From 0cac9d6986e42d0888c94c5d13f50eb98ae85106 Mon Sep 17 00:00:00 2001 From: Ilker Manap Date: Thu, 17 Sep 2026 19:14:04 +0200 Subject: [PATCH] Make the landing page look like something worth having a key for MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It is the only page a stranger sees, and until they have a key it is the whole product: a stack of headings and shell snippets said "internal tool you probably should not be here". Now it opens with what this is and a diagram of why it exists — one address, many GPUs, fabrics that cannot see each other — then the base URL as a headline rather than a line of prose with a colon in it, the route to a key as three cards, and the client examples as tabs instead of a pile. The tabs are radio buttons and a stylesheet. No JavaScript decides what is on the page, so someone who arrives with scripting off still gets every example; the only script is the copy buttons, which exist only where the clipboard API does and so cannot leave a dead control behind. The page says exactly what it said before, including that it lists no models and why. The diagram names no model, no node and no count: fabrics are drawn because llmgw keeps them apart, not because this install has two. Tests cover the parts that are easy to break later: every example lives in the page itself rather than being fetched per tab, the reflected base URL — now in five places rather than one — still cannot write markup, and the diagram carries a label for anyone not looking at it. Co-Authored-By: Claude Opus 5 (1M context) --- internal/web/static/base.css | 92 +++++++++++ internal/web/templates/landing.html | 246 ++++++++++++++++++++++------ internal/web/web_test.go | 43 +++++ 3 files changed, 331 insertions(+), 50 deletions(-) diff --git a/internal/web/static/base.css b/internal/web/static/base.css index 1fb1b69..a692258 100644 --- a/internal/web/static/base.css +++ b/internal/web/static/base.css @@ -170,3 +170,95 @@ select { font: inherit; padding: .35rem .4rem; border: 1px solid var(--line); * design width so scaling up never inflates the 12px labels. */ .ranked { margin: .25rem 0 1rem; } .ranked svg { display: block; width: 100%; max-width: 720px; height: auto; } + +/* The landing page. + * + * It is the only page here a stranger sees, and for most readers it is the whole + * product until they have a key. Everything below is layout and surface: the + * page still says exactly what it said before, and still names no model, no node + * and no count. */ +.banner { border-radius: 6px; padding: .7rem .9rem; margin: 0 0 1.5rem; + border: 1px solid var(--line); background: var(--card); font-size: .875rem; } +.warn-banner { border-color: var(--warn); color: var(--warn); } + +.hero { display: grid; grid-template-columns: minmax(0, 1fr) minmax(0, 420px); + gap: 2rem; align-items: center; border-bottom: 1px solid var(--line); + margin-bottom: 2.25rem; padding-bottom: 2rem; } +.hero h1 { font-size: 2.6rem; line-height: 1; margin: 0 0 .5rem; letter-spacing: -.03em; } +.hero .sub { font-size: 1.15rem; color: var(--fg); font-weight: 500; } +.hero .lede { color: var(--muted); max-width: 34rem; margin: .75rem 0 0; } +.cta { display: flex; flex-wrap: wrap; gap: .6rem; margin: 1.35rem 0 0; } +.btn { display: inline-block; font-size: .9rem; font-weight: 600; text-decoration: none; + background: var(--accent); color: var(--bg); border: 1px solid var(--accent); + border-radius: 6px; padding: .5rem 1.1rem; } +.btn:hover { filter: brightness(1.1); } +.btn.ghost { background: none; color: var(--accent); border-color: var(--line); } +.btn.ghost:hover { border-color: var(--accent); filter: none; } + +.schematic { margin: 0; } +.schematic svg { display: block; width: 100%; height: auto; } +.schematic figcaption { color: var(--muted); font-size: .8rem; margin-top: .6rem; } +@media (max-width: 760px) { + .hero { grid-template-columns: 1fr; gap: 1.5rem; } + .hero h1 { font-size: 2rem; } +} + +/* The base URL is what a reader came for, so it is a headline rather than a + * line of prose with a colon in it. */ +.url-card { display: flex; flex-wrap: wrap; align-items: baseline; gap: .35rem 1rem; + background: var(--card); border: 1px solid var(--line); + border-left: 3px solid var(--accent); border-radius: 6px; + padding: .85rem 1.1rem; } +.url-card .k { font-size: .72rem; font-weight: 600; text-transform: uppercase; + letter-spacing: .06em; color: var(--muted); } +.url-card code { font-size: 1.05rem; word-break: break-all; } + +.steps { list-style: none; padding: 0; margin: .5rem 0 0; + display: grid; gap: 1rem; grid-template-columns: repeat(3, minmax(0, 1fr)); } +.steps > li { margin: 0; background: var(--card); border: 1px solid var(--line); + border-radius: 8px; padding: 1.1rem 1.1rem 1.25rem; } +.steps .n { display: inline-grid; place-items: center; width: 1.6rem; height: 1.6rem; + border-radius: 50%; background: var(--accent); color: var(--bg); + font-size: .8rem; font-weight: 700; } +.steps h3 { font-size: .98rem; margin: .6rem 0 .4rem; } +.steps p { margin: .4rem 0 0; font-size: .9rem; } +/* A card is narrower than a line of shell, and a block that scrolls sideways + * inside a card is one nobody discovers they can scroll. */ +.steps pre { margin: .7rem 0 0; font-size: .78em; + white-space: pre-wrap; overflow-wrap: anywhere; } +@media (max-width: 860px) { .steps { grid-template-columns: 1fr; } } + +/* Tabs with no JavaScript: the radios carry the state, so the panel a reader + * chose survives everything including the script never loading. */ +.tabs { display: flex; flex-wrap: wrap; gap: .35rem; margin: .5rem 0 .75rem; } +.tabs > input { position: absolute; width: 1px; height: 1px; opacity: 0; } +.tabs > label { display: inline-block; margin: 0; cursor: pointer; font-size: .85rem; + font-weight: 600; color: var(--muted); padding: .35rem .85rem; + border: 1px solid transparent; border-radius: 999px; } +.tabs > label:hover { color: var(--fg); } +.tabs > input:checked + label { color: var(--fg); background: var(--card); + border-color: var(--line); } +.tabs > input:focus-visible + label { outline: 2px solid var(--accent); outline-offset: 2px; } +.tabs > .panels { flex: 1 0 100%; } +.tabs .panel { display: none; margin: 0; } +.tabs > input:nth-of-type(1):checked ~ .panels > .panel:nth-child(1), +.tabs > input:nth-of-type(2):checked ~ .panels > .panel:nth-child(2), +.tabs > input:nth-of-type(3):checked ~ .panels > .panel:nth-child(3) { display: block; } + +/* Copy buttons are injected by script, so they are positioned against the block + * they copy rather than reserving space that stays empty without one. */ +pre { position: relative; } +pre .copy { position: absolute; top: .4rem; right: .4rem; padding: .15rem .5rem; + font-size: .72rem; font-weight: 600; background: var(--bg); + color: var(--muted); border: 1px solid var(--line); border-radius: 4px; + opacity: 0; transition: opacity .12s; } +pre:hover .copy, pre .copy:focus-visible { opacity: 1; } +pre .copy:hover { color: var(--fg); filter: none; } + +.endpoints { list-style: none; padding: 0; margin: .5rem 0 0; } +.endpoints > li { display: flex; flex-wrap: wrap; align-items: baseline; gap: .3rem .8rem; + padding: .55rem 0; border-bottom: 1px solid var(--line); margin: 0; } +.endpoints > li:last-child { border-bottom: none; } +.endpoints > li span { color: var(--muted); font-size: .85rem; } +/* A button revealed by hover is a button a phone never has. */ +@media (hover: none) { pre .copy { opacity: 1; } } diff --git a/internal/web/templates/landing.html b/internal/web/templates/landing.html index 9280c6b..6afb79b 100644 --- a/internal/web/templates/landing.html +++ b/internal/web/templates/landing.html @@ -1,68 +1,214 @@ {{define "landing"}} -llmgw +llmgw + +
-
-

