via the Status Panel agent) and raises an alert when a container is not
+  running, then again when everything recovers. Edge-triggered — it notifies
+  once per transition, not on every poll.
+- `stacker monitor --once` runs a single check and exits (cron-friendly);
+  without it, the command loops on an interval. `--interval <seconds>` overrides
+  the configured poll interval.
+- New `monitoring.alerts` block in `stacker.yml` configures the alarm:
+
+  ```yaml
+  monitoring:
+    alerts:
+      interval: 60          # poll interval, seconds (default 60)
+      on_recovery: true     # also notify when containers recover (default true)
+      target:
+        terminal: true                                   # terminal + desktop notification
+        # url: "https://ntfy.example.com/alerts"         # or: HTTP webhook
+        # method: POST
+        # pipe: oncall-notify                            # or: run a declared pipe (deferred)
+  ```
+
+- Alert state is persisted to `.stacker/monitor.state`, so `--once` invocations
+  stay edge-triggered across cron runs.
+- The alarm logic lives in a new, dependency-free `health-monitor` crate (pure
+  health evaluation + transition detection), reusable by the CLI and, later, the
+  agent.
+
+### Added — Declarative pipes (Infrastructure-as-Code)
+
+- New `pipes:` block in `stacker.yml` declares pipes as committable config
+  (name, source/target, endpoints, fields, trigger, retry, handlers) — the
+  same surface as the `pipe create` flags, now version-controllable.
+- `stacker pipe diff` compares the declared `pipes:` against what is deployed
+  and prints a plan: `create` (declared, not deployed), `update` (differs, with
+  the exact field changes), `unchanged`, and `orphan` (deployed, not declared).
+  `--json` for scripting; a clean tree reports "In sync".
+- `stacker pipe apply` reconciles the declaration into the deployment: creates
+  declared-but-missing pipes (template + instance) and, with `--prune`, deletes
+  deployed pipes not in `stacker.yml`. `--dry-run` shows the plan without acting;
+  re-running is idempotent.
+
+### Added — Manual pipe endpoints & non-interactive `pipe create`
+
+- `stacker pipe create` now accepts explicit endpoints, bypassing endpoint
+  discovery entirely: `--source-endpoint "METHOD /path"`,
+  `--target-endpoint "METHOD /path"`, `--source-fields`, `--target-fields`, and
+  `--name`. This lets any app (or arbitrary HTTP endpoint) be piped — including
+  apps whose APIs aren't at auto-discoverable paths — and makes pipe creation
+  fully scriptable.
+
+### Added — Pipe resilience: retry policy & lifecycle handlers
+
+- `stacker pipe create` gained `--retry <n>`, `--retry-backoff-ms <base>`,
+  `--retry-backoff-max-ms <max>`, `--on-failure <pipe>`, and
+  `--on-success <pipe>`. These are persisted in the pipe's typed config
+  (retry policy + handler references) so the runtime can honor them.
+
+### Added — Reverse-proxy routing (traefik, caddy, nginx-proxy-manager)
+
+- `proxy.type` now produces working routing for all supported proxies:
+  Traefik via container labels, Caddy via a generated `Caddyfile`, and
+  Nginx Proxy Manager via auto-created proxy hosts — all driven by
+  `proxy.domains` (`{domain, upstream, ssl}`), which is now forwarded end-to-end
+  to the Install Service.
+- Proxies are **platform-managed**: the synthesized proxy service is stripped
+  from the compose shipped to the server (deployed by its own role instead), so
+  a `proxy:` block no longer double-deploys the proxy or collides on ports
+  80/443/81.
+- For `--target local`, the CLI now renders the proxy config file itself
+  (`.stacker/Caddyfile`) so the bind mount is a real file locally.
+
+### Fixed — Proxy config validation & port-conflict detection
+
+- New `W003` warning: a `proxy:` block plus a service publishing the same
+  ingress host port (80/443/81) is flagged as a likely conflict.
+- Fixed `W001` host-port comparison to correctly handle the
+  `ip:host:container` binding form (previously produced false positives and
+  false negatives on loopback bindings).
+
+### Fixed — SSH backup key saved during provisioning
+
+- The local emergency SSH backup keypair is now saved the moment the server
+  first appears during a cloud deploy's watch loop, instead of only after the
+  watch completes. Interrupting the watch (timeout / Ctrl-C / network) no longer
+  loses SSH access to a successfully-deployed server.
+
+### Fixed — Remote compose cleanup
+
+- Stripping a platform-managed proxy from the remote compose now also prunes the
+  named volumes only that proxy used (e.g. `caddy_data`/`caddy_config`), instead
+  of leaving orphaned top-level volume declarations.
+
 ### Added — Deployment command aliases and `--pinned` flag
 
 - Added `stacker deployment status` as a visible alias for `stacker deployment state`.

