wiki: publish pending edits (2026-10-20)

This commit is contained in:
Tony0410
2026-10-20 18:38:11 +08:00
parent 667ab7f41d
commit fbd4f297eb
46 changed files with 1012 additions and 197 deletions
+9 -3
View File
@@ -3,8 +3,8 @@ title: Cloudflare Tunnel
type: system
status: active
created: 2026-07-22
updated: 2026-07-22
verified_on: 2026-07-22
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [system, network, cloudflare]
sources: [raw/configs/hermes-config-sanitized.txt]
@@ -26,6 +26,12 @@ Exposes internal homelab services, including Hermes, to the public internet with
- `originRequest`: `noTLSVerify: true`, `disableChunkedEncoding: true`, `http2Origin: false` (global defaults; some hostnames override per-entry)
- Ingress is a flat list of `hostname → service` mappings ending in `service: http_status:404` catch-all — see [[infrastructure/domains-and-tunnels]] for the full list and the entries specific to Hermes.
- Two Hermes-relevant entries point at this host's Tailscale IP (`100.118.5.51`): `hermes.martinwa.org` (port 19119) and `hermes-desktop.martinwa.org` (port 5174).
- **Re-verified 2026-10-20:** `cloudflared.service` is still **active** on ct207, the tunnel name/ID
(`ct207-tunnel` / `339a6757-5574-4723-a16c-da4769535caa`) and the credentials-file path are unchanged,
and the global `originRequest` defaults still read `noTLSVerify: true`, `disableChunkedEncoding: true`,
`http2Origin: false`. The Hermes-relevant ingress entries have since **expanded to four**:
`hermes.martinwa.org` (9119), `hermes-chat.martinwa.org` (8787), `tg.hermes.martinwa.org` (19119)
and `hermes-desktop.martinwa.org` (5174). Full live list in [[infrastructure/domains-and-tunnels]].
## Connected systems
- [[systems/hermes-agent]] — the service actually reached through this tunnel
@@ -34,7 +40,7 @@ Exposes internal homelab services, including Hermes, to the public internet with
## Known issues
- This host cannot restart or inspect the tunnel directly (no `cloudflared` installed here); a tunnel-side problem requires access to ct207.
- `mcp-portal.martinwa.org`, referenced in Hermes's own config as an MCP endpoint, was not found in the ct207 ingress list captured 2026-07-22 — **unresolved**, may be on a different tunnel/host.
- `mcp-portal.martinwa.org`, referenced in Hermes's own config as an MCP endpoint, was not found in the ct207 ingress list captured 2026-07-22 — **unresolved**, may be on a different tunnel/host. Re-checked against the live ct207 `config.yml` on 2026-10-20: **still absent**, still unresolved.
## Relevant runbooks
- [[runbooks/diagnose-docker]] (closest existing runbook; no dedicated tunnel-restart runbook exists yet — worth writing one that documents `systemctl restart cloudflared` on ct207, since Hermes has no direct access to perform it)
+8 -4
View File
@@ -3,8 +3,8 @@ title: Guanaco
type: system
status: active
created: 2026-08-14
updated: 2026-08-14
verified_on: 2026-08-14
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [llm, proxy, router, guanaco, ollama]
sources: [live-probe, session-history]
@@ -17,14 +17,18 @@ LLM proxy/router providing model routing and tool-call formatting for the Ollama
## Current Configuration
- **Host:** CT205 (`ollama` LXC container) on Proxmox
- **Version:** v0.8.14 (updated from v0.8.11 on 2026-08-14)
- **Version:** v0.8.15 (updated from v0.8.14 on 2026-08-14; verified at tag `v0.8.15`, commit `5df435e`, 2026-08-14 16:06:13 UTC)
- **Repo path (on CT205):** `/root/.guanaco/repo`
- **API endpoint:** `ollama:11434` (DNS) / `localhost:11434`
- **Key component:** `cmdcode_client.py` — Guanaco's tool-call formatting layer (DSML)
## Version History
### v0.8.14 (current, updated 2026-08-14)
### v0.8.15 (current, updated 2026-08-14)
- **Changelog bump + DSML fixes** — HEAD is tagged `v0.8.15` (commit `5df435e`, "chore: bump to v0.8.15 + changelog"), built from `origin/main` / `origin/develop`.
