<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Olares on Medium]]></title>
        <description><![CDATA[Stories by Olares on Medium]]></description>
        <link>https://medium.com/@olares?source=rss-b748eaf1db0b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*yXwymVBAsC-qOFKBk1Ed9Q.jpeg</url>
            <title>Stories by Olares on Medium</title>
            <link>https://medium.com/@olares?source=rss-b748eaf1db0b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 17:28:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@olares/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Hermes Agent Memory: When Hindsight Is Worth Adding]]></title>
            <link>https://olares.medium.com/hermes-agent-memory-when-hindsight-is-worth-adding-7437f1ed54c3?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/7437f1ed54c3</guid>
            <category><![CDATA[hindsight]]></category>
            <category><![CDATA[ai-memory]]></category>
            <category><![CDATA[hermes-agent]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[self-hosting]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Fri, 20 Nov 2026 01:00:29 GMT</pubDate>
            <atom:updated>2026-11-21T02:01:03.964Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5M0ixafyvSaS9mgyRxpP9A.png" /></figure><p><a href="https://www.olares.com/docs/use-cases/hermes">Hermes Agent</a> has two kinds of built-in memory: two small files, USER.md and MEMORY.md, that load at the start of every session, and session search over past conversations. <a href="https://hindsight.vectorize.io/sdks/integrations/hermes">Hindsight</a>, an open-source memory service, can add a third as a memory provider: a bank that extracts facts from conversations and links related ones.</p><p>The built-in files handle a simple preference well. A reason behind a changed decision is harder to carry forward. If I ask why I switched course, I want Hermes to find the conversation where that reason appeared. That’s when Hindsight is worth adding.</p><h3>Start with Hermes’s built-in memory</h3><h4>Two short files for the things Hermes should always know</h4><p>Hermes loads USER.md and MEMORY.md at the start of a session. The first holds preferences about the person it works with. The second holds working notes about the agent&#39;s environment and ongoing work. You can edit both files directly, and Hermes can update them with its memory tool.</p><p>They are deliberately small: by default, USER.md holds 1,375 characters and MEMORY.md holds 2,200. That space works well for a writing preference, an environment detail, or a convention Hermes needs often.</p><p>Hermes reads both files once, when a session starts. If it saves a note mid-conversation, the file changes right away, but the note only reaches the model’s prompt in the next session. The <a href="https://hermes-agent.nousresearch.com/docs/user-guide/features/memory">Hermes memory docs</a> explain why it works this way.</p><h4>Session search finds the original exchange</h4><p>Hermes can also search messages from earlier sessions. Session search is a full-text keyword search over stored conversations, and it returns the original messages rather than a summary. If you remember roughly what you discussed, it can bring back the exchange and its exact wording.</p><h3>Where Hindsight earns its place</h3><p>Hermes’s short files keep standing notes, and session search reopens an old conversation when you know what to look for. When a decision changes across several chats, its reasons are spread across those exchanges. Hindsight keeps the source text, <a href="https://hindsight.vectorize.io/blog/2026/07/13/inside-retain-agent-memory">extracts facts, and links related records</a> in a named bank. Hermes works with that bank through a memory provider plugin and three tools: hindsight_retain saves, hindsight_recall searches, and hindsight_reflect reasons across saved memories.</p><p>Hindsight is one of several <a href="https://hermes-agent.nousresearch.com/docs/user-guide/features/memory-providers">memory providers Hermes supports</a>, alongside options such as Honcho and Mem0. Only one external provider can be active at a time, and the built-in files keep working alongside it.</p><p>For example, you start a side project on SQLite and later move to Postgres when concurrent writes become a problem. MEMORY.md can hold the current database choice. Hindsight can retain the earlier decision, the reason for switching, and the conversation where you explained it. When you ask why the database changed, Hermes has a history to draw from.</p><p>The project’s current instructions still belong in the project itself, for example in an AGENTS.md file that Hermes loads from your working directory. The bank can keep the story behind them.</p><h4>A bank you can inspect</h4><p>Open the bank when a reply sounds oddly certain. Constellation, Hindsight’s map of a bank’s memories and their links, showed my saved preference for self-hosted tools and the memories connected to it. That gave me a place to check the record and see whether it still fit the current project.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0qcKQsDNi_Dl5Mam2vFZtg.png" /><figcaption>A saved preference in the Hindsight memory bank.</figcaption></figure><h4>What happens in the next conversation</h4><p>In a later chat, Hermes retrieved that hosting preference from Hindsight. It labeled the preference as the direct answer and treated other results as nearby context. That distinction matters more as a bank accumulates history.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*673001qE-fb-EJa0xCWDug.png" /><figcaption>Hermes retrieves the preference in a new conversation.</figcaption></figure><p>The screenshot shows an explicit hindsight_recall call. Auto Recall can search the bank before ordinary replies.</p><h3>Choose what Hermes saves and recalls</h3><h4>Curate writes, then decide on automatic retention</h4><p>Auto Retain sends conversation turns to Hindsight. Auto Recall fetches relevant memories before later replies. Both are on by default, and you can switch each one independently in the Hindsight provider settings on the Hermes Dashboard’s Plugins page. I’d start with Auto Recall on and Auto Retain off, then save decisions and durable preferences deliberately with hindsight_retain. That way you choose which remarks become durable.</p><p>Automatic retention becomes useful once the bank is part of daily work. After a change in plans, check the bank for the old and new decisions together. Old decisions can sound current when they arrive in a confident answer.</p><h3>Hermes built-in memory, session search, and Hindsight compared</h3><p>When a conversation leaves something worth keeping, I ask what I will need later: a standing fact, the exact exchange, or the history behind a decision.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/c8d548e230f690b7201dbe5c8e9ab825/href">https://medium.com/media/c8d548e230f690b7201dbe5c8e9ab825/href</a></iframe><h3>Self-hosted AI agent memory on Olares</h3><p>Hermes and Hindsight can both run on your own hardware. Hindsight also calls a chat model to process retained information and an embedding model to search it. Keeping the full memory path local means running those models locally too.</p><p>On Olares, Hermes Agent and Hindsight are both available as one-click installs in Market. Hindsight uses local chat and embedding models through <a href="https://www.olares.com/docs/use-cases/olares-router">Router</a>; the Olares guide pairs Qwen3.8–27B for chat with EmbeddingGemma for embeddings. To connect the two, run hermes plugins install hindsight in Hermes CLI, then hermes memory setup, and choose Local External with your Hindsight API URL. The <a href="https://www.olares.com/docs/use-cases/hindsight">Olares Hindsight guide</a> walks through each step, including where to find that URL.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WfpAzH509QVCRCUv.png" /><figcaption>Hindsight is available in Olares Market.</figcaption></figure><h3>Summary</h3><p>Keep standing preferences in Hermes’s short files. Add Hindsight when decisions accumulate across conversations and you want their history available to inspect. Start with deliberate saves, then turn on automatic retention when you want routine conversations added too.</p><h3>FAQ</h3><p><strong>Does Hermes Agent have long-term memory?</strong></p><p>Yes. Hermes keeps two small files, USER.md and MEMORY.md, across sessions and can search past conversations with session search. For long-term memory that extracts and links facts, add an external memory provider such as Hindsight.</p><p><strong>When do changes to Hermes’s memory files take effect?</strong></p><p>Hermes loads its memory files when a session starts. Notes saved during a conversation are written to disk right away but only reach the model’s prompt in the next session. Start a new session to pick them up.</p><p><strong>How do I add Hindsight to Hermes?</strong></p><p>Install the plugin with hermes plugins install hindsight, then run hermes memory setup and select hindsight. You can connect to Hindsight Cloud, a local embedded instance, or an external Hindsight server such as the Hindsight app on Olares.</p><p><strong>Can Hindsight run fully locally?</strong></p><p>Yes, as long as its models run locally too. Hindsight needs a chat model to process memories and an embedding model to search them. On Olares, it uses local models from Router, such as Qwen3.8–27B and EmbeddingGemma.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7437f1ed54c3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Olares 1.12.8: Apps on Your Phone, Wise Returns, and File Access Controls]]></title>
            <link>https://olares.medium.com/olares-1-12-8-apps-on-your-phone-wise-returns-and-file-access-controls-14d75ea4bdb4?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/14d75ea4bdb4</guid>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[new-releases]]></category>
            <category><![CDATA[olares]]></category>
            <category><![CDATA[private-cloud]]></category>
            <category><![CDATA[self-hosting]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Wed, 18 Nov 2026 12:08:27 GMT</pubDate>
