docs: refresh devpi reference card and scrub stale references

- docs/reference/services/devpi.md: rewrite for the new launchd
  layout (no more namespace/PVC/ArgoCD/image/ExternalSecret) and
  link to the new how-to.
- docs/reference/infrastructure/tailscale.md: drop tag:devpi from
  the per-service Ingress tag list.
- docs/reference/storage/backups.md: clarify the devpi-cache row
  to call out the new on-indri path.
- docs/how-to/operations/rebuild-minikube-cluster.md: trim the
  "Devpi cold cache after rebuild" section — devpi is no longer in
  minikube — keep a brief note for the still-possible cold-cache
  race after fresh devpi init.
- docs/how-to/operations/devpi-on-indri.md: correct the Backup
  section — server-dir is NOT in borgmatic_source_directories.
- pulumi/tailscale: remove now-dead tag:devpi ACL rule, tagOwner,
  and __main__.py comment. Will need `pulumi up` to apply.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

docs/how-to/operations/devpi-on-indri.md

@@ -66,7 +66,7 @@ Edit `devpi_server_version` / `devpi_web_version` in `ansible/roles/devpi/defaul
 
 ## Backup
 
-The server-dir is included in the borgmatic file backup that already covers `~erichblume/`. Devpi indexes are recoverable from upstream PyPI on a fresh start; only the local index (`root/dev`) is unique state.
+The server-dir is **not** in `borgmatic_source_directories` and is not backed up. The PyPI cache (`+files/`) is re-fetchable from upstream on first request; the local `eblume/dev` index can be republished from source. If retention becomes important, add `/Users/erichblume/devpi/server-dir/` to the borgmatic source list.
 
 ## Related
 

docs/how-to/operations/rebuild-minikube-cluster.md

@@ -235,25 +235,7 @@ mise run services-check
 
 ## Post-Rebuild: Cold Cache Failures
 
-### Devpi (PyPI Cache)
+Devpi runs natively on indri (see [[devpi-on-indri]]) and is unaffected by minikube rebuilds, so the historical "devpi cold cache after rebuild" failure mode no longer applies. If devpi itself goes cold (fresh server-dir), the same lazy-cache race can still cause `404` on the first Dagger build under concurrent load — re-run the build to warm the cache, or pre-warm with `uv pip install --dry-run --index-url https://pypi.ops.eblu.me/root/pypi/+simple/ dagger-io`.
-
-After a rebuild, devpi's package cache is empty. The first Dagger-based container build will trigger a flood of concurrent package downloads. Devpi uses lazy caching — it serves package metadata (simple index) immediately from upstream PyPI but fetches wheel files on demand. Under heavy concurrent load with a cold cache, the upstream fetch can race with the client request, causing devpi to return `no such file` (HTTP 404) for packages it knows about but hasn't finished downloading yet.
-
-**Why devpi, not PyPI?** The repo's `uv.lock` was generated with devpi as the index, so every package source URL points at `pypi.ops.eblu.me`. Dagger's Python SDK runtime does a locked install (`uv sync`), not fresh resolution — it fetches from whatever URLs are in the lockfile. This is intentional (supply chain control), but means all builds — local and CI — depend on devpi being available and warm.
-
-**Symptoms:** Forgejo Actions Dagger builds fail during module initialization with errors like:
-```
-Failed to download `googleapis-common-protos==1.74.0`
-HTTP status client error (404 Not Found) for url (https://pypi.ops.eblu.me/root/pypi/+f/...)
-```
-
-**Fix:** Re-run the failed build. The first attempt warms the cache; subsequent builds succeed. Alternatively, warm the cache manually before triggering CI builds:
-
-```bash
-# From any machine that can reach pypi.ops.eblu.me, install the Dagger SDK
-# to pre-populate the most common packages:
-pip install --dry-run --index-url https://pypi.ops.eblu.me/root/pypi/+simple/ dagger-io
-```
 
 ## Related
 

docs/reference/infrastructure/tailscale.md

@@ -33,7 +33,7 @@ ACLs managed via Pulumi in `pulumi/tailscale/policy.hujson`.
 | `tag:loki` | indri | Loki log aggregation |
 | `tag:k8s-api` | indri | Kubernetes API server (minikube) |
 | `tag:k8s-operator` | (operator pod) | Tailscale operator for k8s — see [[tailscale-operator]] |
-| `tag:k8s` | (Ingress proxy pods) | Kubernetes Tailscale Ingress nodes; each also carries a per-service tag (`tag:grafana`, `tag:kiwix`, `tag:devpi`, `tag:feed`, `tag:pg`) |
+| `tag:k8s` | (Ingress proxy pods) | Kubernetes Tailscale Ingress nodes; each also carries a per-service tag (`tag:grafana`, `tag:kiwix`, `tag:feed`, `tag:pg`) |
 | `tag:ci-gateway` | (ephemeral CI containers) | CI containers pushing images to registry |
 | `tag:flyio-proxy` | (Fly.io proxy container) | Public reverse proxy |
 | `tag:flyio-target` | indri, designated Ingress endpoints | Endpoints reachable by the Fly.io proxy (indri for Caddy routing, Ingress pods for Alloy metrics/logs) |

docs/reference/services/devpi.md

@@ -1,7 +1,7 @@
 ---
 title: Devpi
-modified: 2026-03-23
+modified: 2026-04-29
-last-reviewed: 2026-03-23
+last-reviewed: 2026-04-29
 tags:
   - service
   - python
@@ -9,31 +9,37 @@ tags:
 
 # devpi (PyPI Proxy)
 
-PyPI caching proxy and private package index.
+PyPI caching proxy and private package index. Runs natively on [[indri]] as a LaunchAgent (not in-cluster). See [[devpi-on-indri]] for deploy and operations.
 
