<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"><channel><title>freedzhost.cloud · New articles</title><link>https://freedzhost.cloud/</link><description>Free tutorials to run your own AI agent swarm and homelab at home: hardware, local models, and the services that replace your subscriptions.</description><language>en</language><item><title>Inari is now MIT, and it runs on a core of our own</title><link>https://freedzhost.cloud/inari-mit.html</link><guid>https://freedzhost.cloud/inari-mit.html</guid><description>Inari began as a coat of paint on Cloudroom, an excellent app for Claude Code, Codex and Pi. It is now its own core, interface and Android client, MIT-licensed. What we borrowed, why it is great, and why ours answers our needs better.</description></item><item><title>Inari: Claude Code, Codex and Pi, self-hosted under your own name</title><link>https://freedzhost.cloud/inari.html</link><guid>https://freedzhost.cloud/inari.html</guid><description>Cloudroom&#x27;s desktop app is Apache-2.0 and its core needs no account — self-host it, and a small MIT tool repaints the identity so it is yours: logo, colours, product name. The upstream engine is never forked, only patched in place at build time.</description></item><item><title>About: who writes this, on what machines, and how we measure</title><link>https://freedzhost.cloud/about.html</link><guid>https://freedzhost.cloud/about.html</guid><description>The real homelab behind the tutorials, how we measure, and what we do not promise.</description></item><item><title>turnstile: one model per card, and a service that never lies about what&#x27;s running</title><link>https://freedzhost.cloud/turnstile.html</link><guid>https://freedzhost.cloud/turnstile.html</guid><description>Several local models, one or two GPUs: config files drift and end up describing a machine that doesn&#x27;t exist. turnstile stops believing files and observes the machine: one slot at a time per card, a clean queue, five facts verified every ten seconds.</description></item><item><title>Stacking free tiers: the quota warden</title><link>https://freedzhost.cloud/quota-warden.html</link><guid>https://freedzhost.cloud/quota-warden.html</guid><description>Eight providers each give away a small daily quota. Taken one at a time, that is not a workday. Added up, routed and watched automatically, it is: how a bit of automation, thread teleportation and provider handoff keep a swarm moving between frontier quota cycles.</description></item><item><title>Favorites</title><link>https://freedzhost.cloud/coups-de-coeur.html</link><guid>https://freedzhost.cloud/coups-de-coeur.html</guid><description>Curated external links: hardware, tools and projects that share the local-first vision.</description></item><item><title>Hardware</title><link>https://freedzhost.cloud/materiel.html</link><guid>https://freedzhost.cloud/materiel.html</guid><description>All hardware articles: graphics cards, memory, buses, storage and benchmarks.</description></item><item><title>Training</title><link>https://freedzhost.cloud/formation.html</link><guid>https://freedzhost.cloud/formation.html</guid><description>Learn to develop your own code, assisted by AI, on your own machine — not depend on an AI that codes for you.</description></item><item><title>News</title><link>https://freedzhost.cloud/actualites.html</link><guid>https://freedzhost.cloud/actualites.html</guid><description>The daily briefing on what matters for local AI, hardware, and homelab.</description></item><item><title>An agent&#x27;s state as a portable file: moving a running agent between machines</title><link>https://freedzhost.cloud/portable-agent-state.html</link><guid>https://freedzhost.cloud/portable-agent-state.html</guid><description>Keep an agent&#x27;s whole state in a folder of plain files, switch its model mid-conversation, and move it from one machine to another without ever running it twice. The design, the one ambiguous moment in the protocol, and how to test it with injected failures.</description></item><item><title>A service burns a core and gets killed: find out why in ten minutes</title><link>https://freedzhost.cloud/profile-a-hot-service.html</link><guid>https://freedzhost.cloud/profile-a-hot-service.html</guid><description>A self-hosted app pinned a CPU core and was ended by the out-of-memory killer several times a day. A ten-second profile of the live Node process named the exact loop, and the same session exposed a second, unrelated memory thief: hypervisor ballooning.</description></item><item><title>fh-probes: measure a node without becoming the cause of what you measure</title><link>https://freedzhost.cloud/fh-probes.html</link><guid>https://freedzhost.cloud/fh-probes.html</guid><description>GPU, RAM, memory pressure, model servers: one JSON line per second, read straight from NVML and /proc, without a single subprocess. And fh-history, which lays your runs over the same timeline: &quot;this run failed&quot; becomes &quot;this run failed while memory pressure climbed to 91&quot;.</description></item><item><title>Tailscale as the backbone: everything private, reachable from anywhere</title><link>https://freedzhost.cloud/tailscale-backbone.html</link><guid>https://freedzhost.cloud/tailscale-backbone.html</guid><description>The model gateway, the phone console&#x27;s API, the fingerprint relay, game streaming, the health reports of a laptop on the road: all of it travels over a private network, none of it is on the internet. How we publish a service to the private network only, with real HTTPS, and the defaults worth tightening.</description></item><item><title>ntfy everywhere: the voice of the whole lab</title><link>https://freedzhost.cloud/ntfy-everywhere.html</link><guid>https://freedzhost.cloud/ntfy-everywhere.html</guid><description>One small self-hosted notification server, and every part of the lab can reach your phone: failed services, an agent waiting for your approval, a new message on the contact form. How we use topics, priorities and links that open the exact screen you need, without Google.