            <atom:updated>2026-11-19T07:19:22.365Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6EQtvZkE6zbUbnIP.jpg" /></figure><p>Olares 1.12.8 puts your Olares apps in LarePass on your phone, brings Wise back to Market as a personal reading library, and lets administrators decide who can view or edit External mounts and Common folders in Files.</p><p>Alongside these additions, GPU apps spend less time waiting for access to a shared GPU. File operations also become more reliable, with fixes for folder uploads, paused transfers, cloud video playback, and unavailable external storage.</p><p>Read on for the highlights, or jump to <a href="https://www.olares.com/blog/olares-1-12-8/#how-to-update-to-1-12-8">How to update to 1.12.8</a>. For the full list of changes, see the <a href="https://github.com/beclab/Olares/releases/tag/1.12.8">release notes</a>.</p><h3>At a glance</h3><ul><li>LarePass adds an Apps tab for opening and managing Olares apps from your phone, much like Launchpad on the desktop.</li><li>Wise brings saved webpages, books, PDFs, media, and RSS subscriptions into one reading library, with notes and filtered views to organize what you keep.</li><li>Files adds per-member view and edit permissions for External mounts and top-level Common folders.</li><li>GPU switching in Time slicing mode drops from about 4–5 seconds to 0.1–1.5 seconds, reducing the pause when apps take turns on the same GPU.</li></ul><h3>LarePass brings app access and management to your phone</h3><p>The new <strong>Apps</strong> tab brings the Launchpad you know from the desktop to LarePass, with Market and your installed apps in a single view. The apps still run on Olares while you use them from your phone. Every Olares system app now works on mobile, which makes LarePass a complete way to reach your Olares when you are not at a computer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*0t-Mi4HHjIsHZwyi.png" /><figcaption>Olares apps and Market, accessible from the Apps tab in LarePass</figcaption></figure><p>Wherever you are, you can open the apps you use, browse Market, install something new, and check on or restart what is running, all from one tab on your phone. Each app shows its status and management controls right beside it. Using Olares and looking after it are no longer separate trips.</p><p>System apps have been reworked for smaller screens, with fixes to layouts, scrolling, and touch interactions. Ticket now signs you in to Olares Space with the identity already in LarePass. You no longer have to scan a login code with the phone in your hand.</p><p>See <a href="https://www.olares.com/docs/manual/larepass/manage-apps-with-larepass">Manage Olares apps from your phone</a> for setup and available actions.</p><h3>Wise turns saved reading and feeds into one library</h3><p>Many of us split our reading across a read-it-later app, an RSS reader, and a folder of PDFs. Wise is back in Market to bring them together on your own Olares. Saved webpages, books, PDFs, and media share a library with the feeds you follow, and everything you keep stays on hardware you own.</p><p>While researching a project, you might save an article today and return to it next week. Wise keeps your notes and reading progress with the article, and you pick up where you stopped. Pages captured with the <a href="https://www.olares.com/docs/manual/olares/wise/collect">LarePass browser extension</a> are filed alongside the PDFs and books you have added yourself.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZFeMju-7R_w76a6o.png" /><figcaption>Collect a webpage, download an available file, or subscribe to its RSS feed with the LarePass browser extension</figcaption></figure><p>Meanwhile, the sources you follow are still publishing. <a href="https://www.olares.com/docs/manual/olares/wise/subscribe">RSS subscriptions</a> bring their new posts into Feeds, and anything worth keeping joins the rest of your material.</p><p>As the library grows, a <a href="https://www.olares.com/docs/manual/olares/wise/basics#use-filtered-views">filtered view</a> gathers a project’s reading across saved items and feeds. Matching content is added as it arrives, giving the topic a reading list that grows with it. Your earlier notes are waiting when you come back.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eCDxa9bklf6oWoR4.png" /><figcaption>Organize saved articles with tags and keep a filtered view in the sidebar</figcaption></figure><p>See the <a href="https://www.olares.com/docs/manual/olares/wise/">Wise documentation</a> to start building your library.</p><h3>Decide who can view or edit external storage and Common folders</h3><p>External mounts and Common folders give everyone on the same Olares one set of files to work from. Until now, those locations were open to every member, whatever their role in a project. In 1.12.8, administrators can grant each member view or edit permission based on the work they do.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*uULThNOJjFcqnadK.png" /><figcaption>Access settings for a mounted SMB share in Files</figcaption></figure><p>Consider a project kept on a <a href="https://www.olares.com/docs/manual/olares/files/mount-SMB">mounted NAS share</a>. Contributors add files and revise the work, while others only need to check the results. Both groups open the same project folder in Files. Contributors can edit, and everyone else can view. Nobody has to maintain a second copy just to keep the originals safe.</p><p>We plan to refine permission settings further, making Olares easier to share among users with different roles. See <a href="https://www.olares.com/docs/manual/olares/files/manage-common-external-access">Manage access to Common and External</a> for the current permission model.</p><h3>Other enhancements</h3><h4>Faster switching between GPU apps</h4><p>In <a href="https://www.olares.com/docs/manual/olares/settings/gpu-resource#gpu-modes">Time slicing mode</a>, a chat model and an image-generation app take turns using the same GPU. Moving between them involves handing GPU access from one app to the other, adding a pause before the next app can start its work.</p><p>Olares 1.12.8 updates HAMi and the GPU scheduler to release GPU access more efficiently between apps. Switching time drops from about <strong>4–5 seconds to 0.1–1.5 seconds</strong>, shortening that pause as you move between tasks.</p><h4>More reliable file operations</h4><p>Several fixes address interruptions in everyday file work. Uploading a folder with nested subfolders now creates the missing directories correctly. Cancel all also cancels paused transfer tasks and clears their partial output, instead of leaving them paused in the task list.</p><p>Files fixes video playback from Google Drive and improves how it handles unavailable external storage. When a mount is already reported as unavailable, opening it now returns an error promptly, and the mount can still be removed. This addresses the case where a disconnected NAS left both browsing and unmounting stuck.</p><h3>New in Market</h3><p>The latest Market additions range from research and agent tools to music creation and services you can host on Olares.</p><h4>Research and reading</h4><ul><li><a href="https://www.olares.com/docs/use-cases/local-deep-research"><strong>Local Deep Research</strong></a><strong>:</strong> Research topics across multiple sources and generate reports with citations using your choice of local or cloud language models.</li><li><a href="https://www.olares.com/docs/use-cases/linkwarden"><strong>Linkwarden</strong></a><strong>:</strong> Collect and organize bookmarks, preserve copies of webpages, and share collections with others.</li></ul><h4>Agents and AI tools</h4><ul><li><a href="https://www.olares.com/docs/use-cases/goose"><strong>Goose</strong></a><strong>:</strong> Run an AI agent on Olares for coding, research, writing, and automation, and connect from the native Goose Desktop client.</li><li><strong>Agent J:</strong> Control an AI coding agent running on Olares from your phone, with messages, task progress, and approval prompts.</li><li><a href="https://www.olares.com/docs/use-cases/hindsight"><strong>Hindsight</strong></a><strong>:</strong> Give AI agents persistent memory to retain experiences, recall relevant facts, and reflect on previous interactions.</li><li><a href="https://www.olares.com/docs/use-cases/mempalace"><strong>MemPalace</strong></a><strong>:</strong> Share local memory across AI agents, with original-text storage, semantic search, and MCP access for saving and recalling information.</li><li><a href="https://www.olares.com/docs/use-cases/trustgraph"><strong>TrustGraph</strong></a><strong>:</strong> Turn documents into searchable knowledge graphs and ask questions through document retrieval or Graph RAG.</li><li><a href="https://www.olares.com/docs/use-cases/cognee"><strong>Cognee</strong></a><strong>:</strong> Turn documents and text into searchable knowledge graphs, and connect AI assistants through MCP.</li><li><a href="https://www.olares.com/docs/use-cases/unsloth"><strong>Unsloth</strong></a><strong>:</strong> Run and fine-tune open models in your browser, prepare training datasets, and export trained models, with JupyterLab included.</li></ul><h4>Music creation</h4><ul><li><a href="https://www.olares.com/docs/use-cases/yue2-studio"><strong>YuE2 Studio</strong></a><strong>:</strong> Generate songs from lyrics and a musical style on your own GPU, then edit the score to create another performance.</li><li><a href="https://www.olares.com/docs/use-cases/music-studio"><strong>Music Studio</strong></a><strong>:</strong> Create songs and instrumental music from ideas and lyrics with ACE-Step, regenerate selected sections, and organize finished tracks in your library.</li><li><strong>UVR5 UI:</strong> Separate recordings into vocals, accompaniment, and other tracks through a community-contributed web interface, with results saved to your mounted NAS.</li></ul><h4>Self-hosted services</h4><ul><li><a href="https://www.olares.com/docs/use-cases/qemu"><strong>QEMU Server</strong></a><strong>:</strong> Run a Linux server virtual machine on Olares, with a choice of distributions, SSH access, and a browser-based console.</li><li><a href="https://www.olares.com/docs/use-cases/plex"><strong>Plex Media Server</strong></a><strong>:</strong> Organize your video and music collection on Olares and play it in a browser or Plex client.</li><li><a href="https://www.olares.com/docs/use-cases/pterodactyl"><strong>Pterodactyl Panel</strong></a><strong>:</strong> Manage game server consoles, files, schedules, backups, and user permissions from your browser.</li><li><a href="https://www.olares.com/docs/use-cases/rustdesk"><strong>RustDesk Server</strong></a><strong>:</strong> Host the ID and relay services for RustDesk remote desktop connections, with persistent server keys and a connection setup guide.</li></ul><h4>Personal tools and discovery</h4><ul><li><strong>God’s Eye View:</strong> Explore live aircraft, ships, satellites, and public camera feeds on an interactive 3D globe.</li><li><strong>Mindwtr:</strong> Organize tasks and projects with a GTD workflow and synchronize them across devices through your own Olares-hosted service.</li><li><strong>OpenTodo:</strong> Manage tasks with a self-hosted service that supports CalDAV calendar sync, webhooks, and MCP access for AI assistants.</li><li><strong>NeoDB:</strong> Track, review, and share books, movies, music, and games, with collection imports and connections to the Fediverse.</li><li><strong>udctl:</strong> Bring tasks, notes, files, finances, and AI agent teams into one private workspace.</li></ul><p>Thank you to the community contributors who bring more applications to Olares. To submit an application, see the <a href="https://www.olares.com/docs/developer/develop/submit-apps">app submission guide</a>.</p><h3>How to update to 1.12.8</h3><p>Before you begin, read the <a href="https://www.olares.com/docs/manual/update-guides/1.12.8">Olares 1.12.8 update guide</a> for important changes and any actions needed before or after the update.</p><ol><li>Update LarePass to version <strong>1.12.35</strong>.</li><li>Open LarePass and go to <strong>Settings</strong> &gt; <strong>System</strong> &gt; <strong>System update</strong>.</li><li>Check that <strong>New version</strong> shows <strong>1.12.8</strong>, then tap <strong>Update</strong>.</li><li>Choose an update method and confirm the installation.</li><li>Wait for Olares to finish updating and restart, then refresh your Olares desktop.</li></ol><h3>What’s next</h3><p>Your feedback helps us decide which parts of Olares need more attention. Tell us how the mobile app experience, Wise, and file access permissions work in your setup on the <a href="https://www.olares.com/forum/">Olares Forum</a> or <a href="https://discord.gg/olares">Discord</a>. For device-specific help, contact us through <a href="https://www.olares.com/docs/manual/help/request-technical-support">Ticket</a>.</p><p><em>Originally published at </em><a href="https://www.olares.com/blog/olares-1-12-8/"><em>https://www.olares.com</em></a><em> on November 18, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=14d75ea4bdb4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Run Qwen3.8-Flash-Next Locally on a 24GB GPU with Strata]]></title>