- Live probe 2026-10-20: `/root/.guanaco/repo` on CT205 at tag `v0.8.15`; Ollama API on `localhost:11434` answers **200**.
### v0.8.14 (previous, 2026-08-14)
- **DSML tag leakage fixes** (cumulative across v0.8.12–v0.8.14):
- v0.8.12 — incomplete DSML blocks leaking raw tags as content
- v0.8.13 — comprehensive DSML tag leakage: double-pipe, bare fragments, solo `invoke`
+13 -3
View File
@@ -3,8 +3,8 @@ title: Headless Server
type: system
status: active
created: 2026-07-22
updated: 2026-07-22
verified_on: 2026-07-22
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [system, browser]
sources: [raw/configs/hermes-config-sanitized.txt]
@@ -27,7 +27,17 @@ Headless X11 environment backing Hermes's browser automation and any GUI tooling
| `x11vnc` | `-display :99 -forever -shared -rfbport 5901 -rfbportv6 5901 -noxdamage -quiet -nopw` — **no VNC password set** |
| `websockify` ×2 | `127.0.0.1:6080 → 127.0.0.1:5900` and `127.0.0.1:6081 → 127.0.0.1:5901`, serving noVNC (`/usr/share/novnc`) |
Two VNC/websockify pairs exist (5900/6080 and 5901/6081) — the second (`:99`→5901→6081) is the one confirmed above; the first (5900→6080) predates it and wasn't independently re-verified this pass — **unresolved** whether it's a second live display or a leftover process.
**Re-verified 2026-10-20.** Two websockify instances still run (6080→5900 and 6081→5901), but
**only one Xvfb exists** (`:99`). Port **5900 has no listener at all** — the 6080 websockify points
at a dead port, so it is a **leftover** from an earlier display, not a second live one. The 6080
pair is the **Camofox** noVNC front door (`/home/hermes/services/camofox-browser/plugins/vnc/vnc-watcher.sh`),
which proxies to 5900; Camofox's REST API is separately live on `:9377` (HTTP 200).
The live chain is therefore **`:99` → x11vnc 5901 → websockify 6081**.
Also confirmed still true in 2026-10-20: `x11vnc` still runs with `-nopw` (no VNC password) and
still binds **all interfaces** — `0.0.0.0:5901` and `[::]:5901` — while both websockify ports
are correctly loopback-only (`127.0.0.1:6080`, `127.0.0.1:6081`).
## Connected systems
- [[systems/hermes-agent]] (browser backend uses this display — see [[systems/browser-backend]] / [[comparisons/camoufox-vs-chrome]])
+8 -2
View File
@@ -3,8 +3,8 @@ title: hermes-ui (Browser Hermes Desktop)
type: system
status: live
created: 2026-07-30
updated: 2026-07-30
verified_on: 2026-07-30
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [hermes, webui, desktop, browser]
---
@@ -25,6 +25,12 @@ The official Hermes Desktop renderer, extracted from the `hermes-agent` monorepo
| CT460 (this box) | `/home/hermes/hermes-ui/` | 5174 | Vite dev server | Active |
| CT450 (hermesagent) | `/home/hermes/hermes-ui/` | 5174 | Vite dev server | Active (since 2026-07-30) |
> **Re-verified 2026-10-20 — both deployments are up.** CT460: `0.0.0.0:5174` held by a `node`
> process; CT450: `0.0.0.0:5174` confirmed by `ss -ltn` inside CT450, and `/home/hermes/hermes-ui`
> exists there. Note **5175** on CT460 is a *different* service — a `python3` on `127.0.0.1:5175`
> (the `/ui` Tailscale Serve route) — and **8787** is the Hermes WebUI (`python`, `0.0.0.0:8787`,
> the `/webui` route). Don't conflate 5174 / 5175 / 8787.
## How It Works
- Runs as a Vite dev server, proxying REST/auth/WebSocket to a running Hermes gateway (`http://127.0.0.1:9119` by default).