</description></item><item><title>Your own CI on Forgejo: the runner traps, one image for everything, and tests that cannot lie</title><link>https://freedzhost.cloud/forgejo-ci.html</link><guid>https://freedzhost.cloud/forgejo-ci.html</guid><description>A self-hosted forge with its own CI runner is freedom, until a job waits forever without a word. The traps of Forgejo Actions, a single runner image that removes the tools a job reinstalled at every run, a language ratchet, and the rule that saved us the most time: read the forge&#x27;s log before reproducing anything.</description></item><item><title>Resisting DDoS when your site lives at home</title><link>https://freedzhost.cloud/ddos-at-home.html</link><guid>https://freedzhost.cloud/ddos-at-home.html</guid><description>No setting on a home server can stop a flood that saturates the line. What you can do: make small attacks useless at the proxy, serve the site from memory, and, for real protection, put a cheap rented front with anti-DDoS in front of the house, so its address is never published. Four layers, from free to five euros a month.</description></item><item><title>Before you open a website to the internet: a security audit in ten checks</title><link>https://freedzhost.cloud/web-exposure-audit.html</link><guid>https://freedzhost.cloud/web-exposure-audit.html</guid><description>We audited a self-hosted site before trusting it, without attacking anything: headers, encryption, DNS, the neighbours that share its door, the tools that are supposed to defend it. The site was clean; what sat next to it was not, and the ban tool had never banned anyone. The method, the commands, and the fix for each finding.</description></item><item><title>Your graphics card at x4 instead of x16: diagnose it, and get the x16 back</title><link>https://freedzhost.cloud/pcie-link-width.html</link><guid>https://freedzhost.cloud/pcie-link-width.html</guid><description>A card can silently run on a quarter of its lanes after a reboot. How to read the link the firmware really gave it, why the BIOS has no &#x27;x16&#x27; setting to pick, the one wrong fix to avoid, and what a power cycle did for us: from x4 back to x16, with nothing changed.</description></item><item><title>The fh- suite: sixty small commands that run the lab</title><link>https://freedzhost.cloud/fh-suite.html</link><guid>https://freedzhost.cloud/fh-suite.html</guid><description>Every everyday gesture here is a short command named fh-something: ask the fleet, hand work to an agent, follow it live, deploy, check health, find why something fails, publish a branch. Grouped by job, one line each, with the rule they all share.</description></item><item><title>fh-tools: fourteen deterministic tools for coding agents</title><link>https://freedzhost.cloud/fh-tools.html</link><guid>https://freedzhost.cloud/fh-tools.html</guid><description>Each one answers one question with a fact, always the same way: does this patch apply, which added lines does no test exercise, does this text carry a credential, have we seen this failure before. JSON in, JSON out, exit codes that mean something, and a checker that replays every case three times.</description></item><item><title>The deterministic doctrine: the model proposes, the code decides</title><link>https://freedzhost.cloud/deterministic-doctrine.html</link><guid>https://freedzhost.cloud/deterministic-doctrine.html</guid><description>The one idea running through every layer of this stack: let models do what has many right answers, and give everything with one right answer to plain code. Why it makes small local models reliable, why a good tool is often worth more than a skill, and the numbers that convinced us, layer by layer.</description></item><item><title>Diagrams in code: every figure of this site is a small function</title><link>https://freedzhost.cloud/diagrams-in-code.html</link><guid>https://freedzhost.cloud/diagrams-in-code.html</guid><description>No drawing tool, no image file, no library: each diagram here is a Python function that writes SVG, with its labels in two languages and a caption that screen readers read. A test checks every one is drawn and described. And when you need a richer chart, a diagram skill with forty-five types.</description></item><item><title>ArtCraft: stage the scene before you generate it</title><link>https://freedzhost.cloud/artcraft.html</link><guid>https://freedzhost.cloud/artcraft.html</guid><description>Prompting is guessing; ArtCraft lets you build the shot: layers in 2D, sets and props in 3D, posed characters, a camera you place, then the generation. An open-source creative studio we build from source on Linux, with the one dependency fix it needed.</description></item><item><title>A sentence becomes a film: the local production chain</title><link>https://freedzhost.cloud/studio-film.html</link><guid>https://freedzhost.cloud/studio-film.html</guid><description>One sentence in, a short film out: a local model writes the technical sheet, code refuses what would waste a render, the cards render the shots, add synchronised foley and music, cut and edit. Measured step by step, with the three decisions that make it work.</description></item><item><title>Homelab: a private ChatGPT, every PDF tool, and transcription at home</title><link>https://freedzhost.cloud/homelab-everyday-trio.html</link><guid>https://freedzhost.cloud/homelab-everyday-trio.html</guid><description>Three self-hosted services: Open WebUI to chat with your own models, Stirling-PDF to merge, split, sign and OCR documents, Scriberr to turn recordings into text. Each replaces a subscription or an upload to a stranger&#x27;s server, and each needs the same three safety settings, learned the hard way.