            <link>https://olares.medium.com/run-qwen3-8-flash-next-locally-on-a-24gb-gpu-with-strata-1d1c5c371b8f?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/1d1c5c371b8f</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[local-llm]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[self-hosted]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 14:00:24 GMT</pubDate>
            <atom:updated>2026-11-16T06:08:35.829Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GEE7QjEJ1bXrz2E8" /></figure><p>Running <a href="https://huggingface.co/Qwen/Qwen3.8-Flash-Next">Qwen3.8-Flash-Next</a> at home sounds ambitious. Its main language model has 125 billion parameters, which would occupy about 250GB at 16-bit precision.</p><p>Our setup was an Olares One with a 24GB GPU and 96GB of RAM. Strata made the rest of the machine part of the equation, drawing on the CPU and SSD to help run a more compact version of the model.</p><h3>How Strata makes room for Flash-Next</h3><p>Flash-Next uses about 6 billion of its 125 billion parameters to generate each piece of an answer. Much of the model is divided into smaller networks called experts, and it selects which ones to use as it writes. This design is known as mixture of experts, or MoE.</p><p><a href="https://github.com/Niko1221/Strata/blob/v0.1.42/docs/HOW_IT_WORKS.md">Strata</a> takes advantage of that selective workload. It keeps frequently used experts in GPU memory, where they can run quickly. The full set lives in system RAM. When an expert outside the GPU’s cache is needed, the CPU handles it while the GPU works on its own share.</p><p>The SSD has a role during generation, too. Flash-Next includes a large lookup table for short sequences of text. Strata keeps that table on disk and reads the entries it needs as the model writes, leaving more memory available for the rest of the model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*DtTdKXp2v5FhLoer.png" /><figcaption>How Strata uses the GPU, CPU, RAM, and SSD to run Flash-Next.</figcaption></figure><h3>Flash-Next in practice</h3><p>The <a href="https://www.olares.com/market/app/market.olares/strataqwen38iq4xsv3">Flash-Next app in Olares Market</a> pairs Strata with Unsloth’s UD-IQ4_XS version of the model. This version uses quantization, storing each parameter with fewer bits, roughly four, to reduce the memory needed to hold the model.</p><p>The app reserves 23 GiB of GPU memory, 65 GiB of system RAM, and 120 GiB of disk space. On our <a href="https://www.olares.com/docs/one/spec">Olares One</a>, that takes up most of the 24GB GPU memory and roughly two-thirds of the 96GB RAM.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*x-cBPmjy-85Pwdsz.png" /><figcaption>The Flash-Next app in Olares Market, with its resource requirements.</figcaption></figure><h4>A report in about 21 seconds</h4><p>We gave Flash-Next 96 fictional service records and asked it to turn them into an English operations report. With thinking off, the report took about 21 seconds on average, from sending the request to receiving the full answer.</p><p>As it wrote, generation stayed around 71–73 tokens per second. Tokens are the small pieces of text a model produces, so this number tells you how quickly it writes.</p><p>We then turned thinking up to xhigh, giving the model more room to reason before answering. Generation climbed to about 100 tokens per second, but the average wait at the largest output limit stretched to about 90 seconds. More of that output went into reasoning. The model was generating faster, while taking longer to finish the job.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/6d518ce4d2a74d340737d0722bc48a83/href">https://medium.com/media/6d518ce4d2a74d340737d0722bc48a83/href</a></iframe><p><em>Each row averages three requests with the model already warmed up. No prompt-cache hits were reported across the 18 requests.</em></p><h4>Deploying Wallos with Lares</h4><p>Next, we connected Flash-Next to <a href="https://www.olares.com/docs/use-cases/lares">Lares</a>, the Olares AI assistant, to deploy <a href="https://github.com/ellite/Wallos">Wallos</a>, a self-hosted subscription tracker. We supplied the project’s repository and Docker image, then asked:</p><blockquote><em>Port Wallos to Olares and install it so I can open it from the Olares desktop.</em></blockquote><p>We began with a fresh workspace and a restarted model. Flash-Next worked out the steps, and Lares carried out the commands to package and install Wallos.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zhbjjPGFBG34izF4.png" /><figcaption>Lares completes the Wallos installation with Flash-Next.</figcaption></figure><p>In 13 minutes and 22 seconds, Wallos was ready to open from the Olares desktop. Lares had also restarted the app and checked that its data stayed in place.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mT6vUZwxGBc3rur3.png" /><figcaption>The installed Wallos app, ready to track subscriptions.</figcaption></figure><h3>A smaller model for everyday work</h3><p>For this job, the smaller <a href="https://www.olares.com/market/app/market.olares/llamacppqwen3827bggufv3">Qwen3.8–27B</a> kept pace with Flash-Next. Through Lares on the same Olares One, it deployed Wallos in 13 minutes and 8 seconds at its default thinking setting, just 14 seconds ahead.</p><p>The bigger change came when we turned 27B’s thinking setting down to low. This time, the same job took 5 minutes and 4 seconds. That cut about eight minutes from the wait.</p><p>27B becomes even more appealing when you consider what it leaves available for the rest of your machine. Its app reserves 8 GiB of system RAM, compared with Flash-Next’s 65 GiB, leaving much more room for the other apps you keep running. That smaller memory requirement, together with its performance on the Wallos task, makes it our recommended starting point for everyday agent work.</p><p>With 27B connected to Lares, you can also move from an agent task to creative work on the same GPU. Ask Lares to deploy an app, then create an image or video through <a href="https://www.olares.com/docs/use-cases/flowstudio-lares">FlowStudio</a>. One machine can handle both running your apps and creating content for your projects.</p><h3>Try Flash-Next for yourself</h3><p>To put Flash-Next to work on your own projects, start with the Olares Market app:</p><ol><li>Open <a href="https://www.olares.com/market/app/market.olares/strataqwen38iq4xsv3">Qwen3.8-Flash-Next (Strata)</a> in Olares Market. Click <strong>Get</strong>, then <strong>Install</strong>.</li><li>Click <strong>Open</strong> to view Model Console.<br>Leave the app running while it downloads about 94GB of model weights, prepares the Strata engine, and fetches the additional files for multi-token prediction (MTP), which helps speed up generation. This first setup can take a while, especially on a slower connection.<br>After the main download finishes, wait for <strong>Model</strong> to show <strong>Ready</strong> and <strong>Engine</strong> to show <strong>Running</strong>.</li><li>Open Lares, select Flash-Next, and start a conversation or give it a task.<br>To use Flash-Next from an AI app on your own computer, follow the <a href="https://www.olares.com/docs/use-cases/olares-router">Olares Router documentation</a> to get the Base URL, model name, and API key.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*oJrTat95N8fyiIuU.png" /><figcaption>Flash-Next running locally with the Strata engine.</figcaption></figure><h3>Conclusion</h3><p>With Strata and Olares One, a 125-billion-parameter model becomes something you can put to work at home. Bring your own questions and projects, and see what it can do on hardware you own.</p><p>We’ll be testing more models on Olares One, including models for images, video, and other modalities. Stay tuned for more benchmarks and hands-on reviews.</p><p><em>Originally published at </em><a href="https://www.olares.com/blog/run-qwen3-8-flash-next-24gb-gpu/"><em>https://www.olares.com</em></a><em> on November 5, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1d1c5c371b8f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Inside Olares: How your credentials work together]]></title>
            <link>https://olares.medium.com/inside-olares-how-your-credentials-work-together-c2b1b5f68987?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/c2b1b5f68987</guid>
            <category><![CDATA[password-management]]></category>
            <category><![CDATA[data-privacy]]></category>
            <category><![CDATA[olares]]></category>
            <category><![CDATA[digital-identity]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Fri, 16 Oct 2026 15:46:06 GMT</pubDate>
            <atom:updated>2026-10-19T12:17:05.996Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VaSr-EdfGy7ZtuLDXxK4ug.png" /></figure><p>If you have used 1Password, you may already be familiar with the idea of keeping passwords in an encrypted vault. Olares also uses a <a href="https://www.olares.com/docs/manual/olares/vault/vault-items"><strong>vault</strong></a>, but getting into it is only one part of the story. You may unlock <a href="https://olares.com/larepass"><strong>LarePass</strong></a> with a local password, sign in to Olares Desktop with another password, and use a code or phone confirmation when two-factor authentication is enabled.</p><p>These credentials protect different things. Knowing which one you need matters when setting up a new client, signing in, or recovering access. This article shows where each credential is used, how Vault keeps saved items in sync, and what to back up before a device is lost.</p><h3>Five credentials, five roles</h3><p>The diagram brings together five types of credentials used for identity recovery, local unlocking, Olares sign-in, second-factor verification, and host access. Each has a different role.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9SdoB_zhyLujCHClxlx8Zw.png" /><figcaption>Where each credential is used and protected</figcaption></figure><h3>Mnemonic phrase</h3><p>When you <a href="https://www.olares.com/docs/manual/get-started/create-olares-id"><strong>create an Olares ID</strong></a> using LarePass, it generates a 12-word mnemonic phrase locally. These words can recreate the cryptographic keys that determine its decentralized identifier (DID), the underlying identifier linked to the Olares ID. Using the same phrase on another device lets you import the same identity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BOxhBjgeuF3JJzsd3g48BA.png" /><figcaption>The same mnemonic phrase restores the same Olares identity</figcaption></figure><p>LarePass stores the phrase encrypted locally, outside Vault, protected by the local password. The phrase also helps unlock the associated Vault data, provided that encrypted data remains available.