+72 -42
View File
@@ -1,9 +1,10 @@
---
title: Homepage Dashboard
type: systems
status: active
status: degraded
created: 2026-07-26
updated: 2026-07-26
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [dashboard, homepage, homelab, proxmox, monitoring]
---
@@ -13,18 +14,35 @@ tags: [dashboard, homepage, homelab, proxmox, monitoring]
## Purpose
Central dashboard for all homelab services, with Proxmox integration, per-VM stats, and a management UI.
## URLs
- **Tailscale:** `https://runtipi.kangaroo-eel.ts.net:8443/`
- **Local:** `http://runtipi.kangaroo-eel.ts.net:8082/`
- **Manager UI:** `http://runtipi.kangaroo-eel.ts.net:8084/`
## ⚠ Status: DEGRADED (verified 2026-10-20)
The stock **`homepage` container is not running** on CT201. Only `homepage-manager` (port 8084)
and `homebox` are up; `homebridge` exited two months ago. **Port 8082 has no listener**, and the
`asset_proxy.py` front door on port 8085 proxies to 8082 — so it answers **502 Connection refused**.
## Architecture
The dashboard that *is* live and serving **200** is **Homarr** on port **8102**
(`homar.martinwa.org` in the Cloudflare tunnel ingress, and `homarr` in the Runtipi app list).
The dashboard content itself is now **custom Python tooling** under `/opt/homepage/`
(`manager.py`, `asset_proxy.py`, `dashboard.html/js/css`), not the upstream Homepage image.
**Two independent dashboards are therefore in play — do not assume "the dashboard" means Homepage.**
## URLs
- **Homarr (live, 200):** `http://100.96.244.39:8102/` — public via `https://homar.martinwa.org`
- **Tailscale:** `https://runtipi.kangaroo-eel.ts.net:8443/` → proxies to 8085 → **currently 502**
- **Local:** `http://100.96.244.39:8082/` → **no listener**
- **Manager UI:** `http://runtipi.kangaroo-eel.ts.net:8084/` → `homepage-manager`, up
## Architecture (verified 2026-10-20)
```
CT201 (runtipi)
├── Homepage container (port 8082) — main dashboard
├── Manager container (port 8084) — web UI for adding/removing apps
└── Tailscale serve (port 8443) — HTTPS proxy to Homepage
├── homarr container (port 8102) — live dashboard, HTTP 200
├── homepage-manager container (8084) — web UI for adding/removing apps, up
├── homebox container (7745) — up
├── asset_proxy.py (python3, port 8085) — proxies 8085 → 8082, currently 502
├── manager.py (python3) — dashboard management service, up
└── Tailscale serve (port 8443) — / → 8085, /mgr → 8084, /steel → 5173, /steel-api → 3000
```
No container is listening on **8082** — `docker ps` on CT201 lists no `homepage` container at all.
## Tabs
| Tab | Content |
@@ -44,39 +62,43 @@ CT201 (runtipi)
- **proxmox.yaml:** Config at `/opt/homepage/config/proxmox.yaml`
### VM/LXC Mapping
| Service | VMID | Type |
|---|---|---|
| PVE | (node) | — |
| Portainer docker | 200 | lxc |
| Portainer debianVM | 101 | qemu |
| Portainer openweb | 245 | lxc |
> **Two Proxmox hosts exist and guest IDs collide** — always qualify with a host.
> `ct460` = PVE `100.96.100.82`; `m93p` = PVE `100.104.51.98`. The mapping below was re-read from
> `pct list` on **both** hosts on 2026-10-20; several rows previously listed bare IDs now live on
> `m93p`, and one row no longer exists at all.