</description></item><item><title>One table for everything that happens: fleet events</title><link>https://freedzhost.cloud/fleet-events.html</link><guid>https://freedzhost.cloud/fleet-events.html</guid><description>Pushes on the forge, failed services, nightly security scans, deliveries, agent runs: everything writes one line to the same database table, with roles that can only insert or only read. Then a dashboard, alerts and agents all read the same facts.</description></item><item><title>Route requests with a small trained classifier, not keywords</title><link>https://freedzhost.cloud/intent-classifier.html</link><guid>https://freedzhost.cloud/intent-classifier.html</guid><description>Which model should take this request: the small, the default, the big one? Keywords got it right 63.7 % of the time, less than always answering &#x27;default&#x27; (72.6 %). A small classifier trained on 237 real requests reached 84.0 %, in 20 milliseconds, without a graphics card. And the bigger lesson it revealed: the tool catalogue.</description></item><item><title>Failures that call you: every service reports, every deploy can roll back</title><link>https://freedzhost.cloud/failures-that-call.html</link><guid>https://freedzhost.cloud/failures-that-call.html</guid><description>A service that fails in silence is found days later. Three small systemd habits make every failure call you, without a monitoring suite: a failure template, a daily health ping that tolerates a sleeping laptop, and a delivery script that puts the old version back by itself.</description></item><item><title>From an e-mail to a pull request: the go label</title><link>https://freedzhost.cloud/mail-to-pull-request.html</link><guid>https://freedzhost.cloud/mail-to-pull-request.html</guid><description>Send yourself an e-mail with an idea; it becomes a drafted issue waiting for you. Put one label on it, and an agent picks it up, works, opens a pull request, and closes the issue when it is merged. One rule keeps it safe: no code ever adds the label.</description></item><item><title>Take your Claude.ai history home: claude-archive, step by step</title><link>https://freedzhost.cloud/claude-archive.html</link><guid>https://freedzhost.cloud/claude-archive.html</guid><description>Your conversations with Claude.ai, as files on your disk: one Markdown file per conversation, every code block as a real file, a checklist of attachments to fetch by hand, and a manifest that proves nothing was lost. One script, no dependency, no network.</description></item><item><title>Omarchy in depth: extend it without breaking updates, and five &#x27;bugs&#x27; that are not</title><link>https://freedzhost.cloud/omarchy-deep.html</link><guid>https://freedzhost.cloud/omarchy-deep.html</guid><description>Omarchy turns Arch and Hyprland into a ready workstation. Extending its menu the supported way, and five strange behaviours we met, each with its real cause: a screen dark at &#x27;100 %&#x27;, a bar that vanishes for good, four kinds of title bars, milky transparency, a menu entry that cannot be pressed.</description></item><item><title>ACE-Step: three minutes of music in 27 seconds, on your own card</title><link>https://freedzhost.cloud/ace-step-music.html</link><guid>https://freedzhost.cloud/ace-step-music.html</guid><description>An open-source music model fast enough to be a tool: whole tracks from a few style tags, repaint of a passage, variations. How we use it for an ambience bank, a soundtrack for automatic videos, and why it never runs while the music plays.</description></item><item><title>billion-context: a context-compression proxy, measured honestly</title><link>https://freedzhost.cloud/billion-context.html</link><guid>https://freedzhost.cloud/billion-context.html</guid><description>A local proxy that sits between your coding agent and its model and promises to stretch the context. We installed it carefully, plugged it into Claude Code, Kimi Code and OpenFox, and measured. Below 100k tokens it saved nothing, and once it cost three times more. When it might help, and how to try it without risk.</description></item><item><title>Strata: a 177-billion-parameter model on one gaming card</title><link>https://freedzhost.cloud/strata-flash-next.html</link><guid>https://freedzhost.cloud/strata-flash-next.html</guid><description>Strata is an open-source engine built to run Qwen3.8-Flash-Next on an ordinary PC: one 12 to 16 GB card and enough RAM for the experts. How we plugged it into our GPU grid as one more slot, and the three details that decide whether it starts.</description></item><item><title>The workshop: every tool behind this site, one line each</title><link>https://freedzhost.cloud/workshop.html</link><guid>https://freedzhost.cloud/workshop.html</guid><description>About forty repositories built and run here, grouped by job, each in one sentence, with the tutorial that explains it. Open-source licences; the method in each one is yours to copy.</description></item><item><title>A local music studio that keeps the project: fh-seq</title><link>https://freedzhost.cloud/fh-seq-music.html</link><guid>https://freedzhost.cloud/fh-seq-music.html</guid><description>Online song generators hand you an opaque MP3. fh-seq keeps the song as a project: sections and lyrics, a tree of renders where every retouch keeps its parent, repaint of a time range, stems you can mix, a clip shot by shot. All on open-source engines, on your own cards.</description></item><item><title>A workstation assistant that points, never types</title><link>https://freedzhost.cloud/workstation-assistant.html</link><guid>https://freedzhost.cloud/workstation-assistant.html</guid><description>Say a sentence, a small local model picks one function from a closed table, and the script runs it. Questions are answered from your own notes first, then by a cascade of models that must say &#x27;insufficient&#x27; when they do not know. What the measurements imposed, one rule at a time.