</p><p><strong>Back up the phrase separately and keep it private.</strong> If no accessible client or backup remains, Olares cannot recreate the original phrase.</p><h3>LarePass local password</h3><p>The local password protects the mnemonic phrase and access to Vault and other protected features within a particular LarePass client. Each client or browser extension has its own local password, which does not sync through Vault.</p><p>Setting it gives the client a local lock after an identity has been imported. On mobile, biometric unlock can be enabled after a local password is set. It offers a convenient way to unlock protected features and can also reveal the local password after successful biometric verification.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yNeT3YtPYt0UtxdQlI-6dg.png" /><figcaption>Set a local password and use biometric verification in LarePass Mobile</figcaption></figure><h3>Olares login password</h3><p>This password signs you in to Olares Desktop. LarePass may also request it when it needs to authenticate with Olares again.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3na8R2wkraNA3xohtvbS0Q.png" /><figcaption>Sign in to Olares Desktop</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vX1xC6qnAvEuMPoZqrBpeA.png" /><figcaption>Log in to Olares from LarePass Desktop</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YZBZcninuRh1_PVqxk5Iiw.png" /><figcaption>Log in to Olares from LarePass Mobile</figcaption></figure><p>When Olares asks you to sign in, use the Olares login password whether the prompt appears in Olares Desktop or LarePass. Olares authentication services verify it. This is different from the local password used to unlock a particular LarePass client.</p><h3>TOTP secret and 6-digit code</h3><p>During Olares activation, a TOTP secret is created for that installation and saved in Vault. LarePass uses the secret to generate a temporary six-digit code that changes regularly. The secret is stored and synced as encrypted Vault data. The code itself is generated when needed. Activating the same Olares ID on another Olares device creates a new TOTP secret for that installation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IKOGbSVZcDm0NLav.png" /><figcaption>Olares Desktop asks for a six-digit code when 2FA is enabled</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wj6X90kOu3MSq6Pc.png" /><figcaption>Olares Desktop asks you to confirm the sign-in in LarePass Mobile when 2FA is enabled</figcaption></figure><p>When two-factor authentication is enabled, Olares Desktop requires an additional check after the login password. You can enter the current code from LarePass or confirm the sign-in in LarePass Mobile. When signing in to Olares Desktop through LarePass, the client can generate and submit the required TOTP code automatically, so you may not see a separate code prompt.</p><h3>Host / SSH password</h3><p>The host password gives access to the underlying Linux system through SSH or a physical terminal. It is separate from the Olares login password and is managed by the host operating system.</p><p>On Olares One, a copy is saved in Vault for retrieval. Vault holds the copy; the host operating system verifies the password when you <a href="https://www.olares.com/docs/one/access-terminal-ssh"><strong>access the terminal</strong></a>.</p><h3>How Vault keeps encrypted data in sync</h3><p>Vault keeps saved passwords, TOTP secrets, and other sensitive items available across configured LarePass clients. Like 1Password’s encrypted syncing,<strong> it encrypts item contents on a client before syncing them</strong> and decrypts them on another client when unlocked. In Olares, the Vault server that stores and syncs the encrypted data runs on your Olares device.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hDxBofptq6w2FVrScaZEhQ.png" /><figcaption>Vault items are encrypted on one client and decrypted on another</figcaption></figure><p>When an item changes, the Vault server makes its encrypted version available to other clients. It does not need a readable copy of the item to keep those clients in sync.</p><p><strong>The mnemonic phrase stays outside Vault sync by default.</strong> LarePass keeps it encrypted locally, rather than storing it on Olares with synced Vault data. If the device is lost, you can still use a separate mnemonic backup or an accessible LarePass client.</p><p>Vault items, including the TOTP secret, are a separate matter: recovering them still requires an available copy of the encrypted data. <strong>Sync does not replace an independent backup.</strong></p><h3>Simple habits for safer access to Olares</h3><p>A few preparations help protect everyday access and preserve recovery options.</p><ul><li><strong>Back up the mnemonic phrase offline.</strong> Keep every word in the correct order, somewhere private, readable, and accessible over time. Avoid keeping the only backup on the device it is meant to help recover.</li><li><strong>Protect each LarePass client.</strong> Set a strong local password and keep a secure record of it. Another trusted client provides an additional access point, but does not replace an independent mnemonic backup.</li><li><strong>Check before resetting or replacing a device.</strong> Confirm that both the mnemonic backup and important data backups are available before removing an Olares ID, uninstalling LarePass, or reinstalling Olares.</li></ul><p>The phrase deserves special care because someone who obtains it can recreate the underlying identity keys. Its backup needs to be protected from both loss and unwanted access.</p><h3>Lost a credential? Start here</h3><p>The next step depends on what is missing and what remains accessible. These credentials have different recovery paths; they do not all replace one another.</p><p><strong>Mnemonic phrase</strong></p><p>Check an existing LarePass client where the Olares ID remains accessible, then reveal and <a href="https://www.olares.com/docs/manual/larepass/back-up-mnemonics#back-up-mnemonic-phrase"><strong>back up the phrase</strong></a>. Without an accessible client or a backup, the phrase cannot be recovered.</p><p><strong>LarePass local password</strong></p><p>On mobile, previously enabled biometric unlock can reveal it. The desktop client cannot reveal a forgotten local password. If this route is unavailable, check another accessible client for the mnemonic phrase and follow the recovery guide before removing anything.</p><p><strong>Olares login password</strong></p><p>An administrator with host terminal access can follow the <a href="https://www.olares.com/docs/manual/help/ts-forget-login-password?utm_source=chatgpt.com"><strong>password reset guide</strong></a>.</p><p><strong>6-digit code</strong></p><p>Open an accessible LarePass client containing the corresponding authenticator. If a sign-in notification does not arrive, switch to code verification.</p><p><strong>Host / SSH password</strong></p><p>For Olares One, check the saved Vault entry. If Olares Desktop remains accessible with the required permissions, follow the <a href="https://www.olares.com/docs/one/access-terminal-ssh"><strong>SSH password reset steps</strong></a>. Other installations use their host system’s recovery process.</p><p>See the <a href="https://www.olares.com/docs/manual/help/ts-access-without-mnemonic?utm_source=chatgpt.com"><strong>access recovery guide</strong></a> for platform-specific steps and conditions.</p><p><strong>Before removing an Olares ID or uninstalling LarePass, make sure the mnemonic phrase is safely backed up.</strong> An existing client may hold the only remaining copy.</p><p>Like 1Password, Olares relies on client-side encryption to keep synced Vault contents unreadable to the server. But protecting the data is only part of staying in control of it. Keep an independent backup of the mnemonic phrase, maintain another accessible LarePass client when possible, and back up important data separately. That way, you are prepared not just to keep credentials private, but to use the right one and regain access when circumstances change.</p><p><em>Originally published at </em><a href="https://www.olares.com/blog/olares-credentials-explained/"><strong><em>https://www.olares.com</em></strong></a><em> on October 16, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c2b1b5f68987" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Olares Showcases Private AI at Intel’s AI NAS and Thunderbolt™ Technology Forum in Japan]]></title>
            <link>https://olares.medium.com/olares-showcases-private-ai-at-intels-ai-nas-and-thunderbolt-technology-forum-in-japan-78a939abd63a?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/78a939abd63a</guid>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[intel]]></category>
            <category><![CDATA[olares]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 13:00:42 GMT</pubDate>
            <atom:updated>2026-10-19T12:18:00.172Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZjD9IF98Zleztm7b" /></figure><p><em>The open-source platform brings agentic AI to hardware people own, giving users greater control over how their AI runs and how their data is used.</em></p><p>TOKYO, October 15, 2026 — Olares, the company behind the open-source <a href="https://www.olares.com/os"><strong>Olares OS</strong></a>, today showcased its operating system on Intel® Core™ Ultra-powered AI NAS at Intel’s AI NAS &amp; Thunderbolt™ Technology Forum in Tokyo. The showcase highlighted how users can run AI agents that work with their files and applications on hardware they own.</p><p>Running a model is only one part of building a useful personal AI system. Agents also need access to the files, applications and tools involved in a task. Olares OS brings these resources together in one environment, helping users put AI agents to work in their everyday workflows.</p><p>“Choosing a personal AI system is a long-term decision,” said Peng Peng, Founder and CTO of Olares. “People should be able to change the models they use, add applications and decide where their data is processed as their needs evolve. That flexibility is central to how we’re building Olares OS.”</p><h3>Agents, models and apps in one personal cloud</h3><p>Key capabilities of Olares OS include:</p><ul><li><strong>Shared access to AI models.</strong> Olares Router gives agents and applications a common gateway to local models and other configured AI services, with model connections and access controls managed in one place.</li><li><strong>Task execution in plain language.</strong> <a href="https://www.olares.com/blog/meet-lares-olares-native-agent/"><strong>Lares</strong></a>, the Olares-native AI agent, uses configured models and tools to carry out multi-step tasks across files, applications and system services. Users can ask Lares to handle everyday tasks such as organizing files and checking system status. It can also take on more complex work, including deploying applications, conducting in-depth research, and generating images and videos through FlowStudio, an AI media creation app.