| Service | Host | VMID | Type | Live state |
|---|---|---|---|---|
| PVE | both | (node) | — | — |
| Portainer docker | ct460 | 200 | lxc | running (8000, 9443) |
| Portainer debianVM | ct460 | 101 | qemu | running |
| Portainer openweb | **m93p** | 245 | lxc | running |
| Runtipi | ct460 | 201 | lxc | running |
| Gitea | **m93p** | 221 | lxc | running |
| FreshRSS | **m93p** | **1260** | lxc | running |
| NextCloud | ct460 | 270 | lxc | running |
| LiteLLM | **m93p** | 500 | lxc | running |
| OmniRoute | ct460 | 600 | lxc | running (also `Omniroute:20128` on m93p 101) |
| Ntfy | **m93p** | **1241** | lxc | running |
| Home Assistant | ct460 | 100 | qemu | running |
| N8n | ct460 | 215 | lxc | running |
| Mattermost | ct460 | 255 | lxc | **stopped** |
| SFTPGo | ct460 | 299 | lxc | running |
| Cliproxy API | ct460 | 103 | lxc | running |
| OpenClaw | — | — | qemu | **absent — no VM 403 on either host** |
| Zeroclaw | ct460 | 333 | lxc | running |
| Hermes | ct460 | 460 | lxc | running |
| Hindsight | ct460 | 450 | lxc | running |
> Portainer on CT245 (`openweb.kangaroo-eel.ts.net:9443`) now uses a Tailscale/Let's Encrypt cert (installed 2026-08-09). Files: `/root/portainer-certs/portainer.{crt,key}` mounted into the container, `--sslcert/--sslkey`. Renewal: `/root/renew-portainer-cert.sh` via crontab (04:00 on the 1st monthly) — `tailscale cert --min-validity=20d` then `docker restart portainer`. Cert validity ~90 days.
| Runtipi | 201 | lxc |
| Gitea | 221 | lxc |
| FreshRSS | 260 | lxc |
| NextCloud | 270 | lxc |
| LiteLLM | 500 | lxc |
| OmniRoute | 600 | lxc |
| Ntfy | 241 | lxc |
| Home Assistant | 100 | qemu |
| N8n | 215 | lxc |
| Mattermost | 255 | lxc |
| SFTPGo | 299 | lxc |
| Cliproxy API | 103 | lxc |
| OpenClaw | 403 | qemu |
| Zeroclaw | 333 | lxc |
| Hermes | 460 | lxc |
| Hindsight | 450 | lxc |
## Configuration Files
All at `/opt/homepage/config/` on CT201:
- `services.yaml` — All apps, groups, and Proxmox mappings
- `settings.yaml` — Theme, layout, tabs
- `custom.css` — Warm amber accent overrides
- `widgets.yaml` — System health, clock
- `proxmox.yaml` — PVE API token
- `docker.yaml` — Docker socket integration
`/opt/homepage/` on CT201 now holds **custom Python tooling**, not the upstream Homepage bundle:
- `manager.py` — dashboard management service (running)
- `asset_proxy.py` — port 8085 → 8082 front door (running, currently 502 upstream)
- `dashboard.html` / `dashboard.js` / `dashboard.css` — the custom dashboard frontend
- `config/` — legacy Homepage config directory (still present, upstream `homepage` container not running)
- `manager.py.bak.*` — a long tail of dated manager backups (brand, chat2, cookie, hp, motion, owtools …)
## Manager UI
Web form at port 8084 for adding/removing apps without SSH.
@@ -96,7 +118,15 @@ Type anywhere on the dashboard to search services by name/description. Google se
Port 443 occupied by Traefik on CT201 — Tailscale uses port 8443 instead.
## Maintenance
- **Restart Homepage:** `docker restart homepage`
- **Restart Manager:** `docker restart homepage-manager`
- **Edit apps manually:** Edit `/opt/homepage/config/services.yaml` then restart
- **Docker auto-discovery:** Add `homepage.group`, `homepage.name`, `homepage.href` labels to containers
- **Restart Homarr:** `docker restart homarr`
- **Restart the custom dashboard:** the `manager.py` / `asset_proxy.py` processes run directly on CT201
(not in Docker) — restart via whatever supervises them; check `ps aux | grep -E 'manager.py|asset_proxy'`
- **Recover the upstream Homepage (8082):** the image `ghcr.io/gethomepage/homepage` is still present on
CT201 (4 months old) but no container uses it — recreate it to restore 8082 and the 8085 proxy path
- **Tailscale serve on CT201:** `/` → 8085, `/mgr` → 8084, `/steel` → 5173, `/steel-api` → 3000
## Recent Changes
- **2026-10-20 (live probe):** `homepage` container absent, 8082 dead, 8085 → 502, Homarr on 8102 → 200.
Status changed `active` → `degraded`. VM/LXC table re-read from both PVE hosts; corrected IDs and
host assignments, dropped the non-existent OpenClaw VM 403, marked Mattermost CT255 stopped.
