llmgw/internal/web/templates/landing.html

{{define "landing"}}<!doctype html>
<html lang="en"><head>
<meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1">
<title>llmgw</title>
<meta name="description" content="One OpenAI-compatible endpoint in front of self-hosted models.">
<link rel="stylesheet" href="{{stylesheet}}">
</head><body><div class="wrap">

{{if .Notice}}<p class="banner note">{{.Notice}}</p>{{end}}
{{if not .AuthEnabled}}
<p class="banner warn-banner">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.</p>
{{end}}

<header class="hero">
  <div class="hero-text">
    <h1>llmgw</h1>
    <p class="sub">One OpenAI-compatible endpoint in front of your own GPUs.</p>
    <p class="lede">Point any OpenAI client at this address. The API is the one you
    already have a client for — only the base URL changes.</p>
    <p class="cta">
      {{if .SignupOpen}}<a class="btn" href="/signup">Create an account</a>
      <a class="btn ghost" href="/login">Sign in</a>
      {{else if .AuthEnabled}}<a class="btn" href="/login">Sign in</a>{{end}}
    </p>
  </div>

  {{/* 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. */}}
  <figure class="schematic" aria-labelledby="schematic-cap">
    <svg viewBox="0 0 420 200" role="img"
         aria-label="Clients reach one endpoint, which places work on GPU nodes across separate fabrics.">
      <defs>
        <marker id="a" viewBox="0 0 8 8" refX="7" refY="4" markerWidth="6" markerHeight="6" orient="auto">
          <path d="M0 0 L8 4 L0 8 z" fill="var(--viz-grid)"/>
        </marker>
      </defs>

      <!-- clients -->
      <g font-size="10" fill="var(--muted)">
        <rect x="2" y="42" width="84" height="26" rx="6" fill="var(--card)" stroke="var(--line)"/>
        <text x="44" y="59" text-anchor="middle" fill="var(--fg)">your client</text>
        <rect x="2" y="86" width="84" height="26" rx="6" fill="var(--card)" stroke="var(--line)"/>
        <text x="44" y="103" text-anchor="middle" fill="var(--fg)">editor agent</text>
        <rect x="2" y="130" width="84" height="26" rx="6" fill="var(--card)" stroke="var(--line)"/>
        <text x="44" y="147" text-anchor="middle" fill="var(--fg)">a script</text>
      </g>
      <g fill="none" stroke="var(--viz-grid)" stroke-width="2" marker-end="url(#a)">
        <path d="M88 55 C 118 55 118 92 146 92"/>
        <path d="M88 99 h58"/>
        <path d="M88 143 C 118 143 118 108 146 108"/>
      </g>

      <!-- the gateway -->
      <rect x="150" y="70" width="72" height="60" rx="8"
            fill="var(--accent)" opacity=".12" stroke="var(--accent)" stroke-width="1.5"/>
      <text x="186" y="96" text-anchor="middle" font-size="12" font-weight="600"
            fill="var(--fg)">llmgw</text>
      <text x="186" y="112" text-anchor="middle" font-size="9"
            fill="var(--muted)">one address</text>

      <g fill="none" stroke="var(--viz-grid)" stroke-width="2" marker-end="url(#a)">
        <path d="M224 88 C 254 88 254 46 286 46"/>
        <path d="M224 112 C 254 112 254 146 286 146"/>
      </g>

      <!-- two fabrics that cannot see each other -->
      <g font-size="9" fill="var(--muted)">
        <rect x="288" y="18" width="128" height="58" rx="8"
              fill="none" stroke="var(--line)" stroke-dasharray="4 3"/>
        <text x="296" y="32">fabric</text>
        <rect x="296" y="40" width="26" height="26" rx="4" fill="var(--viz-1)" opacity=".85"/>
        <rect x="330" y="40" width="26" height="26" rx="4" fill="var(--viz-1)" opacity=".85"/>
        <rect x="364" y="40" width="26" height="26" rx="4" fill="var(--viz-1)" opacity=".85"/>

        <rect x="288" y="118" width="128" height="58" rx="8"
              fill="none" stroke="var(--line)" stroke-dasharray="4 3"/>
        <text x="296" y="132">fabric</text>
        <rect x="296" y="140" width="26" height="26" rx="4" fill="var(--viz-3)" opacity=".85"/>
        <rect x="330" y="140" width="26" height="26" rx="4" fill="var(--viz-3)" opacity=".85"/>
        <rect x="364" y="140" width="26" height="26" rx="4" fill="var(--viz-3)" opacity=".85"/>
      </g>
    </svg>
    <figcaption id="schematic-cap">Clients see one address. Work is placed on GPUs that
    can actually run it — and a model split across machines never straddles two fabrics.</figcaption>
  </figure>
</header>