 ## Quick Reference
 
 | Property | Value |
 |----------|-------|
-| **URL** | https://pypi.ops.eblu.me |
+| **URL** | `https://pypi.ops.eblu.me` |
-| **Namespace** | `devpi` |
+| **Listen** | `127.0.0.1:3141` (loopback only; reached via Caddy) |
-| **ArgoCD App** | `devpi` |
+| **Service** | LaunchAgent `mcquack.eblume.devpi` on indri |
-| **Storage** | 50Gi PVC |
+| **Server-dir** | `/Users/erichblume/devpi/server-dir/` |
-| **Image** | `registry.ops.eblu.me/blumeops/devpi` (see `argocd/manifests/devpi/kustomization.yaml` for current tag) |
+| **Runtime** | uv-managed venv at `/Users/erichblume/devpi/venv/` |
+| **Ansible role** | `ansible/roles/devpi/` |
+| **Versions** | Pinned in `ansible/roles/devpi/defaults/main.yml` (`devpi_server_version`, `devpi_web_version`) |
 
 ## Indices
 
 | Index | Purpose |
 |-------|---------|
-| `root/pypi` | PyPI mirror/cache (auto-created) |
+| `root/pypi` | PyPI mirror/cache (auto-created by `devpi-init`) |
-| `eblume/dev` | Private packages (inherits from root/pypi) |
+| `eblume/dev` | Private packages (inherits from `root/pypi`) |
 
 ## Credentials
 
-Root password stored in 1Password (blumeops vault), injected via ExternalSecret.
+Root password stored in 1Password (`blumeops` vault, item `devpi`, field `root password`). Fetched via `op read` in the `ansible/playbooks/indri.yml` `pre_tasks` and passed to the role on first init.
+
+## Backup
+
+The server-dir is **not** backed up. The PyPI cache (`+files/`) is re-fetchable from upstream on first request. The local `eblume/dev` index metadata is small but also not critical to retain — packages can be republished from source. If retention becomes important, add `/Users/erichblume/devpi/server-dir/` to `borgmatic_source_directories`.
 
 ## Related
 
-- [[use-pypi-proxy]] - Client configuration and package uploads
+- [[devpi-on-indri]] — Deploy, verify, and version-bump procedures
-- [[argocd]] - Deployment
+- [[use-pypi-proxy]] — Client configuration and package uploads
-- [[1password]] - Secrets management
+- [[1password]] — Secrets management

docs/reference/storage/backups.md

@@ -62,7 +62,7 @@ Other data lives directly on [[sifaka]] (music via [[navidrome]], video via [[je
 | ZIM archives (`~/transmission/`) | Re-downloadable via torrent |
 | Prometheus metrics | Ephemeral, in k8s PVC |
 | Loki logs | Ephemeral, in k8s PVC |
-| devpi cache | Re-fetchable from PyPI |
+| devpi cache (`~/devpi/server-dir/` on indri) | Re-fetchable from PyPI on first request |
 
 ## Retention Policy
 

pulumi/tailscale/__main__.py

@@ -37,7 +37,7 @@ acl = tailscale.Acl(
 
 # indri - Mac Mini M1, primary homelab server
 # Hosts forge, loki, zot registry, and the k8s control plane.
-# Other services (grafana, kiwix, devpi, etc.) run in k8s with their own Tailscale devices.
+# Other services (grafana, kiwix, etc.) run in k8s with their own Tailscale devices.
 indri = tailscale.get_device(name="indri.tail8d86e.ts.net")
 indri_tags = tailscale.DeviceTags(
     "indri-tags",

pulumi/tailscale/policy.hujson

@@ -20,7 +20,8 @@
 		},
 
 		// --- Members: user-facing services only ---
-		// Kiwix, Forge, devpi, Miniflux, PostgreSQL
+		// Kiwix, Forge, Miniflux, PostgreSQL
+		// (devpi moved off-cluster to indri; reachable via Caddy on tag:flyio-target)
 		{
 			"src": ["autogroup:member"],
 			"dst": ["tag:kiwix"],
@@ -31,11 +32,6 @@
 			"dst": ["tag:forge"],
 			"ip":  ["tcp:443", "tcp:22"],
 		},
-		{
-			"src": ["autogroup:member"],
-			"dst": ["tag:devpi"],
-			"ip":  ["tcp:443"],
-		},
 		{
 			"src": ["autogroup:member"],
 			"dst": ["tag:feed"],
@@ -152,7 +148,6 @@
 		"tag:grafana": ["autogroup:admin", "tag:blumeops"],
 		"tag:kiwix": ["autogroup:admin", "tag:blumeops"],
 		"tag:forge": ["autogroup:admin", "tag:blumeops"],
-		"tag:devpi": ["autogroup:admin", "tag:blumeops"],
 		"tag:loki": ["autogroup:admin", "tag:blumeops"],
 		"tag:pg": ["autogroup:admin", "tag:blumeops"],
 		"tag:feed": ["autogroup:admin", "tag:blumeops"],