<!DOCTYPE html>
<html lang="en">
{{template "views/partials/head" .}}
<body class="bg-gray-900 text-gray-200">
<div class="flex flex-col min-h-screen" x-data="{}">
{{template "views/partials/navbar" .}}
<div class="container mx-auto px-4 flex-grow">
<div class="workers mt-12 text-center">
<div class="animation-container">
<canvas id="networkCanvas"></canvas>
<div class="text-overlay">
<header class="text-center py-12">
<h1 class="text-5xl font-bold text-gray-100">
<i class="fa-solid fa-circle-nodes mr-2"></i> Distributed inference with P2P
</h1>
<p class="mt-4 text-lg">
Distribute computation by sharing and loadbalancing instances or sharding model weights.
<a href="https://localai.io/features/distribute/" target="_blank">
<i class="fas fa-circle-info pr-2"></i>
</a>
</p>
</header>
</div>
</div>
<h5 class="mb-4 text-justify">LocalAI uses P2P technologies to enable distribution of work between peers. It is possible to share an instance with Federation and/or split the weights of a model across peers (only available with llama.cpp models). You can now share computational resources between your devices or your friends!</h5>
<div class="bg-gray-800 p-6 rounded-lg shadow-lg mb-12 text-left">
<p class="text-lg font-bold mb-4 mt-1">
Network token
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">{{.P2PToken}}</code><br>
The network token can be used to either share the instance or join a federation or a worker network. Below you will find a few examples on how to start a new instance or a worker with the token and you will be able to see the available workers and federated nodes.
</div>
<!-- Warning box if p2p token is empty and p2p is enabled -->
{{ if and .IsP2PEnabled (eq .P2PToken "") }}
<div class="bg-red-500 p-4 rounded-lg shadow-lg mb-12 text-left">
<p class="text-xl font-semibold text-white"> <i class="fa-solid fa-exclamation-triangle"></i> Warning: P2P mode is disabled or no token was specified</p>
<p class="mb-4">You have to enable P2P mode by starting LocalAI with <code>--p2p</code>. Please restart the server with <code>--p2p</code> to generate a new token automatically that can be used to automatically discover other nodes. If you already have a token specify it with <code>export TOKEN=".."</code> <a href="https://localai.io/features/distribute/" target="_blank">
Check out the documentation for more information.
</a> </p>
</div>
{{ else }}
<!-- Federation Box -->
<div class="bg-gray-800 p-6 rounded-lg shadow-lg mb-12 text-left">
<p class="text-xl font-semibold text-gray-200"> <i class="text-gray-200 fa-solid fa-circle-nodes"></i> Federated Nodes: <span hx-get="/p2p/ui/workers-federation-stats" hx-trigger="every 1s"></span> </p>
<p class="mb-4">You can start LocalAI in federated mode to share your instance, or start the federated server to balance requests between nodes of the federation.</p>
<div class="grid grid-cols-1 sm:grid-cols-2 md:grid-cols-3 gap-4 mb-12">
<div hx-get="/p2p/ui/workers-federation" hx-trigger="every 1s"></div>
</div>
<hr class="border-gray-700 mb-12">
<h3 class="text-2xl font-semibold text-gray-100 mb-6"><i class="fa-solid fa-book"></i> Start a federated instance</h3>
<!-- Tabs navigation -->
<ul class="mb-5 flex list-none flex-row flex-wrap ps-0" role="tablist" data-twe-nav-ref>
<li role="presentation" class="flex-auto text-center">
<a href="#tabs-federated-cli" class="tablink my-2 block border-0 bg-gray-800 px-7 pb-3.5 pt-4 text-xs font-medium uppercase leading-tight text-white hover:bg-gray-700 focus:bg-gray-700 data-[twe-nav-active]:border-yellow-500 data-[twe-nav-active]:text-yellow-500 data-[twe-nav-active]:bg-gray-700 active" data-twe-toggle="pill" data-twe-target="#tabs-federated-cli" data-twe-nav-active role="tab" aria-controls="tabs-federated-cli" aria-selected="true"><i class="fa-solid fa-terminal"></i> CLI</a>