<section class="url-card">
  <span class="k">Base URL</span>
  <code>{{.BaseURL}}/v1</code>
</section>

<h2>Three steps</h2>
<ol class="steps">
  <li>
    <span class="n">1</span>
    <h3>Get a key</h3>
    {{if .SignupOpen}}
    <p><a href="/signup">Create an account</a>. Someone approves it, and then you issue
    your own keys from your <a href="/profile">profile</a> — as many as you need,
    revoking them yourself when a laptop is retired or a script is done.</p>
    {{else if .Contact}}
    <p>Ask {{if .ContactHref}}<a href="{{.ContactHref}}">{{.Contact}}</a>{{else}}<strong>{{.Contact}}</strong>{{end}}
    for one.{{if .AuthEnabled}} If you already have an account,
    <a href="/login">sign in</a>.{{end}}</p>
    {{else}}
    <p>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.</p>
    {{end}}
    <p class="note">A key is shown once and never again — only its hash is kept here —
    so put it in a password manager straight away.</p>
  </li>
  <li>
    <span class="n">2</span>
    <h3>Point your client at it</h3>
    <p>Set the base URL above and your key. Every OpenAI SDK, and every tool built on
    one, takes both as configuration.</p>
    <pre><code>export OPENAI_BASE_URL={{.BaseURL}}/v1
export OPENAI_API_KEY=$LLMGW_API_KEY</code></pre>
  </li>
  <li>
    <span class="n">3</span>
    <h3>Pick a model</h3>
    {{if .AuthEnabled}}
    <p><code>GET /v1/models</code> 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.</p>
    {{else}}
    <p><code>GET /v1/models</code> lists what is available.</p>
    {{end}}
    <p class="note">Sending the placeholder <code>&lt;model-name&gt;</code> unchanged is
    answered with the list of models you may use, rather than with an error.</p>
  </li>
</ol>

<h2>Connect your client</h2>
<div class="tabs">
  <input type="radio" name="client" id="tab-curl" checked><label for="tab-curl">curl</label>
  <input type="radio" name="client" id="tab-py"><label for="tab-py">Python</label>
  <input type="radio" name="client" id="tab-oc"><label for="tab-oc">OpenCode</label>
  <div class="panels">
    <pre class="panel"><code>curl {{.BaseURL}}/v1/chat/completions \
  -H "Authorization: Bearer $LLMGW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "&lt;model-name&gt;",
       "messages": [{"role": "user", "content": "hello"}]}'</code></pre>

    <pre class="panel"><code>import os
from openai import OpenAI

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

reply = client.chat.completions.create(
    model="&lt;model-name&gt;",
    messages=[{"role": "user", "content": "hello"}],
)
print(reply.choices[0].message.content)</code></pre>

    <pre class="panel"><code>// ~/.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": { "&lt;model-name&gt;": { "name": "&lt;model-name&gt;" } }
    }
  }
}</code></pre>
  </div>
</div>
<p class="note">Anything that speaks the chat-completions API works, whether or not it
is listed here — there is nothing llmgw-specific to install.</p>

<h2>Once you have a key</h2>
<ul class="endpoints">
  <li><code>GET {{.BaseURL}}/v1/models</code><span>what your key may use</span></li>
  <li><code>GET {{.BaseURL}}/me</code><span>its limits, and what it has spent</span></li>
</ul>

<footer><span><a href="/about">llmgw {{.Version}}</a></span><span>{{.Now}}</span></footer>
</div>

<script>
// Vanilla and inline, for the same reason this page has no build step. It is an
// enhancement: the button only exists where the clipboard API does, so nothing
// here can leave a dead control on the page.
if (navigator.clipboard) {
  for (const pre of document.querySelectorAll('pre')) {
    // Read the text before the button joins the element, or "copy" ends up in
    // everything anyone pastes.
    const text = (pre.querySelector('code') || pre).textContent;
    const b = document.createElement('button');
    b.type = 'button';
    b.className = 'copy';
    b.textContent = 'copy';
    b.addEventListener('click', () => {
      navigator.clipboard.writeText(text).then(() => {
        b.textContent = 'copied';
        setTimeout(() => { b.textContent = 'copy'; }, 1200);
      });
    });
    pre.appendChild(b);
  }
}
</script>
</body></html>{{end}}