+10 -2
View File
@@ -3,8 +3,8 @@ title: Nextcloud
type: system
status: active
created: 2026-08-17
updated: 2026-08-17
verified_on: 2026-08-17
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [nextcloud, file-sharing, self-hosted, infrastructure]
sources: []
@@ -28,6 +28,14 @@ Self-hosted file sync and sharing platform running on Proxmox LXC container.
|---|---|---|---|
| memories | 8.1.0 | active | Photo management |
| metadata | 0.24.0 | active | |
> **Re-verified 2026-10-20** — every fact on this page still holds: CT270 `nextcloud` is running,
> `occ status` reports **version 33.0.2.2** (installed, not in maintenance, no DB upgrade needed),
> the container holds `192.168.178.56` (LAN) and `100.90.102.19` (Tailscale), and both
> `memories 8.1.0` and `metadata 0.24.0` are enabled. `https://nextcloud.martinwa.org/` answers
> **302** both publicly and via the Tailscale IP. `occ` lives at `/var/www/nextcloud/occ`
> (not `/var/www/html/nextcloud`).
> Note: `occ` must be run from that path and as `www-data`.
| ncdownloader | 1.0.23 | **disabled** | Incompatible with NC33 — calls removed method `OC\Server::getAppConfig()` |
| news | 28.7.0 | active | RSS reader |
| nextcloud_announcements | 5.0.0 | active | |
+17 -1
View File
@@ -3,7 +3,8 @@ title: Obsidian Vault
type: systems
status: active
created: 2026-07-26
updated: 2026-07-26
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [obsidian, notes, vault, homelab]
---
@@ -19,6 +20,21 @@ Personal knowledge base — notes, research, projects, and agent documentation.
- **Structure:** NUMBERED folders (00-99)
- **Sync:** Git-based (manual or cron-triggered)
> **Re-verified 2026-10-20 — the "NUMBERED folders (00-99)" description is no longer accurate.**
> On disk there are now only two vault-shaped directories on CT460:
> - `/home/hermes/obsidian-vault/` — contains `50 - Research/` (with `Restaurants and Cafes/`) and
> `Cron Outputs/` (with `Weekend Adventure/`). **No `00 - Home`, no `10 - Inbox`, no `20 - Agent Notes`.**
> This directory is **not a git repo** (`git -C` → "not a git repository").
> - `/home/hermes/wiki/` — a **local-only** git repo (no remote configured) whose tracked content is
> the `agent-estate-wiki/` mirror plus `reports/uber-eats/`. Its `.obsidian/` dir is under
> `wiki/agent-estate-wiki/.obsidian`.
>
> So the numbered-folder tree described below is **largely absent**, and the vault that *is* mirrored
> to Gitea is the **agent-estate-wiki** repo (`Anthony/agent-estate-wiki`), not `Anthony/obsidian-vault`.
> Treat the folder table as historical. Separately, an **Obsidian LiveSync** service runs on **CT100
> (`obsidian-livesync`) on the `m93p` host** (`/opt/obsidian-livesync/`, holds only a `.env`).
> Full folder table, left for reference:
## Folder Structure
| Folder | Purpose |
|---|---|
+8 -2
View File
@@ -3,8 +3,8 @@ title: OpenCode Go
type: system
status: active
created: 2026-07-29
updated: 2026-07-29
verified_on: 2026-07-29
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [opencode, provider, models, subscription, limits]
sources: [https://opencode.ai/docs/go/]
@@ -22,6 +22,12 @@ Low cost subscription for open coding models — $5 first month, then **$10/mont
> Only one member per workspace can subscribe.
> **Note (2026-10-20):** the pricing and dollar-limit tables below are **vendor-published figures
> for the OpenCode Go product**, not a measurement of this estate. Nothing in `~/.hermes/config.yaml`
> or `jobs.json` shows an OpenCode Go provider in use — the estate routes through LiteLLM,
> OmniRoute (`omni-101`) and `Omniroute:20128`. Treat this page as reference material for the
> subscription, not as a description of a configured provider.
## Usage Limits
| Window | Dollar Limit |
+10 -2
View File
@@ -3,8 +3,8 @@ title: OpenRouter
type: system
status: active
created: 2026-07-22
updated: 2026-07-22
verified_on: 2026-07-22
updated: 2026-10-20
verified_on: 2026-10-20
confidence: medium
tags: [system, openrouter, provider]
sources: [raw/configs/hermes-config-sanitized.txt]
@@ -21,6 +21,14 @@ Primary key, fallback models, rate limits.