llmgw

-

An OpenAI-compatible endpoint in front of self-hosted models.

+ +{{if .Notice}}{{end}} +{{if not .AuthEnabled}} + +{{end}} + +
+
+

llmgw

+

One OpenAI-compatible endpoint in front of your own GPUs.

+

Point any OpenAI client at this address. The API is the one you + already have a client for — only the base URL changes.

+

+ {{if .SignupOpen}}Create an account + Sign in + {{else if .AuthEnabled}}Sign in{{end}} +

+
+ + {{/* A schematic of what this is, not of what is installed here. It names no + model, no node and no count: this page is served to anyone who can reach + the address. */}} +
+ + + + + + + + + + + your client + + editor agent + + a script + + + + + + + + + + llmgw + one address + + + + + + + + + + fabric + + + + + + fabric + + + + + +
Clients see one address. Work is placed on GPUs that + can actually run it — and a model split across machines never straddles two fabrics.
+
-

Point any OpenAI client at this address. The API is the one you already have a -client for; only the base URL changes.

+
+ Base URL + {{.BaseURL}}/v1 +
-

Base URL

-
{{.BaseURL}}/v1
+

Three steps

+
    +
  1. + 1 +

    Get a key

    + {{if .SignupOpen}} +

    Create an account. Someone approves it, and then you issue + your own keys from your profile — as many as you need, + revoking them yourself when a laptop is retired or a script is done.

    + {{else if .Contact}} +

    Ask {{if .ContactHref}}{{.Contact}}{{else}}{{.Contact}}{{end}} + for one.{{if .AuthEnabled}} If you already have an account, + sign in.{{end}}

    + {{else}} +

    Keys are issued by whoever operates this gateway. If you do not know who that + is, the address you reached this page at is the place to start asking.

    + {{end}} +

    A key is shown once and never again — only its hash is kept here — + so put it in a password manager straight away.

    +
  2. +
  3. + 2 +

    Point your client at it

    +

    Set the base URL above and your key. Every OpenAI SDK, and every tool built on + one, takes both as configuration.

    +
    export OPENAI_BASE_URL={{.BaseURL}}/v1
    +export OPENAI_API_KEY=$LLMGW_API_KEY
    +
  4. +
  5. + 3 +

    Pick a model

    + {{if .AuthEnabled}} +

    GET /v1/models lists what is available to your key. It needs the + key, which is why this page does not list them: this page is served to anyone who + can reach the address.

    + {{else}} +

    GET /v1/models lists what is available.

    + {{end}} +

    Sending the placeholder <model-name> unchanged is + answered with the list of models you may use, rather than with an error.

    +
  6. +
-

Example

-
curl {{.BaseURL}}/v1/chat/completions \
+

Connect your client

+
+ + + +
+
curl {{.BaseURL}}/v1/chat/completions \
   -H "Authorization: Bearer $LLMGW_API_KEY" \
   -H "Content-Type: application/json" \
-  -d '{"model": "<model-name>", "messages": [{"role":"user","content":"hello"}]}'
-

- Replace <model-name> with a name from - GET {{.BaseURL}}/v1/models, which needs your key. Sending the - placeholder unchanged is answered with the list of models that are available. -

+ -d '{"model": "<model-name>", + "messages": [{"role": "user", "content": "hello"}]}'
-
from openai import OpenAI
-client = OpenAI(base_url="{{.BaseURL}}/v1", api_key=os.environ["LLMGW_API_KEY"])
+
import os
+from openai import OpenAI
 
-

Which models

-{{if .AuthEnabled}} -

GET /v1/models lists what is available to your key. It needs the key, -which is why this page does not list them: this page is served to anyone who can -reach the address.

-{{else}} -

GET /v1/models lists what is available.

-

This gateway has no API keys configured, so it serves anyone who can -reach it. That is a deployment choice and only safe on a trusted network.

-{{end}} +client = OpenAI(base_url="{{.BaseURL}}/v1", + api_key=os.environ["LLMGW_API_KEY"]) -

Getting a key

-{{if .SignupOpen}} -

Create an account. Someone will approve it, and then you -can issue your own keys from your profile — as many as you -need, revoking them yourself when a laptop is retired or a script is done.

-

Already have an account? Sign in.

-{{else if .Contact}} -

Ask {{if .ContactHref}}{{.Contact}}{{else}}{{.Contact}}{{end}} -for one. {{if .AuthEnabled}}If you already have an account, -sign in.{{end}}

-{{else}} -

Keys are issued by whoever operates this gateway. If you do not know who that -is, the address you reached this page at is the place to start asking.

-{{end}} -

A key is shown once and never again — only its hash is kept here — so put it in -a password manager straight away.

+reply = client.chat.completions.create( + model="<model-name>", + messages=[{"role": "user", "content": "hello"}], +) +print(reply.choices[0].message.content)
-{{if .Notice}}

{{.Notice}}

{{end}} +
// ~/.config/opencode/opencode.json
+{
+  "$schema": "https://opencode.ai/config.json",
+  "provider": {
+    "llmgw": {
+      "npm": "@ai-sdk/openai-compatible",
+      "name": "llmgw",
+      "options": {
+        "baseURL": "{{.BaseURL}}/v1",
+        "apiKey": "{env:LLMGW_API_KEY}"
+      },
+      "models": { "<model-name>": { "name": "<model-name>" } }
+    }
+  }
+}
+
+ +

Anything that speaks the chat-completions API works, whether or not it +is listed here — there is nothing llmgw-specific to install.

-

Once you have one

-