</li><li><strong>An ecosystem of more than 200 apps.</strong> <a href="https://www.olares.com/market/"><strong>Olares Market</strong></a> offers one-click installation of apps and supported AI models. Users can deploy models such as Qwen3.8–27B alongside apps including Home Assistant for smart home integration and n8n for workflow automation.</li><li><strong>AI across work, creativity and home.</strong> From transcribing meeting recordings to generating images, videos and music locally, Olares supports a range of personal AI workflows. With the relevant integrations configured, users can also ask Lares to control connected smart home devices in natural language.</li></ul><p>These capabilities are accessible across devices, at home or on the go, while the underlying apps, local models and data remain on users’ own hardware.</p><p><a href="https://www.olares.com/one"><strong>Olares One</strong></a>, the company’s personal AI cloud device, pairs an Intel Core Ultra 9 275HX processor with an NVIDIA GeForce RTX 5090 Mobile GPU to support demanding local generative AI workloads. Olares OS can also be installed on other compatible systems, with supported workloads depending on the hardware configuration.</p><p>“Home is the new edge,” said Gordon Sylin, Global Head of NAS at Intel. “The AI NAS is growing from the place you store your life into the place your agents live, and Intel Core Ultra was built for exactly this: edge AI that runs all day, at home, on hardware people own. Olares OS shows what our partner ecosystem can deliver when the operating system is designed for agents from day one.”</p><h3>An open platform for partners</h3><p>Olares is inviting AI NAS manufacturers, original design manufacturers (ODMs) and system integrators to explore Olares OS for compatible Intel Core Ultra-based systems. Its open-source software and application ecosystem provide a foundation for products that combine personal storage with AI agents and applications.</p><p>Olares One raised $2.34 million on Kickstarter in January 2026 and has fulfilled all backer orders. To learn more about Olares OS and explore its open-source code, visit <a href="https://www.olares.com/"><strong>olares.com</strong></a> and <a href="https://github.com/beclab/Olares"><strong>github.com/beclab/Olares</strong></a>.</p><h3>Availability</h3><p>Olares One is available at <a href="https://shop.olares.com/products/olares-one"><strong>shop.olares.com</strong></a>, with Black Friday offers planned for November.</p><p><strong>About Olares</strong></p><p>Founded in 2022 and headquartered in Singapore, Olares develops personal AI cloud software and devices. Its open-source operating system, Olares OS, enables individuals and teams to run AI models, agents and applications on their own hardware, with control over their data. Olares One is the company’s personal AI cloud device, built to run Olares OS. Olares has raised more than $100 million, including a Series A led by SIG. Learn more at and <a href="https://github.com/beclab/Olares"><strong>github.com/beclab/Olares</strong></a>.</p><p><em>Source: This announcement was first released on </em><a href="https://www.prnewswire.com/news-releases/olares-showcases-private-ai-at-intels-ai-nas--thunderbolt-technology-forum-in-japan-302909516.html"><strong><em>PR Newswire</em></strong></a><em>.</em></p><p><em>Originally published at </em><a href="https://www.olares.com/blog/olares-intel-ai-nas-forum-japan/"><em>https://www.olares.com</em></a><em> on October 15, 2026.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78a939abd63a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Run MiniMax H3 Locally: Image to Video with FlowStudio]]></title>
            <link>https://olares.medium.com/run-minimax-h3-locally-image-to-video-with-flowstudio-e90a9671aa83?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/e90a9671aa83</guid>
            <category><![CDATA[olares]]></category>
            <category><![CDATA[image-to-video]]></category>
            <category><![CDATA[minimax-h3]]></category>
            <category><![CDATA[flowstudio]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 13:00:50 GMT</pubDate>
            <atom:updated>2026-10-14T11:38:46.997Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w0goW6Y997l3j0URfnky7Q.png" /></figure><p>You can run MiniMax H3 locally to turn an image into video with sound. On Olares, FlowStudio provides ready-made workflows and prepares the models and runtime needed to use them.</p><p>FlowStudio is an app for generating images, editing images, and creating videos. It turns ComfyUI workflows into reusable scenes, with a form for your prompt, reference files, and generation settings.</p><p>Here is one example made in FlowStudio. We generated a landscape image, then used MiniMax H3 to create a short video with movement and ambient sound.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F6EMZ9iLnVxo%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D6EMZ9iLnVxo&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F6EMZ9iLnVxo%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/96716039a7ca3daa02b79c94b0c67f0d/href">https://medium.com/media/96716039a7ca3daa02b79c94b0c67f0d/href</a></iframe><h3>What FlowStudio handles for a local H3 setup</h3><p><a href="https://docs.comfy.org/tutorials/video/minimax/minimax-h3"><strong>ComfyUI already supports MiniMax H3</strong></a><strong> </strong>with templates for text, image, and reference-based video generation. FlowStudio builds on ComfyUI workflows and manages their setup on Olares.</p><p>Choose a ready-made workflow in <strong>Workflow Plaza</strong> and create a scene. FlowStudio prepares its required models and runtime. Once initialization finishes, the scene is available for generation.</p><ul><li><strong>Model preparation.</strong> The scene identifies and downloads the model files required by its workflow.</li><li><strong>Runtime setup.</strong> FlowStudio prepares the environment that runs the workflow.</li><li><strong>Repeat use.</strong> Open the prepared scene to change inputs and generate again without rebuilding its node graph.</li></ul><p>The generation workload runs on your Olares device. You still need enough GPU memory and storage for the selected scene, and the first run includes downloads and initialization.</p><h3>Choose how the image guides your video</h3><p>An image can play different roles in an H3 workflow. Choose based on whether you need an exact opening frame or visual references for a new shot.</p><p><strong>For a fixed opening frame:</strong> FlowStudio offers <strong>FastVideo FastH3: Image to Video</strong>. Supply a starting image and, optionally, an ending image. The output follows the starting image’s aspect ratio.</p><p><strong>For reference-based generation:</strong> choose <strong>MiniMax H3 NVFP4 · Reference to Audio-Video</strong>. It accepts image, video, and audio references to guide the result. A reference image is not a locked first frame.</p><p>The distinction matters for a product photo or an approved composition. If the clip must open on that exact image, choose the first-frame workflow.</p><h3>Our example: one landscape image, then video</h3><p>We created this source image with <strong>Z-Image-Turbo Int8: Text to Image</strong> in FlowStudio. You can also start with an image you already have.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*H7twD_Hyj1S7lTyU.jpg" /><figcaption>The source image for our MiniMax H3 video.</figcaption></figure><p>For the video, we used <strong>MiniMax H3 NVFP4 · Reference to Audio-Video</strong> and supplied only the landscape image. The reference video and audio inputs were left empty.</p><p>The requested result was a calm lakeside shot with rippling water, moving flowers, drifting mist, and birdsong. H3 generated the video and audio together. The clip at the top shows the output from this workflow.</p><p>The image and video are separate outputs. You can keep the still and use it as the reference for another generation when you want to try a different result.</p><h3>You can also request generation through Lares</h3><p>You can work directly in a FlowStudio scene, or describe what you want in Lares. Router connects Lares to the available FlowStudio capabilities, and FlowStudio runs the generation task.</p><p>Prepare the required scenes first. Then ask Lares which generation capabilities are available and describe your request. You can name a model or let Lares choose from the available options.</p><p>The <a href="https://www.olares.com/docs/use-cases/flowstudio-lares"><strong>Lares integration guide</strong></a> covers the required apps and local model setup. The landscape demo above used FlowStudio directly, so this is a separate way to use its generation capabilities.</p><h3>What to expect when running MiniMax H3 locally</h3><p>Resolution, clip length, reference inputs, and available memory affect what your device can run. Use the scene’s <strong>Pre-run checks</strong> to check resources before initialization.</p><p>For the reference workflow, the current Olares guide recommends 480p. Start with a short clip to check the result before spending more time on a longer generation.</p><p>Local H3 output also differs from MiniMax’s full hosted 2K pipeline. The <a href="https://huggingface.co/MiniMaxAI/MiniMax-H3"><strong>official model card</strong></a><strong> </strong>describes a 768p base model and separate hosted stages for the complete 2K workflow.</p><p>To get started, use Olares OS 1.12.7 or later and install or update FlowStudio from Market. The <a href="https://www.olares.com/docs/use-cases/flowstudio-create"><strong>FlowStudio setup guide</strong></a> covers installation, scene preparation, and generation settings.</p><p>For more output examples and the Olares One hardware option, explore <a href="https://www.olares.com/scenes/minimax-h3"><strong>MiniMax H3 on Olares</strong></a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e90a9671aa83" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing Lares, the Olares-Native Agent]]></title>
            <link>https://olares.medium.com/introducing-lares-the-olares-native-agent-292b3bd60fa1?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/292b3bd60fa1</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[olares]]></category>
            <category><![CDATA[personal-ai-cloud]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 07:26:29 GMT</pubDate>