</description></item><item><title>Three small tools for your workstation: your AI history, a context meter, a fleet inventory</title><link>https://freedzhost.cloud/workstation-small-tools.html</link><guid>https://freedzhost.cloud/workstation-small-tools.html</guid><description>Take your Claude.ai conversations home as readable files, see how full your local agent&#x27;s context is right in the status bar, and know what changed on each of your machines since last week. Each one is a single script with no dependency.</description></item><item><title>Knowledge packs: facts with a source, a test and an expiry date</title><link>https://freedzhost.cloud/knowledge-packs.html</link><guid>https://freedzhost.cloud/knowledge-packs.html</guid><description>Models answer from memory, and memory goes stale. A knowledge pack is a small unit of verified knowledge served to agents: each claim has a dated source, each procedure an executable assertion, and each fact an expiry date that opens an issue when it passes.</description></item><item><title>The account that reasons holds no secret: agents and authority</title><link>https://freedzhost.cloud/agent-authority.html</link><guid>https://freedzhost.cloud/agent-authority.html</guid><description>Agents deploy, push and create repositories here, yet none of them holds a token or a password. The rule that makes it safe, and three small services built on it: a deploy daemon with a closed list, a forge helper behind a socket, and a repository bootstrapper that never lets a repository live without its guard.</description></item><item><title>Why is it failing? Give the model facts, and a machine that remembers</title><link>https://freedzhost.cloud/facts-before-diagnosis.html</link><guid>https://freedzhost.cloud/facts-before-diagnosis.html</guid><description>Asked why something fails, a local model explores blindly and invents files. Handed the facts, the same model answers right: 5/5 in 72 seconds instead of 2.75/5 in 9 minutes. Two small tools: one that gathers the facts, one that records what the machine was living through.</description></item><item><title>Skills: teach your agent a procedure once</title><link>https://freedzhost.cloud/agent-skills.html</link><guid>https://freedzhost.cloud/agent-skills.html</guid><description>A skill is a folder with one Markdown file: when to use it, and how. The agent loads it only when the task calls for it, so a hundred skills cost nothing until needed. The ones we use every day, from shaping answers for an ADHD reader to a full security review, and how to write your own.</description></item><item><title>A fully local judge in two passes: Tev1 first, Nimble when in doubt</title><link>https://freedzhost.cloud/judge-two-passes.html</link><guid>https://freedzhost.cloud/judge-two-passes.html</guid><description>A 0.8B model answers in a twentieth of a second; a 9B one is slower but surer. Code sends every question to the small one, keeps its answer when it is confident, and only wakes the large one for the rest. On real e-mails, the pair matches the hosted judge, with nothing leaving the house.</description></item><item><title>Which local model for which job: a bench that measures real work</title><link>https://freedzhost.cloud/model-bench.html</link><guid>https://freedzhost.cloud/model-bench.html</guid><description>Benchmarks favour the fast model; real work favoured the reliable one. A bench where code judges every answer, then each model works for real on a small repository. What it showed: small and well-framed beats big and vague, and four settings matter more than the model.</description></item><item><title>The lab notebook: what we tried on the infrastructure, what we kept</title><link>https://freedzhost.cloud/lab-notebook.html</link><guid>https://freedzhost.cloud/lab-notebook.html</guid><description>Every experiment on our machines, with its verdict and the number that decided it. Kept, rejected, paused. The fastest way to know what is worth an evening before you spend yours.</description></item><item><title>Long context on a 16 GB card: KVMem, and the build we run</title><link>https://freedzhost.cloud/kvmem-long-context.html</link><guid>https://freedzhost.cloud/kvmem-long-context.html</guid><description>Agents read long conversations. KVMem keeps the whole history in system RAM and only the relevant blocks on the card: 48 out of 48 recall checks, flat speed past 100k tokens, 4.9 GB of VRAM saved. And what our build adds: one binary for two card generations, experts in host memory, token probabilities, a GPU expert cache that runs a 177B model from a single card.</description></item><item><title>Inframagotchi: a cat whose health is your homelab&#x27;s</title><link>https://freedzhost.cloud/inframagotchi.html</link><guid>https://freedzhost.cloud/inframagotchi.html</guid><description>Monitoring you actually look at. A manga stray cat lives in a window on your desktop: bright-eyed when everything runs, sweating, bandaged, then knocked out as services fail. Real probes underneath, a drawing that only knows one number, all drawn in code.</description></item><item><title>Purrd: music while the agent works, silence when it is done</title><link>https://freedzhost.cloud/purrd.html</link><guid>https://freedzhost.cloud/purrd.html</guid><description>Stop staring at a terminal. A small open-source program plays generated lo-fi music while a long task runs; the return of silence tells you it finished. It wraps any command, follows Claude Code through hooks, and can speak a line when the agent needs you.</description></item><item><title>Your non-Steam PC games on Linux: Lutris and Ubisoft Connect</title><link>https://freedzhost.cloud/lutris-ubisoft.html</link><guid>https://freedzhost.cloud/lutris-ubisoft.html</guid><description>Steam is not the only shop. Lutris installs the Ubisoft, EA, GOG and Epic launchers inside Proton, on the disk of your choice, and they run like on Windows. Ours holds Assassin&#x27;s Creed Valhalla, Immortals Fenyx Rising and Star Wars Outlaws, ready to play from the sofa.