hermes: finalize dashboard exposure + local-inference config

Confirmed against upstream: dashboard binds 0.0.0.0:9119 by default
(HERMES_DASHBOARD_HOST/PORT), so bridge + PublishPort=5100:9119 needs no
override. LLM backend uses Hermes' `custom` OpenAI-compatible provider
pointed at the local sovereign inference (hanzalova.internal:31313/v1).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011D3YeWKpjg5bT488fVanCH

images/hermes/hermes.container

@@ -1,21 +1,21 @@
-# DRAFT reference quadlet for deploying Hermes on bob (bob.hanzalova.internal).
+# Reference quadlet for deploying Hermes on bob (bob.hanzalova.internal).
 # Deploy to /etc/containers/systemd/hermes.container (rootful, matching the
 # existing agent-zero.container and open-webui.container), then:
 #   sudo install -d -o 10000 -g 10000 /var/lib/hermes      # /opt/data owner = HERMES_UID
-#   # drop config.yaml + .env into /var/lib/hermes (LLM backend, secrets) — see readme.md
+#   sudo install -o 10000 -g 10000 /path/to/config.yaml /var/lib/hermes/config.yaml
+#   sudo install -o 10000 -g 10000 /path/to/.env          /var/lib/hermes/.env   # if needed
 #   sudo systemctl daemon-reload && sudo systemctl start hermes.service
 #
-# Once git.lair.cafe/lair/hermes:latest is published by the `images` workflow,
+# Gated on git.lair.cafe/lair/hermes:latest being published by the `images`
-# this is a normal pull + AutoUpdate=registry quadlet — same lifecycle as the
+# workflow first. After that it's a normal pull + AutoUpdate=registry quadlet —
-# other two services on bob.
+# same lifecycle as the other two services, and now enrolled in the (enabled)
+# podman-auto-update.timer.
 #
-# UNRESOLVED before first deploy (confirm against hermes dashboard docs):
+# Dashboard: the image binds the dashboard on 0.0.0.0:9119 by default
-#   The dashboard binds 127.0.0.1:9119 by default. To expose it on the LAN at
+# (HERMES_DASHBOARD_HOST / HERMES_DASHBOARD_PORT), so bridge networking +
-#   :5100 (the agent-zero=5080 / open-webui=5090 convention) the dashboard must
+# PublishPort below exposes it on the LAN at :5100 with no override needed.
-#   be told to bind 0.0.0.0 INSIDE the container — set that in
+# ⚠ The dashboard stores provider API keys and has NO auth — keep it on a trusted
-#   /var/lib/hermes/config.yaml (or a hermes dashboard-host env) and keep the
+# LAN only; front it with an authenticating reverse proxy for anything wider.
-#   PublishPort below. ⚠ It stores provider API keys and has no auth, so only
-#   expose on a trusted LAN — consider a reverse proxy with auth for anything wider.
 
 [Unit]
 Description=Hermes Agent
@@ -26,20 +26,15 @@ Wants=network-online.target
 Image=git.lair.cafe/lair/hermes:latest
 ContainerName=hermes
 AutoUpdate=registry
-# Bridge + PublishPort keeps the 50X0 LAN convention. Requires the dashboard to
+# Keeps the 50X0 LAN convention (agent-zero=5080, open-webui=5090, hermes=5100).
-# bind 0.0.0.0:9119 inside the container (see note above). If you instead accept
-# host networking like upstream's compose, replace the next two lines with
-# `Network=host` and configure the dashboard bind/port directly.
 PublishPort=5100:9119
 Volume=/var/lib/hermes:/opt/data:Z
 # Upstream drops to the non-root hermes user (uid/gid 10000); /var/lib/hermes
-# must be owned by 10000:10000 on the host (see install -d above).
+# must be owned 10000:10000 on the host (see install -d above).
 Environment=HERMES_UID=10000
 Environment=HERMES_GID=10000
-# LLM backend: point hermes at the local sovereign inference at
+# LLM backend (local sovereign inference) is configured in
-# http://hanzalova.internal:31313/v1 (same endpoint open-webui uses). Hermes is
+# /var/lib/hermes/config.yaml via provider: "custom" -> see readme.md.
-# OpenRouter-first with per-provider base URLs and no plain OpenAI slot, so the
-# model routing is configured in /var/lib/hermes/config.yaml, not here. See readme.md.
 
 [Service]
 Restart=always

images/hermes/readme.md

@@ -40,20 +40,31 @@ existing `agent-zero` / `open-webui` services on bob. Summary:
 1. `git.lair.cafe/lair/hermes:latest` must be published first (run the `images`
    workflow).
 2. `sudo install -d -o 10000 -g 10000 /var/lib/hermes`
-3. Drop `config.yaml` + `.env` into `/var/lib/hermes`:
+3. Drop `config.yaml` into `/var/lib/hermes` (owned `10000:10000`) — **LLM backend
-   - **LLM backend → local sovereign inference.** Point hermes at
+   → local sovereign inference.** Hermes exposes a `custom` provider for any
-     `http://hanzalova.internal:31313/v1` (the endpoint open-webui already uses).
+   OpenAI-compatible endpoint, so point it at the same endpoint open-webui uses:
-     Hermes is OpenRouter-first with per-provider base URLs and no plain
+
-     OpenAI-base slot, so define the OpenAI-compatible provider/model in
+   ```yaml
-     `config.yaml` (confirm the exact schema against hermes docs).
+   # /var/lib/hermes/config.yaml
-   - Secrets (any provider keys, tool keys) go in `.env`, not the quadlet.
+   model:
+     provider: "custom"                              # OpenAI-compatible endpoint
+     base_url: "http://hanzalova.internal:31313/v1"
+     api_key: "beast"                                # matches open-webui's OPENAI_API_KEY
+     default: "<model-id-your-endpoint-serves>"      # see: curl http://hanzalova.internal:31313/v1/models
+     # context_length: 32768                         # optional
+     # max_tokens:     4096                           # optional, output ceiling
+   ```
+
+   Any other secrets (web-search/tool keys, messaging tokens) go in
+   `/var/lib/hermes/.env`, never in the quadlet.
 4. Install `hermes.container` to `/etc/containers/systemd/`, `daemon-reload`,
    `start hermes.service`.
 
-### Open item: dashboard LAN exposure
+### Dashboard LAN exposure (resolved)
 
-The dashboard defaults to `127.0.0.1:9119` and **stores API keys with no auth**.
+The image binds the dashboard on **`0.0.0.0:9119` by default**
-The draft quadlet publishes it on the LAN at `:5100` (the 5080/5090 convention),
+(`HERMES_DASHBOARD_HOST` / `HERMES_DASHBOARD_PORT`), so bridge networking +
-which requires telling the dashboard to bind `0.0.0.0` inside the container (a
+`PublishPort=5100:9119` in the quadlet exposes it on the LAN at `:5100` with no
-`config.yaml`/env setting to confirm). Only expose on a trusted LAN; front it
+override. ⚠ The dashboard **stores provider API keys and has no auth** — keep it
-with an authenticating reverse proxy for anything wider.
+on a trusted LAN only, and front it with an authenticating reverse proxy for any
+wider exposure.