644f8829af
Introduce a new "AI Inference" item in the sidebar, linking to the "Inference Endpoints" index page. This page displays a list of all inference endpoints relevant to the user and project. Each model is presented in a visual card format, featuring key details: * Model name * Model type (e.g., Text Generation, Guard, Embedding) * Model URL * Pricing information * Usage example
74 lines
2.5 KiB
Text
74 lines
2.5 KiB
Text
<% @page_title = "Inference Endpoints" %>
|
|
<%== render("inference/tabbar") %>
|
|
|
|
<% @inference_endpoints.sort_by { |e| [e[:model_type], e[:model_name]]}.each_with_index do | ie, index |
|
|
|
|
path, model_type, model_icon, curl_message = case ie[:model_type]
|
|
when :text_generation
|
|
[
|
|
"/v1/chat/completions",
|
|
"Text Generation",
|
|
"hero-chat-bubble-bottom-center-text",
|
|
<<-MSG
|
|
"messages": [{"role": "user", "content": "say something"}],
|
|
"stream": true
|
|
MSG
|
|
]
|
|
when :guard
|
|
[
|
|
"/v1/chat/completions",
|
|
"Guard",
|
|
"hero-shield-check",
|
|
<<-MSG
|
|
"messages": [{"role": "user", "content": "is this safe?"}]
|
|
MSG
|
|
]
|
|
when :embedding
|
|
["/v1/chat/embeddings", "Embedding", "document-arrow-down", "\"input\": \"embed me!\"\n"]
|
|
else
|
|
fail "Unknown model type"
|
|
end
|
|
|
|
|
|
curl_snippet = <<-CURL
|
|
curl #{ie[:url]}#{path} \\
|
|
-H <span class="text-green-400">"Content-Type: application/json"</span> \\
|
|
-H <span class="text-green-400">"Authorization: Bearer <span class="text-orange-500">$INFERENCE_TOKEN</span>"</span> \\
|
|
-d <span class="text-green-400">'{
|
|
"model": "#{ie[:model_name]}",
|
|
#{curl_message} }'</span>
|
|
CURL
|
|
%>
|
|
|
|
<div class="inline-block w-auto bg-white shadow-md rounded-lg p-4 mr-4 <%= (index > 0) ? "mt-6" : "" %>">
|
|
<h2 class="text-lg font-bold mb-4 text-gray-800"><%= ie[:name] %></h2>
|
|
<div class="grid grid-cols-2 gap-4">
|
|
<div>
|
|
<div class="mb-4">
|
|
<h3 class="text-sm font-semibold text-gray-900">Type</h3>
|
|
<p class="text-gray-500"><%== render("components/icon", locals: { name: model_icon, classes: "inline-block w-5 h-5 flex-shrink-0 text-gray-400" }) %> <%= model_type %></p>
|
|
</div>
|
|
</div>
|
|
<div class="mb-4">
|
|
<h3 class="text-sm font-semibold text-gray-900">Pricing per million tokens</h3>
|
|
<p class="text-gray-500">$<%= ie[:price_million_tokens] %></p>
|
|
</div>
|
|
</div>
|
|
<div class="grid grid-cols-1 gap-4">
|
|
<div class="mb-4">
|
|
<h3 class="text-sm font-semibold text-gray-900">URL</h3>
|
|
<p class="text-gray-500"><%== render("components/copyable_content", locals: { content: ie[:url], message: "Copied URL" }) %></p>
|
|
</div>
|
|
</div>
|
|
<div class="grid grid-cols-1 gap-2">
|
|
<div class="mb-2">
|
|
<h3 class="text-sm font-semibold text-gray-900">
|
|
CURL usage example
|
|
</h3>
|
|
<div class="mt-2">
|
|
<pre class="text-sm bg-gray-800 text-white p-1 rounded-lg"><%== curl_snippet %></pre>
|
|
</div>
|
|
</div>
|
|
</div>
|
|
</div>
|
|
<% end %>
|