</description></item><item><title>Let an agent see and drive your desktop, under your eyes</title><link>https://freedzhost.cloud/agent-screen-control.html</link><guid>https://freedzhost.cloud/agent-screen-control.html</guid><description>Some tasks have no command line: a settings page, a checkbox, a form. An agent that takes screenshots and moves the mouse can do them while you watch. Two open-source tool servers, four constants that make it work, and a referee that says whether a click actually did something.</description></item><item><title>SSH for an agent workstation: instant, safe, and four traps</title><link>https://freedzhost.cloud/ssh-for-agents.html</link><guid>https://freedzhost.cloud/ssh-for-agents.html</guid><description>When agents run hundreds of SSH commands a day, the details matter. Shared connections that drop each call from 300 ms to 12, free text that must never travel in a command line, the right way to open a window on another machine, and the kill that takes your own shell with it.</description></item><item><title>One fingerprint for the whole lab: sudo on the server, your vault on the laptop</title><link>https://freedzhost.cloud/fingerprint-everywhere.html</link><guid>https://freedzhost.cloud/fingerprint-everywhere.html</guid><description>The laptop&#x27;s fingerprint reader approves a sudo on another machine, and unlocks your password manager. A signed answer, a fresh random number per request, the private key never leaves the laptop, and the password always remains as a fallback.</description></item><item><title>Your GPU container runs on the CPU: three traps after an update or a reboot</title><link>https://freedzhost.cloud/gpu-container-on-cpu.html</link><guid>https://freedzhost.cloud/gpu-container-on-cpu.html</guid><description>The model says healthy in seven seconds, then crawls: it is running on the processor. The card is visible, CUDA is not. A stale device description, a device number that changes at every boot, and the bisection that found them.</description></item><item><title>The context floor: where an agent&#x27;s tokens really go</title><link>https://freedzhost.cloud/agent-context-floor.html</link><guid>https://freedzhost.cloud/agent-context-floor.html</guid><description>We measured an agent run token by token. The conversation was 1 % of the bill; the rest was the same 30,939 tokens of instructions and tool descriptions, sent again at every turn. How to measure your own floor, and which levers actually move it.</description></item><item><title>Slow for no reason? It is re-reading: the KV cache trap</title><link>https://freedzhost.cloud/kv-cache-trap.html</link><guid>https://freedzhost.cloud/kv-cache-trap.html</guid><description>A local model that generates fast but takes ten minutes to answer is not slow: it is reading the prompt badly. One flag made it 21 times slower, silently. How to measure it in two minutes, and the two settings to use instead.</description></item><item><title>ComfyUI: images, video and 3D on your own cards</title><link>https://freedzhost.cloud/comfyui-studio.html</link><guid>https://freedzhost.cloud/comfyui-studio.html</guid><description>The node editor for generative models, run as a service that only listens on the machine itself, one instance per card. Which model for what, how long a five-second video shot really takes, and how to drive it from scripts and agents.</description></item><item><title>Vibe coding and digital art: when describing becomes making</title><link>https://freedzhost.cloud/vibe-coding-art.html</link><guid>https://freedzhost.cloud/vibe-coding-art.html</guid><description>Say what you want, let a model write the code, judge by what appears on screen. Vibe coding was born as a joke about programming; in digital art it is a real shift. Generative visuals, shaders, live-coded music, motion design as code, all reachable by description, with a local model. What it opens, what it costs, and where we draw the line.</description></item><item><title>Create with local AI: every tool of the studio</title><link>https://freedzhost.cloud/create-tour.html</link><guid>https://freedzhost.cloud/create-tour.html</guid><description>Images, video shots, whole short films, music, songs, 3D objects, diagrams, animated explainers: everything we create here runs on our own cards or in plain code. The map of the studio, what each tool is good at, and where to start.</description></item><item><title>Reverse engineering with evidence: let the agent prove, not guess</title><link>https://freedzhost.cloud/agent-rea.html</link><guid>https://freedzhost.cloud/agent-rea.html</guid><description>Before wiring your stack into someone else&#x27;s program, understand it with proof. An open-source reverse-engineering toolkit that agents can call: analyze once, keep a snapshot, ask follow-up questions, and get answers with their evidence and their limits.</description></item><item><title>BMAD with small models: the model proposes, code decides</title><link>https://freedzhost.cloud/agent-bmad.html</link><guid>https://freedzhost.cloud/agent-bmad.html</guid><description>A planning method for agent work, adapted to local models: stories with a schema, scenarios written as data and run by pytest, one story alive at a time, and no model ever delivering a verdict.</description></item><item><title>Rooms and FoxDesk: one conversation, any machine, any agent</title><link>https://freedzhost.cloud/agent-rooms-foxdesk.html</link><guid>https://freedzhost.cloud/agent-rooms-foxdesk.html</guid><description>A rooms controller keeps every agent conversation durable and replayable, across several machines. An open-source agent core, never patched, with your agents plugged in by adapters. And a desktop client served on every workstation, so one machine down never cuts you off.</description></item><item><title>Inari: your agents in your pocket</title><link>https://freedzhost.cloud/agent-swarmfox.html</link><guid>https://freedzhost.cloud/agent-swarmfox.html</guid><description>The laptop sleeps, the phone does not. A native Android console that follows coding sessions live, talks to a local orchestrator, and holds the one decision that stays human: the go on a task. Built without Google, published without a store.