Cargo.toml

@@ -1,11 +1,11 @@
 [package]
 name = "stacker"
-version = "0.3.1"
+version = "0.3.2"
 edition = "2021"
 default-run= "server"
 
 [workspace]
-members = ["crates/pipe-adapter-sdk", "crates/pipe-adapter-mail", "crates/td-audit"]
+members = ["crates/pipe-adapter-sdk", "crates/pipe-adapter-mail", "crates/td-audit", "crates/health-monitor"]
 resolver = "2"
 
 [lib]
@@ -39,6 +39,7 @@ config = "0.13.4"
 reqwest = { version = "0.11.23", features = ["json", "blocking", "stream", "native-tls"] }
 serde = { version = "1.0.195", features = ["derive"] }
 td-audit = { path = "crates/td-audit" }
+health-monitor = { path = "crates/health-monitor" }
 tokio = { version = "1.28.1", features = ["full"] }
 tracing = { version = "0.1.40", features = ["log"] }
 tracing-bunyan-formatter = "0.3.8"

README.md

@@ -152,6 +152,10 @@ monitoring:
   healthcheck:
     endpoint: /health
     interval: 30s
+  alerts:                       # container-down alarm for `stacker monitor` (0.3.2)
+    interval: 30
+    target:
+      terminal: true            # or: { url: "https://ntfy.example.com/alerts" }
 ```
 
 Full schema reference: [docs/STACKER_YML_REFERENCE.md](docs/STACKER_YML_REFERENCE.md)
@@ -239,14 +243,17 @@ The end-user tool. No server required for local deploys.
 | `stacker pipe scan` | Discover local endpoints/resources from running containers (when target is `local`) |
 | `stacker pipe scan --containers [filter]` | Discover local endpoints/resources for matching containers |
 | `stacker pipe scan --app <app>` | Probe a remote app for API endpoints |
-| `stacker pipe create <src> <tgt>` | Create a data pipe between two containers (interactive) |
+| `stacker pipe create <src> <tgt>` | Create a data pipe between two containers (interactive; or non-interactive with `--source-endpoint`/`--target-endpoint`/`--name` — added in 0.3.2). `--retry`/`--on-failure`/`--on-success` attach a retry policy + lifecycle handlers |
+| `stacker pipe diff` | Compare the declared `pipes:` block against deployed pipes (added in 0.3.2) |
+| `stacker pipe apply` | Reconcile declared pipes into the deployment; `--prune` removes orphans, `--dry-run` previews (added in 0.3.2) |
 | `stacker pipe list` | List pipe instances for the current deployment |
 | `stacker pipe activate <id>` | Activate a pipe (start listening for triggers) |
 | `stacker pipe deactivate <id>` | Pause an active pipe |
 | `stacker pipe trigger <id>` | One-shot pipe execution with optional input data |
 | `stacker pipe deploy <id>` | Promote a local pipe to a remote deployment |
 | `stacker pipe history <id>` | View execution history for a pipe |
 | `stacker pipe replay <exec-id>` | Re-run a previous pipe execution |
+| `stacker monitor` | Watch container health and alert on problems (added in 0.3.2); `--once` for a single check (cron-friendly). Configure via `monitoring.alerts` in `stacker.yml` |
 | `stacker target [local\|cloud\|server]` | Switch deployment target mode |
 | `stacker env [local\|dev\|prod]` | Show or persist the active deploy environment/profile used by app-only updates |
 | `stacker whoami` | Show the active login, subscription plan, and current project deployment context |
@@ -376,7 +383,9 @@ curl -sL https://marketplace.try.direct/<purchase-token>/install.sh | sh
 - **Agent control** — `stacker agent` subcommand to manage remote Status Panel agents (health, logs, restart, deploy, proxy) with `--json` output
 - **SSH key management** — generate, view, upload, and repair server SSH keys
   (Vault-backed), with automatic local backup SSH access after cloud deploy
-- **Reverse proxy** — auto-detects Nginx / Nginx Proxy Manager, configures domains + SSL
+- **Reverse proxy** — Traefik (labels), Caddy (Caddyfile), and Nginx Proxy Manager, platform-managed and driven by `proxy.domains` end-to-end (0.3.2)
+- **Container-health alarm** — `stacker monitor` watches container health and alerts on problems (terminal, webhook, or pipe), configured via `monitoring.alerts` (0.3.2)
+- **Declarative pipes (IaC)** — declare pipes in `stacker.yml` and reconcile with `stacker pipe diff` / `pipe apply [--prune]` (0.3.2)
 - **Cloud deployment** — Hetzner, DigitalOcean, AWS, Linode, with provider firewall operations and paused/failed install IP retention
 - **MCP Server** — 85+ tools, including deployment, agent control, config, proxy, firewall, and remote service secret management
 - **Marketplace** — submit stacks for review, auto-publish on approval, check status from CLI
@@ -520,6 +529,7 @@ cargo test user_service_client     # User Service connector
 cargo test marketplace_webhook     # Marketplace webhook flows
 cargo test deployment_validator    # Deployment validation
 cargo test --test security_cli     # CLI endpoint IDOR security tests
+SQLX_OFFLINE=true cargo test --lib -- proxy_domains 2>&1 | tail -10
 ```
 
 ---

access_control.conf

@@ -0,0 +1,14 @@
+[request_definition]
+r = sub, obj, act
+
+[policy_definition]
+p = sub, obj, act
+
+[role_definition]
+g = _, _
+
+[policy_effect]
+e = some(where (p.eft == allow))
+
+[matchers]
+m = g(r.sub, p.sub) && keyMatch2(r.obj, p.obj) && r.act == p.act

crates/health-monitor/Cargo.toml

@@ -0,0 +1,12 @@
+[package]
+name = "health-monitor"
+version = "0.1.0"
+edition = "2021"
+description = "Pure container-health alarm engine for TryDirect Stacker: evaluate container health snapshots, edge-detect down/recovery transitions, and format alerts. No I/O, network, or scheduler — the CLI/agent drive it."
+
+[dependencies]
+serde = { version = "1", features = ["derive"] }
+serde_json = "1"
+
+[features]
+default = []