OpenRouter is not configured as a standalone provider in `~/.hermes/config.yaml` `custom_providers`.
OmniRoute’s catalog includes provider names and model IDs consistent with OpenRouter-style slugs, but live routing via OpenRouter directly is not present in current Hermes config.
> **Re-verified 2026-10-20 — the claim above still holds.** `custom_providers` in
> `/home/hermes/.hermes/config.yaml` is a list of exactly five entries:
> `litellm`, `guacano`, `Omniroute:20128`, `command-code`, `omni-101`. There is **no `openrouter`
> entry**. OpenRouter-style slugs (`openrouter/auto`, `freeapi/openrouterfree`,
> `kilo-openrouter-auto`, `or-openrouter-fusion`, and many more) exist only inside the **OmniRoute
> model catalog**, not as directly-routable Hermes providers. Routing to OpenRouter therefore still
> goes through OmniRoute, exactly as recorded here.
## History
- **User-provided:** Previously used; now largely replaced by LiteLLM and OmniRoute paths
- **Verified:** Decision page `keep-openrouter-as-fallback` exists, indicating retain intent
+2 -2
View File
@@ -3,8 +3,8 @@ title: Installed Skills
type: system
status: active
created: 2026-07-22
updated: 2026-10-19
verified_on: 2026-10-19
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [skills, hermes]
sources: []
+8 -3
View File
@@ -3,8 +3,8 @@ title: Terminal Backend
type: system
status: active
created: 2026-07-22
updated: 2026-07-22
verified_on: 2026-07-22
updated: 2026-10-20
verified_on: 2026-10-20
confidence: high
tags: [terminal, shell, backend, persistence]
sources: [raw/configs/hermes-config-sanitized.txt]
@@ -17,12 +17,17 @@ Execute shell commands, scripts, and interactive sessions for Hermes.
## Current Configuration
- **Backend:** local
- **Working directory:** `/home/hermes/workspace`
- **Working directory:** `/home/hermes/workspace/`
- **Timeout:** 180 s
- **Home mode:** auto
- **Container CPU:** 1
- **Persistent shell:** true
Live `terminal:` block re-read from `/home/hermes/.hermes/config.yaml` on 2026-10-20 also carries:
`lifetime_seconds: 300`, `container_memory: 5120`, `container_disk: 51200`,
`container_persistent: true`, `docker_mount_cwd_to_workspace: false`, `docker_run_as_host_user: true`.
`backend: local` and `persistent_shell: true` are unchanged since 2026-07-22.
## Persistence
- Shell state (venv, env vars, aliases) persists across terminal calls
- Working directory fixed to `/home/hermes/workspace`
+14 -6
View File
@@ -3,9 +3,9 @@ title: Hermes Tool Search
type: system
status: active
created: 2026-07-25
updated: 2026-07-25
review_after: 2026-10-23
verified_on: 2026-07-25
updated: 2026-10-20
review_after: 2026-11-20
verified_on: 2026-10-20
version_applies_to: Hermes Agent on CT460
confidence: high
tags: [hermes, automation, system]
@@ -32,11 +32,19 @@ File: `/home/hermes/.hermes/config.yaml`
tools:
tool_search:
enabled: "on"
threshold_pct: 10
search_default_limit: 5
max_search_limit: 20
threshold_pct: 20
search_default_limit: 8
max_search_limit: 50
listing_max_tokens: 8
```
> **Live config re-read 2026-10-20** — the values above are the current ones in
> `/home/hermes/.hermes/config.yaml`. They have drifted from the original 2026-07-25 rollout:
> `threshold_pct` 10 → **20**, `search_default_limit` 5 → **8**, `max_search_limit` 20 → **50**,
> and `listing_max_tokens: 8` has been added. `enabled: "on"` is unchanged.
> The token-budget figures in "Verified result" below were measured under the *original* settings
> and have not been re-measured since the tuning.
`enabled: "on"` is the important setting: it makes Tool Search active on every
request instead of waiting for the tool schemas to cross an automatic token
threshold.