</description></item><item><title>Kimi K3.1 Pro is out: what changes, what it costs, how to try it</title><link>https://freedzhost.cloud/kimi-k3-1-pro.html</link><guid>https://freedzhost.cloud/kimi-k3-1-pro.html</guid><description>Moonshot AI releases Kimi K3.1 Pro, the first model of its K3.1 series: K3 post-trained with far more compute, a one-million-token context, open weights announced before 4 November. What is announced, who measured what, what it costs on the API and in the Kimi Code subscription, and a referral link to start with a bonus.</description></item><item><title>Kimi Code: the long tasks, on a server, from anywhere</title><link>https://freedzhost.cloud/agent-kimi-code.html</link><guid>https://freedzhost.cloud/agent-kimi-code.html</guid><description>A second hosted agent with a one-million-token context, kept running on the server in a terminal multiplexer and reached from laptop or phone. Plus the two traps that cost us an evening: the tool-list size limit and a resume tied to the folder.</description></item><item><title>Claude Code, two ways: on the subscription and on your own models</title><link>https://freedzhost.cloud/agent-claude-code.html</link><guid>https://freedzhost.cloud/agent-claude-code.html</guid><description>The same agent in two roles: the operations agent on the subscription, with rules and hooks that refuse dangerous commands, and an isolated junior on local models through three environment variables.</description></item><item><title>OpenFox: the coding agent that takes local models seriously</title><link>https://freedzhost.cloud/agent-openfox.html</link><guid>https://freedzhost.cloud/agent-openfox.html</guid><description>Most coding agents are built for a frontier model and tolerate a local one. OpenFox is built the other way round: a plan becomes an immutable contract of acceptance criteria, the agent loops until every criterion passes, and you see the prefill and generation speed of your own card live. Open source, free per token, and our agent of every day.</description></item><item><title>FH Coder: a local coding agent that works while you sleep, and never merges unread</title><link>https://freedzhost.cloud/fh-coder.html</link><guid>https://freedzhost.cloud/fh-coder.html</guid><description>Drop an intention, get a tested pull request. FH Coder is our production chain for local models: a shared queue, a runner in a separate account, a verification gate that fixes the style by itself, a report that opens the pull request and feeds the knowledge base. Plus the fh-coder agent that brings the same doctrine into OpenFox. How it is wired, and how to set it up.</description></item><item><title>What an agent harness is, and how to choose yours</title><link>https://freedzhost.cloud/harness-choose.html</link><guid>https://freedzhost.cloud/harness-choose.html</guid><description>The model writes text; the harness turns it into work: it reads files, runs commands, asks permission, manages the context, and decides when to stop. Six harnesses run here, each for a reason. What sets them apart, and five questions to pick the right one for each job.</description></item><item><title>One client, your whole lab: the tour</title><link>https://freedzhost.cloud/agent-stack-tour.html</link><guid>https://freedzhost.cloud/agent-stack-tour.html</guid><description>From a single app on your phone or desktop: coding agents on every machine, images and video on your cards, the guard machine&#x27;s alerts, your workflows and your house. The concept sounds big. Each brick is small, readable, and already running here.</description></item><item><title>Your own Android app, outside the Play Store, with its own updates</title><link>https://freedzhost.cloud/android-self-publish.html</link><guid>https://freedzhost.cloud/android-self-publish.html</guid><description>Build in a container, sign properly, serve updates from your own server. And the rule that silently breaks updates if you forget it: a version number is used once.</description></item><item><title>Secrets that never touch a file: passwords for scripts and agents</title><link>https://freedzhost.cloud/secrets-pipe.html</link><guid>https://freedzhost.cloud/secrets-pipe.html</guid><description>A vault for people&#x27;s passwords, a secrets manager for machines, and one rule: a secret goes from the vault to its program through a pipe, never through a file, a command line or a screen.</description></item><item><title>Rebuild your machines in one command, and know what you changed by hand (Ansible)</title><link>https://freedzhost.cloud/ansible-workstation.html</link><guid>https://freedzhost.cloud/ansible-workstation.html</guid><description>Describe each machine in a repository; --check --diff tells you what drifted, without touching anything. A first playbook, and the beginner traps.</description></item><item><title>A second-hand laptop as a comfortable Linux workstation (Omarchy)</title><link>https://freedzhost.cloud/linux-laptop.html</link><guid>https://freedzhost.cloud/linux-laptop.html</guid><description>Check what you buy, install a ready-to-work Linux desktop, and the three surprises that look like bugs but are not.</description></item><item><title>A breach warning: changing dozens of passwords without locking yourself out</title><link>https://freedzhost.cloud/password-rotation.html</link><guid>https://freedzhost.cloud/password-rotation.html</guid><description>The order that never leaves you outside, how to choose what to change first, and why an outdated vault entry is worse than an empty one.</description></item><item><title>Undo any change: snapshots of your home folder with snapper</title><link>https://freedzhost.cloud/snapshots.html</link><guid>https://freedzhost.cloud/snapshots.html</guid><description>Instead of .bak copies everywhere, a photo of your folder before every risky change, taken in a second, and a one-line undo.