            <atom:updated>2026-10-14T11:31:09.808Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v4h57mLmQObZaFkTFVX-IQ.png" /></figure><p>With Olares v1.12.7, we are introducing Lares: an Olares-native AI agent that runs directly on your device. Ask it to check your system, organize files, research a topic, or help deploy an application, all through a conversation.</p><p>Lares brings together your Olares workspace, platform tools, and the models you configure through Router. It can plan a task, use the available tools, and produce a result you can keep working with, from a research brief to a deployed app.</p><p>For the broader release, including Router, new local models, and other platform updates, read <a href="https://www.olares.com/blog/olares-1-12-7/"><strong>Olares 1.12.7: Router, Lares, and a Broader Local AI Experience</strong></a>.</p><h3>What can Lares do today?</h3><p>Start with the work you already do on Olares. The tasks Lares can complete depend on your configured models, tools, skills, and permissions.</p><ul><li><strong>Manage your device and apps.</strong> Check system status, inspect configuration, install applications, and investigate problems.</li><li><strong>Work with files.</strong> Add attachments or reference files and earlier sessions, organize materials, and create drafts or summaries in your workspace.</li><li><strong>Research a topic.</strong> With web research tools configured, find sources, read relevant pages, and bring the findings together with links.</li><li><strong>Create media.</strong> Use supported image, video, audio, or 3D generation capabilities exposed through Router, and view the results in the conversation.</li></ul><p>Lares can manage installed apps through Olares platform capabilities. To perform tasks inside an app, it needs that app to expose supported tools, integrations, or Agent Skills; installing an app alone does not make all its functions available to Lares.</p><h3>From a request to a running app</h3><p>A task becomes more useful when Lares can carry it across several steps. Our onboarding guide includes an example of deploying Wealthfolio from its GitHub repository.</p><p><strong>The request:</strong></p><pre>Deploy this app to Olares: https://github.com/wealthfolio/wealthfolio and make sure it has a desktop icon.</pre><p><strong>The work:</strong> Lares inspects the project, prepares its Olares Application Chart, and updates the required manifest files. The conversation keeps the steps and the resulting deployment details together.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Netjff4sJORm3qb1.png" /><figcaption>Lares reports the deployment details and verification steps. Screenshot from the Olares onboarding guide; select to enlarge.</figcaption></figure><p><strong>The result:</strong> In this example, Wealthfolio is available from the Launchpad and My Olares. See the <a href="https://www.olares.com/docs/one/olares-onboarding#deploy-an-app-to-olares"><strong>app deployment walkthrough</strong></a> for the full example.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*coGKAtgHJRHd6AMj.png" /><figcaption>Wealthfolio appears in My Olares after deployment.</figcaption></figure><p>The same approach applies to research: ask Lares to gather sources, summarize the findings, and draft a short article. If a supported image-generation capability is available through Router, you can also ask for an accompanying image.</p><p><strong>Before researching the web:</strong> Configure the required search and page-reading tools in Router. For a search tool, also select it under <strong>Settings &gt; Web Search</strong> in Lares. The <a href="https://www.olares.com/docs/use-cases/lares#run-a-deep-research-task"><strong>Lares research tutorial</strong></a> walks through the setup.</p><h3>How Lares works with Router</h3><p>Lares is built on DeepSeek Harness and adapted for Olares. It uses two complementary parts of the platform to carry out your requests:</p><ul><li><strong>Router provides access to AI models and tools.</strong> Manage local models, cloud providers, and supported AI capabilities in one place. Lares uses those connections, so you do not need to configure the same model providers again inside Lares.</li><li><strong>Olares CLI and Agent Skills support system actions.</strong> They let Lares work with your files, apps, and system services under your account’s permissions. In Lares, you also choose the permission level for the task: Read Only, Write, or Full Access.</li></ul><p>Lares is not tied to one model or provider. You can expand its available capabilities as your needs change. See the <a href="https://www.olares.com/docs/use-cases/olares-router"><strong>Router guide</strong></a> for how local and external capabilities connect.</p><h3>Alongside the agents you already use</h3><p>You can continue using Codex, OpenCode, Claude Code, or other agents for your preferred workflows. Agents with access to the same Olares CLI Agent Skills can perform overlapping system tasks. Lares provides an Olares-native starting point, with your device context and Router connections close at hand.</p><p>For more on where agents run and how they access your device, read <a href="https://www.olares.com/blog/olares-cli-ai-agents-explained/"><strong>Olares, Olares CLI, and AI Agents: The Relationship Explained</strong></a>.</p><h3>Try Lares</h3><p>Start on Olares v1.12.7 or later, with Lares, Router, and a connected model ready to use. Devices shipped with the Olares One v1.12.7 factory image include Lares, Router, and Qwen3.8–27B (llama.cpp). If you upgraded an existing device or run a self-hosted installation, follow the <a href="https://www.olares.com/docs/one/olares-onboarding#step-1-prepare-your-environment"><strong>environment setup guide</strong></a> to install the required apps from Market.</p><ol><li>Open Lares from the Launchpad.</li><li>Choose your workspace, model, and permission level. Read Only is a useful starting point for inspecting your device.</li><li>Send your first request:</li></ol><pre>Check this device&#39;s configuration.</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kgJTrNrMut2DksOP.png" /><figcaption>Choose your workspace, permissions, and model before starting.</figcaption></figure><blockquote><strong>Run one task at a time.</strong> When using Qwen3.8–27B (llama.cpp) on Olares One, we recommend running one request at a time to preserve the available context and model accuracy.</blockquote><p>Continue with the <a href="https://www.olares.com/docs/one/olares-onboarding#step-3-try-common-tasks"><strong>common-task examples</strong></a>, or configure web research with the <a href="https://www.olares.com/docs/use-cases/lares"><strong>Lares tutorial</strong></a>.</p><h3>Where Lares is going</h3><p>Today, Lares helps you operate and create within Olares using the capabilities available to it. Over time, we plan to extend that role: understanding broader goals, identifying the right specialized tools or agents, and coordinating their work across your personal AI cloud.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=292b3bd60fa1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing Ticket: Get Support Right from Your Olares Device]]></title>
            <link>https://olares.medium.com/introducing-ticket-get-support-right-from-your-olares-device-058ab619000f?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/058ab619000f</guid>
            <category><![CDATA[technical-support]]></category>
            <category><![CDATA[olares]]></category>
            <category><![CDATA[self-hosting]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Tue, 29 Sep 2026 06:00:59 GMT</pubDate>
            <atom:updated>2026-10-14T11:43:01.404Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OXWeR3XXojuAV3kPl1_e5Q.jpeg" /></figure><p>We take helping you seriously. And most of the time, the fastest help comes from the community.</p><p>The <a href="https://www.olares.com/forum/"><strong>Olares Forum</strong></a>, <a href="https://github.com/beclab/Olares"><strong>GitHub</strong></a>, and our <a href="https://discord.gg/olares"><strong>Discord</strong></a> community are where most questions should start. Someone has probably run into the same thing before you, and when the answer is public, it helps the next person too. How-to questions, setup advice, show and tell: post them in the open, and the community, including our team, will jump in.</p><p>But sometimes your question involves things you shouldn’t post publicly: system logs, device details, account information, or the shipping status of your Olares One order. Or the issue is specific to your machine and needs a one-on-one diagnosis from our team.</p><p>For those, today we’re introducing <strong>Ticket</strong>, an official, private channel for reaching the Olares team. Every submission is trackable, with a clear status and a complete conversation history from start to resolution.</p><h3>How to create a ticket from Olares</h3><h4>Create a ticket from the UI</h4><p>Install Ticket directly from Olares Market. Open the app, then scan the QR code with LarePass to sign in.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*289ceUUKEbMWpPBK" /></figure><p>Be sure to use the Olares ID that you used to activate your device so that your ticket is associated with the correct machine.</p><p>Describe what’s happening in your own words. For technical issues, you can attach system logs right in the app to save a round trip.</p><p>And if your device is temporarily unavailable, open Olares Space in any browser and select <strong>Tickets</strong> in the sidebar. From there, you can create a new ticket or continue an existing conversation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IKWFHVG5E1RZ6QLM" /></figure><h4>Create a ticket from the CLI</h4><p>We’ve also upgraded the Olares CLI so you can use it to create a ticket with the system logs automatically collected.</p><p>Either in the Ticket app or in Olares Space, click the terminal icon. The <strong>Upload logs with Olares CLI</strong> window shows a ready-to-use command with a verification code.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IUX4ErJA-U2Wwpgs" /></figure><p>The code is time-limited and refreshes when the countdown ends, so copy the command while it’s still valid. Then run it on your Olares device, either from the local terminal or over SSH.</p><h3>Which channel should you use?</h3><p>A quick rule of thumb:</p><ul><li><strong>Start on the</strong> <a href="https://www.olares.com/forum/"><strong>Olares Forum</strong></a> for how-to questions, setup advice, and experience sharing. The answer you get stays searchable and helps everyone.</li><li><strong>Open a</strong> <a href="https://github.com/beclab/Olares/issues/new"><strong>GitHub Issue</strong></a> for bugs you’d like to report and track publicly.</li><li><strong>Submit a ticket</strong> when the issue is specific to your device, involves account or order details, or is simply something you’d rather keep private.</li></ul><h3>Get started</h3><p>Ticket is available on Olares OS starting from v1.12.7. If you haven’t upgraded yet, see <a href="https://www.olares.com/docs/one/update"><strong>Update Olares</strong></a> for instructions.</p><p>Whenever you need help, Ticket gives you a direct, private line to the Olares team. For a comprehensive guide, see the <a href="https://www.olares.com/docs/manual/help/request-technical-support"><strong>documentation</strong></a>.</p><p>We hope reaching us now feels as smooth as the rest of Olares. The next time something needs our eyes on your device, send us a ticket. For everything else, the community is right there.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=058ab619000f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Olares 1.12.7: Router, Lares, and a Broader Local AI Experience]]></title>
            <link>https://olares.medium.com/olares-1-12-7-router-lares-and-a-broader-local-ai-experience-2816abac70f6?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/2816abac70f6</guid>
            <category><![CDATA[new-releases]]></category>