</li>
<li role="presentation" class="flex-auto text-center">
<a href="#tabs-federated-docker" class="tablink my-2 block border-0 bg-gray-800 px-7 pb-3.5 pt-4 text-xs font-medium uppercase leading-tight text-white hover:bg-gray-700 focus:bg-gray-700 data-[twe-nav-active]:border-yellow-500 data-[twe-nav-active]:text-yellow-500 data-[twe-nav-active]:bg-gray-700" data-twe-toggle="pill" data-twe-target="#tabs-federated-docker" role="tab" aria-controls="tabs-federated-docker" aria-selected="false"><i class="fa-solid fa-box-open"></i> Container images</a>
</li>
</ul>
<!-- Tabs content -->
<div class="mb-6">
<div class="tabcontent hidden opacity-100 transition-opacity duration-150 ease-linear data-[twe-tab-active]:block p-4" id="tabs-federated-cli" role="tabpanel" aria-labelledby="tabs-federated-cli" data-twe-tab-active>
<p class="text-lg font-bold mb-4 mt-1">
To start a new instance to share:
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">
# Start a new instance to share with --federated and a TOKEN<br>
export TOKEN="<span class="token">{{.P2PToken}}</span>"<br>
local-ai run --federated --p2p
</code>
<p class="mt-2">Note: If you don't have a token do not specify it and use the generated one that you can find in this page.</p>
<p class="text-lg font-bold mb-4 mt-1">
To start a new federated load balancer:
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">
export TOKEN="<span class="token">{{.P2PToken}}</span>"<br>
local-ai federated
</code>
<p class="mt-2">Note: Token is needed when starting the federated server.</p>
<p class="mt-2">For all the options available, please refer to the <a href="https://localai.io/features/distribute/#starting-workers" target="_blank" class="text-yellow-300 hover:text-yellow-400">documentation</a>.</p>
</div>
<div class="tabcontent hidden opacity-0 transition-opacity duration-150 ease-linear data-[twe-tab-active]:block p-4" id="tabs-federated-docker" role="tabpanel" aria-labelledby="tabs-federated-docker">
<p class="text-lg font-bold mb-4 mt-1">
To start a new federated instance:
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">
docker run -ti --net host -e TOKEN="<span class="token">{{.P2PToken}}</span>" --name local-ai -p 8080:8080 localai/localai:latest-cpu run --federated --p2p
</code>
<p class="text-lg font-bold mb-4 mt-1">
To start a new federated server with Docker (port to 9090):
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">
docker run -ti --net host -e TOKEN="<span class="token">{{.P2PToken}}</span>" --name local-ai -p 9090:8080 localai/localai:latest-cpu federated
</code>
<p class="mt-2">For all the options available and see what image to use, please refer to the <a href="https://localai.io/basics/container/" target="_blank" class="text-yellow-300 hover:text-yellow-400">Container images documentation</a> and <a href="https://localai.io/advanced/#cli-parameters" target="_blank" class="text-yellow-300 hover:text-yellow-400">CLI parameters documentation</a>.</p>
</div>
</div>
</div>
<!-- Llama.cpp Box -->
<div class="bg-gray-800 p-6 rounded-lg shadow-lg mb-12 text-left">
<p class="text-xl font-semibold text-gray-200"> <i class="text-gray-200 fa-solid fa-circle-nodes"></i> Workers (llama.cpp): <span hx-get="/p2p/ui/workers-stats" hx-trigger="every 1s"></span> </p>
<p class="mb-4">You can start llama.cpp workers to distribute weights between the workers and offload part of the computation. To start a new worker, you can use the CLI or Docker.</p>