</description></item><item><title>n8n: route, repair and generate workflows, with guardrails</title><link>https://freedzhost.cloud/n8n-route-and-heal.html</link><guid>https://freedzhost.cloud/n8n-route-and-heal.html</guid><description>A request in plain words that loads the right model; a workflow that reloads what crashed, within strict limits; and workflows written by code, versioned like the rest.</description></item><item><title>n8n: an agentic clock that notices what stopped</title><link>https://freedzhost.cloud/n8n-agentic-clock.html</link><guid>https://freedzhost.cloud/n8n-agentic-clock.html</guid><description>Every job declares its cadence; a tick compares each one&#x27;s last sign of life every minute; a watchdog outside n8n watches the tick itself. How an agent network learns that something went silent.</description></item><item><title>n8n against Zapier, Make, IFTTT and Power Automate: why self-host your automations</title><link>https://freedzhost.cloud/n8n-vs-saas.html</link><guid>https://freedzhost.cloud/n8n-vs-saas.html</guid><description>The hosted services bill every step and carry your data through their cloud. Self-hosted n8n runs at home, calls your local models, and costs nothing per run. And when the SaaS is still the better choice.</description></item><item><title>Homelab 18: ban the scanners with fail2ban, and the Docker trap that makes bans useless</title><link>https://freedzhost.cloud/homelab-fail2ban.html</link><guid>https://freedzhost.cloud/homelab-fail2ban.html</guid><description>Read the proxy&#x27;s logs, ban an address after a few errors, and make sure the ban sits where the traffic really passes.</description></item><item><title>Homelab 17: your errors find you, with GlitchTip instead of Sentry</title><link>https://freedzhost.cloud/homelab-glitchtip.html</link><guid>https://freedzhost.cloud/homelab-glitchtip.html</guid><description>A self-hosted error tracker in three containers, compatible with Sentry&#x27;s SDKs: every crash of every service lands in one place, grouped and dated.</description></item><item><title>Homelab 16: keep everything you read, forever, with Karakeep</title><link>https://freedzhost.cloud/homelab-karakeep.html</link><guid>https://freedzhost.cloud/homelab-karakeep.html</guid><description>Replace Pocket and browser bookmarks: full-page archives that survive the original site, tags and summaries by a local model, and your agents can search them.</description></item><item><title>Homelab 15: Home Assistant with a local voice, in French</title><link>https://freedzhost.cloud/homelab-voice.html</link><guid>https://freedzhost.cloud/homelab-voice.html</guid><description>Speak to your house without any cloud: speech recognition and synthesis on your own GPU, a CPU fallback when the card is busy, and a local model to understand.</description></item><item><title>Hybrid: local first, the cloud only when it is worth it</title><link>https://freedzhost.cloud/hybrid-escalation.html</link><guid>https://freedzhost.cloud/hybrid-escalation.html</guid><description>A queue that sends each task to the cheapest agent able to do it: local models, then a cheaper hosted agent, then a frontier one. Daily caps, sandboxes, and your confirmation before anything climbs.</description></item><item><title>A 177-billion-parameter model on two consumer cards: placement beats bandwidth</title><link>https://freedzhost.cloud/big-moe-two-cards.html</link><guid>https://freedzhost.cloud/big-moe-two-cards.html</guid><description>A mixture-of-experts model that does not fit in VRAM can still run at home. Where its weights live matters more than the PCIe link, and two settings decide between a fast server and a frozen machine.</description></item><item><title>Homelab 14: automate without a Zapier subscription, with n8n and a decision cascade</title><link>https://freedzhost.cloud/homelab-n8n.html</link><guid>https://freedzhost.cloud/homelab-n8n.html</guid><description>Workflows that connect your services, and the pattern that keeps them cheap and private: rules first, a small local judge, a hosted model only for public data, you for the rest.</description></item><item><title>One browser per agent: web automation that does not trip over itself</title><link>https://freedzhost.cloud/browser-per-agent.html</link><guid>https://freedzhost.cloud/browser-per-agent.html</guid><description>Twenty agents in one browser looked efficient and produced false results. Separate browsers, clean profiles and encrypted DNS fixed it; here is the setup and the limit of a laptop.</description></item><item><title>Homelab 13: a task list for you and your agents with Vikunja</title><link>https://freedzhost.cloud/homelab-vikunja.html</link><guid>https://freedzhost.cloud/homelab-vikunja.html</guid><description>Replace Todoist or Microsoft To Do, import your lists, and give your agents a shared queue where the decisions only you can take wait for you.</description></item><item><title>Homelab 12: every paper you own, scanned, read and searchable with Paperless-ngx</title><link>https://freedzhost.cloud/homelab-paperless.html</link><guid>https://freedzhost.cloud/homelab-paperless.html</guid><description>Drop a scan or a photo in a folder; it is read, tagged and filed. Search a word on any invoice of the last ten years.</description></item><item><title>Does your test catch anything? Prove it by breaking the code (mutation testing)</title><link>https://freedzhost.cloud/mutation-testing.html</link><guid>https://freedzhost.cloud/mutation-testing.html</guid><description>A green test only says it found nothing, which is also what a broken test does. Inject defects one by one and check that each one turns a test red.</description></item><item><title>Homelab 11: your files and office suite instead of Google Drive and Docs</title><link>https://freedzhost.cloud/homelab-office.html</link><guid>https://freedzhost.cloud/homelab-office.html</guid><description>Nextcloud for files, calendars and contacts, OnlyOffice to edit documents in the browser. The three-party mechanism that explains every setup error.