            <category><![CDATA[olares]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Mon, 28 Sep 2026 14:43:42 GMT</pubDate>
            <atom:updated>2026-09-28T14:43:42.376Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*lPuM821FKJJ9gfxz.jpg" /></figure><p>Olares 1.12.7 centers on two additions: Router and Lares. Router provides a shared foundation for models and tools across Olares, and Lares uses that foundation to offer a first-party agent experience for your personal cloud. Together, they make the platform’s growing AI ecosystem easier to configure, extend, and use.</p><p>That shared foundation arrives with a broader model catalog spanning chat, speech, translation, OCR, retrieval, and music generation. The release also includes improvements across search, app lifecycle reliability, hardware support, and multi-node management, alongside recent updates to LarePass and Market.</p><p>Read on for the highlights, or see the <a href="https://github.com/beclab/Olares/releases/tag/1.12.7">complete release notes</a> for the full change list.</p><p>Before you begin, see <a href="#how-to-update-to-1-12-7">How to update to 1.12.7</a>.</p><h3>At a glance</h3><ul><li>Router gives AI applications one managed interface for local models, compatible external providers, and tools such as web search, web fetch, translation, and OCR.</li><li>Lares is the first-party Olares agent for planning and carrying out multi-step work across apps, files, and system services.</li><li>The model catalog expands beyond chat across speech, translation, OCR, embeddings and reranking, and music creation.</li><li>Recent LarePass updates make LAN access easier, translate webpages through Router, and save downloads to Olares with Fetch.</li><li>Support for five additional languages begins rolling out across the Olares experience: French, German, Italian, Spanish, and Japanese.</li><li>Ticket provides a private path to official support from an Olares device or Olares Space.</li></ul><h3>A consistent foundation for AI across Olares</h3><p>Local models, external providers, and AI tools all have different APIs, names, credentials, and operating characteristics. Router gives Olares applications a common way to reach these capabilities while keeping provider configuration, access control, and usage records in one place.</p><p>Lares uses that shared layer to choose models and tools, then combines it with Olares CLI and Agent Skills to work across the rest of the system.</p><p>The two products have distinct roles: Router coordinates access to AI capabilities, while Lares turns a goal into a sequence of actions inside your personal cloud.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Mm_D0YNqt6G-Vg2x.png" /><figcaption>Router provides a consistent model-and-tool interface</figcaption></figure><h3>One interface for models and tools with Router</h3><p>Router organizes available capabilities into four broad groups:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/860/1*Jy3EPgYqUepai_XssAjlUg.png" /></figure><p>For an AI app, Router provides a single connection point for configured models and tools. Provider connections, access policies, and default models are managed centrally in Router, so applications that use a Router default automatically follow the active selection when it changes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*krH8nP0jPrPgvkcH.png" /><figcaption>Connect an application running inside Olares without an API key.</figcaption></figure><p>Applications inside Olares can use platform identity without an API key. External clients and third-party integrations can use Router-issued API keys with defined model scopes, expiration dates, and quotas. Administrators can revoke these keys when needed without exposing upstream provider credentials.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*q6_jWE1DvzNFw15K.png" /><figcaption>Create an API key in Router</figcaption></figure><p>See <a href="https://www.olares.com/docs/use-cases/olares-router">Using Olares Router</a> for architecture, authentication, and connection examples.</p><h3>Meet Lares, the first-party agent for your personal cloud</h3><p>Lares gives your Olares personal cloud a conversational interface. Ask it to organize files, research a topic, deploy apps, inspect your system, or manage your Olares environment in plain language.</p><p>Lares coordinates the AI models and tools configured in Router, then uses Olares CLI and Agent Skills to carry out system actions. Because it runs as an approved Olares application, it can use the credentials provided by the platform. You do not need to create a separate CLI profile or configure the same providers again inside Lares.</p><p>Lares is designed for tasks that involve more than one step. A research request can search the web, open promising sources, extract relevant material, and return a structured answer with links. An app-porting request can start with a repository or Docker Compose file, turn it into an Olares Application Chart, deploy it, inspect the result, and iterate on issues it finds. The value is the ability to keep the steps, tools, decisions, and results together in one task.</p><blockquote><strong><em>Note:</em></strong><em> When using Lares with Qwen3.8–27B on Olares One, we recommend running one request at a time to preserve the 100K context window and model accuracy.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*TK8BGpIUTEQEWysA.png" /><figcaption>Lares coordinates Router, Olares Skills, and system tools to carry a multi-step request through to a result.</figcaption></figure><p>Start with the <a href="https://www.olares.com/docs/use-cases/lares">Lares tutorial</a>, or use the <a href="https://www.olares.com/docs/one/olares-onboarding">Olares onboarding guide</a> for a broader introduction to working with agents on Olares.</p><h3>More models for more kinds of work</h3><p>Router arrives alongside a much broader set of installable model applications. With Olares 1.12.7, the model lineup expands across conversation, speech, retrieval, document understanding, translation, and music generation:</p><ul><li><strong>Conversation and reasoning:</strong> Gemma 4 is available for vLLM and Ollama, alongside Qwen3.5, several Qwen3.6 variants, Qwen3.8, Qwopus3.6, Muse Glimmer, and Ornith. The range of sizes and runtimes gives you more options for different hardware profiles.</li><li><strong>Speech and audio:</strong> Breeze TTS 2 and FireRedTTS3 Instruct support speech generation, while Faster-Whisper Large V3, Whisper Large V3 (FW), and Qwen3-ASR 1.7B cover transcription. Qwen3 Forced Aligner, Pyannote Embedding, Speakrs Diarization, Silero VAD V5, and SpeechBrain Enhance handle alignment, speaker analysis, voice activity detection, and audio enhancement.</li><li><strong>Translation, document understanding, and retrieval:</strong> Hy-MT2 supports local translation. dots.ocr, PaddleOCR Hybrid, and Surya OCR 2 extract content from documents and images. EmbeddingGemma, Jina CLIP V2, and BGE-Reranker-v2-m3 support search and retrieval workflows.</li><li><strong>Creative generation:</strong> ACE-Step Music supports local music generation.</li></ul><p>The point is choice by capability, not one model for every job. A browser translation feature can use a compact translation model, while a research workflow can combine search, reranking, OCR, and a reasoning model.</p><p>Over the coming weeks, more Olares applications will expose these capabilities in focused workflows. We will introduce them as the experiences are ready, along with practical guides for selecting models that fit your hardware.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*J1_TnN-EextndWeS.png" /><figcaption>The Olares model catalog now spans language, speech, retrieval, translation, OCR, and music workloads.</figcaption></figure><h3>Recent LarePass updates: local access, translation, and Fetch</h3><p>LarePass ships on its own application cadence, so these features are not all tied to the system release itself. Together, they make it easier to reach Olares on a local network and move information into it.</p><h3>Use local domains without turning on VPN</h3><p>On Windows and macOS, LarePass Desktop can manage host mappings for Olares applications. When your computer and Olares are on the same LAN, those mappings let supported Olares domains resolve locally without first enabling LarePass VPN. This is useful for everyday access at home or in the office, while VPN remains available when you are away.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7Ikni2tvKpjgrTBH.png" /><figcaption>The Olares model catalog now spans language, speech, retrieval, translation, OCR, and music workloads.</figcaption></figure><p>See <a href="https://www.olares.com/docs/manual/get-started/local-access">local access methods</a> for the setup and decision guide.</p><h3>Translate webpages with a local model</h3><p>The LarePass Chrome extension can translate a webpage through Router using the locally hosted Hy-MT2–1.8B model. It supports 33 languages and keeps the translation path under your control instead of sending the page to a separate translation service.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YzfNKNYgbN6hpNu4.jpg" /><figcaption>Translate webpages through Router with a locally hosted model.</figcaption></figure><p>See <a href="https://www.olares.com/docs/manual/tutorial/translate-webpages-with-larepass">translating webpages with LarePass</a> for installation and usage.</p><h3>Send downloads to Olares with Fetch</h3><p>Fetch in LarePass Desktop accepts HTTP and HTTPS links, magnet links, torrent files, Hugging Face repositories, and supported media URLs. Add a resource from your computer and let Olares continue the transfer in the background, even after you close the desktop client.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CMcbV2_IRlIm4GTd.png" /><figcaption>Send supported resources to Olares and let Fetch continue the transfer in the background.</figcaption></figure><p>See <a href="https://www.olares.com/docs/manual/larepass/save-resources-to-olares">saving resources to Olares</a> for supported inputs and destination settings.</p><h3>Five additional language options</h3><p>Olares now includes French, German, Italian, Spanish, and Japanese as language options alongside English and Simplified Chinese.</p><p>Coverage rolls out application by application. LarePass supports all seven languages, while the browser desktop and new first-party applications such as Router, Lares, and Ticket begin with English and Simplified Chinese in this release. We will continue expanding localized interfaces and content in subsequent application updates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*cJmZIErU36bcP29a.png" /><figcaption>LarePass now supports English, Simplified Chinese, French, German, Italian, Spanish, and Japanese.</figcaption></figure><h3>Get private support with Ticket</h3><p>Ticket gives you a direct but private channel to official Olares support from your device or Olares Space. Submit a device-specific issue, attach system information and logs when needed, and track replies and status in one place.</p><p>Ticket keeps your request, diagnostic details, and follow-up conversation together, making it easier to handle issues that require deeper investigation or private context.