<div class="grid grid-cols-1 sm:grid-cols-2 md:grid-cols-3 gap-4 mb-12">
<div hx-get="/p2p/ui/workers" hx-trigger="every 1s"></div>
</div>
<hr class="border-gray-700 mb-12">
<h3 class="text-2xl font-semibold text-gray-100 mb-6"><i class="fa-solid fa-book"></i> Start a new llama.cpp P2P worker</h3>
<!-- Tabs navigation -->
<ul class="mb-5 flex list-none flex-row flex-wrap ps-0" role="tablist" data-twe-nav-ref>
<li role="presentation" class="flex-auto text-center">
<a href="#tabs-cli" class="tablink my-2 block border-0 bg-gray-800 px-7 pb-3.5 pt-4 text-xs font-medium uppercase leading-tight text-white hover:bg-gray-700 focus:bg-gray-700 data-[twe-nav-active]:border-yellow-500 data-[twe-nav-active]:text-yellow-500 data-[twe-nav-active]:bg-gray-700 active" data-twe-toggle="pill" data-twe-target="#tabs-cli" data-twe-nav-active role="tab" aria-controls="tabs-cli" aria-selected="true"><i class="fa-solid fa-terminal"></i> CLI</a>
</li>
<li role="presentation" class="flex-auto text-center">
<a href="#tabs-docker" class="tablink my-2 block border-0 bg-gray-800 px-7 pb-3.5 pt-4 text-xs font-medium uppercase leading-tight text-white hover:bg-gray-700 focus:bg-gray-700 data-[twe-nav-active]:border-yellow-500 data-[twe-nav-active]:text-yellow-500 data-[twe-nav-active]:bg-gray-700" data-twe-toggle="pill" data-twe-target="#tabs-docker" role="tab" aria-controls="tabs-docker" aria-selected="false"><i class="fa-solid fa-box-open"></i> Container images</a>
</li>
</ul>
<!-- Tabs content -->
<div class="mb-6">
<div class="tabcontent hidden opacity-100 transition-opacity duration-150 ease-linear data-[twe-tab-active]:block p-4" id="tabs-cli" role="tabpanel" aria-labelledby="tabs-cli" data-twe-tab-active>
<p class="text-lg font-bold mb-4 mt-1">
To start a new worker, run the following command:
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">
export TOKEN="<span class="token">{{.P2PToken}}</span>"<br>
local-ai worker p2p-llama-cpp-rpc
</code>
<p class="mt-2">For all the options available, please refer to the <a href="https://localai.io/features/distribute/#starting-workers" target="_blank" class="text-yellow-300 hover:text-yellow-400">documentation</a>.</p>
</div>
<div class="tabcontent hidden opacity-0 transition-opacity duration-150 ease-linear data-[twe-tab-active]:block p-4" id="tabs-docker" role="tabpanel" aria-labelledby="tabs-docker">
<p class="text-lg font-bold mb-4 mt-1">
To start a new worker with docker, run the following command:
<i class="fa-solid fa-copy copy-icon float-right"></i>
</p>
<code class="block bg-gray-700 text-yellow-300 p-4 rounded-lg break-words" @click="copyClipboard($el.textContent)">
docker run -ti --net host -e TOKEN="<span class="token">{{.P2PToken}}</span>" --name local-ai -p 8080:8080 localai/localai:latest-cpu worker p2p-llama-cpp-rpc
</code>
<p class="mt-2">For all the options available and see what image to use, please refer to the <a href="https://localai.io/basics/container/" target="_blank" class="text-yellow-300 hover:text-yellow-400">Container images documentation</a> and <a href="https://localai.io/advanced/#cli-parameters" target="_blank" class="text-yellow-300 hover:text-yellow-400">CLI parameters documentation</a>.</p>
</div>
</div>
</div>
<!-- Llama.cpp Box END -->
{{ end }}
</div>
</div>
{{template "views/partials/footer" .}}
</div>
<script src="/static/p2panimation.js"></script>
<style>
.token {
word-break: break-all;
}
.workers .grid div {
display: flex;
flex-direction: column;
justify-content: space-between;
}
</style>
</body>
</html>