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NTxu8yxtRiEA32H_.png" /><figcaption>Contact official Olares support and follow a device-specific request in one place.</figcaption></figure><p>See <a href="https://www.olares.com/docs/manual/help/request-technical-support">Get technical support</a> for instructions on creating and following a support request.</p><h3>Other improvements and fixes</h3><p>Olares 1.12.7 also includes a broad set of platform improvements:</p><ul><li>Global search now covers Drive, Sync, Google Drive, Dropbox, and Knowledge, with better indexing for Office files, PDFs, and scanned documents.</li><li>App installs, updates, stops, and offline recovery are more reliable, while Market search and categories are easier to navigate.</li><li>Overlay Gateway switching is safer, and multi-node setup and management are simpler.</li><li>Files gains clearer folder-usage information and broader access to shared storage, alongside fixes for media playback, Sync, and downloads.</li></ul><h3>New in Market</h3><p>Alongside the model applications covered above, Market also adds:</p><ul><li><a href="https://www.olares.com/docs/use-cases/open-design"><strong>Open Design</strong></a>: Create slide decks, landing pages, wireframes, prototypes, and design documents locally.</li><li><a href="https://www.olares.com/docs/use-cases/minecraft"><strong>Minecraft</strong></a>: Host a Minecraft Java Edition server on Olares, with a browser-based console and LAN or VPN access.</li><li><a href="https://www.olares.com/docs/use-cases/palworld"><strong>Palworld</strong></a>: Host a dedicated Palworld server on Olares, with console access and LAN or VPN connections.</li><li><strong>VRAMarket</strong>: A community-contributed app for renting GPU memory peer to peer through Olares GPU memory slicing. Thanks to <a href="https://github.com/jeffreyturov-dev">@jeffreyturov-dev</a> for bringing VRAMarket to Olares.</li><li><a href="https://www.olares.com/docs/use-cases/blender"><strong>Blender</strong></a>: Create 3D models, animations, and renders in your browser, and connect Lares or other compatible agents through MCP to work on your scene.</li><li><a href="https://www.olares.com/docs/use-cases/concat"><strong>Concat</strong></a>: Edit videos in your browser with multiple tracks, transitions, and titles, with API and MCP access for automated editing.</li><li><strong>Junior Investor</strong>: A community-contributed app for kids and teens to learn investing through lessons, historical market scenarios, and simulated trading with virtual money.</li><li><strong>Locius</strong>: A community-contributed AI agent for tasks across Gmail, Notion, Slack, and the web, with approval prompts before sending, posting, or deleting.</li><li><strong>Unbound</strong>: Run your own DNS resolver with DNSSEC validation and caching, or use it alongside AdGuard Home.</li></ul><p>Want to bring an application to Olares? See the <a href="https://www.olares.com/docs/developer/develop/submit-apps">app submission guide</a>.</p><h3>How to update to 1.12.7</h3><p>Before you start, read the <a href="https://www.olares.com/docs/manual/update-guides/1.12.7">Olares 1.12.7 update guide</a> for the correct update path from your current version and any application-specific notes.</p><ol><li>Update LarePass to the latest version available from your app store.</li><li>Open LarePass and go to <strong>Settings</strong> &gt; <strong>System</strong> &gt; <strong>System update</strong>.</li><li>Verify that Olares 1.12.7 is available, then select <strong>Update</strong>.</li><li>Choose an update method and confirm the installation.</li><li>Wait for Olares to restart, then refresh the desktop.</li><li>Install Router, Lares, and the model applications you want from Market. The update guide recommends Qwen3.8–27B as a starting point where the hardware fits.</li></ol><p>If this is your first time using Olares agents, continue with the <a href="https://www.olares.com/docs/one/olares-onboarding">onboarding guide</a>. It covers the initial agent setup and introduces the Olares Skills available to supported applications.</p><h3>What’s next</h3><p>Every Olares release is shaped by the questions, bug reports, and ideas you share with us. Thank you for continuing to tell us what works and where Olares can improve.</p><p>Router and Lares establish a shared foundation that more Olares applications can build on. In the coming weeks, we will publish additional tutorials and introduce focused experiences for the expanded model catalog, with practical examples for research, documents, speech, translation, and creative work.</p><p>Update to Olares 1.12.7, try Lares with a task from your own workflow, and explore the new applications in Market. Tell us what works and what we should improve next on the <a href="https://www.olares.com/forum">Olares Forum</a>, <a href="https://discord.gg/olares">Discord</a>, or directly in Olares Ticket.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2816abac70f6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Olares, Olares CLI, and AI Agents: The Relationship Explained]]></title>
            <link>https://olares.medium.com/olares-olares-cli-and-ai-agents-the-relationship-explained-bf381447dff0?source=rss-b748eaf1db0b------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf381447dff0</guid>
            <category><![CDATA[olares-cli]]></category>
            <category><![CDATA[olares]]></category>
            <dc:creator><![CDATA[Olares]]></dc:creator>
            <pubDate>Fri, 18 Sep 2026 09:09:16 GMT</pubDate>
            <atom:updated>2026-09-18T09:09:16.837Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Z3FOO5RfQsRv9oMx5rn8QA.jpeg" /></figure><p><a href="https://www.olares.com/os">Olares</a> is a personal cloud operating system. It looks after your apps, your files, your models, your compute, and your network. Asking an agent to operate it usually means more than a single function call, because one task can cross several of those systems and run through a chain of operations that depend on each other.</p><p>In May we published <a href="https://www.npmjs.com/package/@olares/cli">the npm package for Olares CLI</a> along with its Skills, which let agents such as Codex and Claude Code work with a running Olares. <strong>As we keep adding commands and refining the Skills, more and more of the work that used to happen in the Olares interface can be handed to an agent instead.</strong></p><p>Along the way we also learned where people actually get stuck. Where is the agent supposed to run? Do I have to use Olares CLI inside one particular agent, <a href="https://www.youtube.com/watch?v=7gWQgeLxRbU">the way we showed with OpenCode</a>? These are the questions this article is here to answer. <strong>Simply put, any agent that can run a command can manage your Olares. The agent itself can run on Olares, or on your own computer, and either way it reaches the same Olares through the Olares CLI in whatever environment it happens to be in.</strong></p><h3>Where an agent can run</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CkBbVbFgLp6bV_7E.png" /><figcaption>Several agents on one computer can share a single Olares CLI. Agents running on Olares are isolated from each other, each using Olares CLI inside its own container.</figcaption></figure><p>When Codex, Claude Code, and the rest run in the same macOS, Windows, or Linux user environment, they can share whichever Olares CLI is installed on that system.</p><p><strong>So if you already have an agent and a model you like on your own computer, there is no need to set up a second agent on Olares first.</strong> Install Olares CLI locally and it will connect over the network and operate your Olares from there.</p><p><strong>An agent can also run as an app on Olares itself.</strong> Codex and Claude Code installed on Olares, along with Lares, the native agent we are releasing soon, each run in their own container. Because those containers are isolated from one another, every agent app that needs to operate Olares carries its own copy of Olares CLI and its Skills. <strong>Agent apps already adapted for Olares, like the ones in the diagram, bring those pieces with them, so there is nothing for you to install.</strong></p><p>Which means one person’s agents can end up spread across a few computers and a few containers on Olares, each calling the Olares CLI in its own environment. That raises the obvious question. How does any of those CLIs know which Olares to connect to, and which user it is acting for? That is what a profile settles.</p><h3>A profile points every agent at the same Olares</h3><p><strong>A profile holds the Olares to connect to and the Olares ID to connect as.</strong> A single installation of Olares CLI can keep several of them, for different machines or different identities.</p><p>Agents sharing a local user environment share more than the CLI itself. They share the profiles stored in it and the login state that goes with them, so the CLI only has to be installed and logged in once. A CLI on another computer, or inside another container on Olares, keeps its own profile. <strong>As long as those profiles point at the same Olares with the same Olares ID, all of those agents are working on the same system.</strong></p><p>The profile also settles whose authority the agent is using. Log in with an Olares ID that has admin privileges and the agent can do what an admin can do. Log in with a regular Olares ID and it is held to whatever that ID is allowed to do. Either way, Olares checks each request against that user’s permissions as it arrives.</p><p>Lares is the exception to the setup. When you install it, Olares passes the installing user’s CLI credentials into its runtime, so <strong>Lares can act as that user from the moment it starts.</strong> There is no profile to create and no profile login to run. Install a model and it is ready to use.</p><h3>Starting from your own computer</h3><p>If you are already using Codex, Claude Code, or something similar locally, start here. The installer sets up Olares CLI and puts the Skills where agents will find them.</p><pre>npx @olares/cli@latest install</pre><p>Then log in, replacing &lt;your-olares-id&gt; with your own.</p><pre>olares-cli profile login --olares-id &lt;your-olares-id&gt;</pre><p>Enter your Olares password when prompted. If you have two-step verification turned on, you will also need the six-digit code from LarePass. The CLI creates the profile once you are through, and olares-cli profile list will show it as logged-in.</p><p>From there an agent can check system state, manage your files and apps, or help move an existing app onto Olares.</p><h3>Summary</h3><p>Olares CLI means where an agent runs stops being a constraint. You keep the agent and the model you already know, and Olares supplies the part that has to stay up: the apps, the data, the compute, and the permissions around them.</p><p>More tutorials and examples of agents working with Olares CLI are on the way. We would like to see yours as well. If you get an agent doing something useful on your Olares, or hit a place where a command or a Skill does not fit the job, share it on <a href="https://discord.gg/olares">Discord</a> or in the <a href="https://forum.olares.com/">forum</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf381447dff0" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>