</description></item><item><title>A RAG on your own journal: never solve the same problem twice</title><link>https://freedzhost.cloud/journal-rag.html</link><guid>https://freedzhost.cloud/journal-rag.html</guid><description>Write down what you solved, let a search service index it, and let your agents ask it before they guess. The chunk size that matters, and the three ways it silently serves stale answers.</description></item><item><title>Homelab 10: notifications from your servers and agents, on your phone</title><link>https://freedzhost.cloud/homelab-ntfy.html</link><guid>https://freedzhost.cloud/homelab-ntfy.html</guid><description>ntfy turns one curl line into a phone notification: backups, agents waiting for you, a failed job. Self-hosted, with accounts.</description></item><item><title>A tiny local judge for the small decisions</title><link>https://freedzhost.cloud/local-judge.html</link><guid>https://freedzhost.cloud/local-judge.html</guid><description>Yes or no, which of these, how much: a sub-1B model on the processor answers typed questions in 0.05 s with an honest confidence. And why the code, not the prompt, must send it the question.</description></item><item><title>Homelab 9: your own search engine, for you and your agents</title><link>https://freedzhost.cloud/homelab-searxng.html</link><guid>https://freedzhost.cloud/homelab-searxng.html</guid><description>SearXNG asks dozens of engines for you, without an account or a profile. And it gives your local agents a web search they can read.</description></item><item><title>Speculative decoding (MTP): up to twice as fast on the same card</title><link>https://freedzhost.cloud/speculative-decoding.html</link><guid>https://freedzhost.cloud/speculative-decoding.html</guid><description>A small head guesses the next tokens, the model checks them in one pass. The right flags, the value to measure yourself, the real gain in production, and the crash to watch for.</description></item><item><title>Homelab 8: your own code forge, where nothing reaches main without a green check</title><link>https://freedzhost.cloud/homelab-forgejo.html</link><guid>https://freedzhost.cloud/homelab-forgejo.html</guid><description>Forgejo with its CI, a protected branch, and the rules that let AI agents propose code without ever breaking main.</description></item><item><title>A second graphics card through OCuLink: more VRAM without a new motherboard</title><link>https://freedzhost.cloud/second-gpu-oculink.html</link><guid>https://freedzhost.cloud/second-gpu-oculink.html</guid><description>When the motherboard has one big slot, an M.2 slot can lend its four lanes to an external card. What worked, what did not, and how to read the link.</description></item><item><title>Homelab 7: replace Google Photos with Immich</title><link>https://freedzhost.cloud/homelab-immich.html</link><guid>https://freedzhost.cloud/homelab-immich.html</guid><description>Automatic phone backup, faces, maps and search, at home. And the two things you must back up, not one.</description></item><item><title>RAM, zram and swap on an AI machine: the three thieves that cost us a factor of ten</title><link>https://freedzhost.cloud/ram-zram-swap.html</link><guid>https://freedzhost.cloud/ram-zram-swap.html</guid><description>Before suspecting the GPU, measure the RAM. Three settings were eating it; fixing them took a model from 1.9 to about 20 tokens per second.</description></item><item><title>Several models, two graphics cards: a grid instead of a juggling act</title><link>https://freedzhost.cloud/gpu-grid.html</link><guid>https://freedzhost.cloud/gpu-grid.html</guid><description>One container per model and per card, swaps that cost half a second per gigabyte, a voice service that gives way, and why a model must never fill its card.</description></item><item><title>Homelab 6: reach home from anywhere without opening a port</title><link>https://freedzhost.cloud/homelab-remote.html</link><guid>https://freedzhost.cloud/homelab-remote.html</guid><description>Only the public sites go through the box. Administration, models and private services go through an encrypted private network.</description></item><item><title>Free cloud models behind your gateway, and routes that say who answers</title><link>https://freedzhost.cloud/free-models-routes.html</link><guid>https://freedzhost.cloud/free-models-routes.html</guid><description>When the cards are busy or off, eight official free tiers can take over: Groq, Cerebras, Google AI Studio, Cloudflare, OpenRouter, Cohere, NVIDIA and Mistral. How to wire them into the gateway without one key in a file, how to group them into routes named auto-… that fall back in order, how we proved the fallback, and the traps that cost us an evening.</description></item><item><title>Homelab 5: backups that survive an attack on your server</title><link>https://freedzhost.cloud/homelab-backups.html</link><guid>https://freedzhost.cloud/homelab-backups.html</guid><description>Two copies, two machines, and the detail that matters most: the second copy is pulled with a read-only key, never pushed.</description></item><item><title>One gateway for all your models, and three ways a local model freezes your agents</title><link>https://freedzhost.cloud/model-gateway.html</link><guid>https://freedzhost.cloud/model-gateway.html</guid><description>One address for every client, model names that never lie, and the three settings that froze our agents for days.</description></item><item><title>Giving code work to a local agent without it breaking everything</title><link>https://freedzhost.cloud/agent-intents.html</link><guid>https://freedzhost.cloud/agent-intents.html</guid><description>Local models write plausible code that does not run. The rules we measured: four deliverables at most, the real data, a test that executes, and a pull request someone reads.</description></item><item><title>Homelab 4: block ads and trackers for the whole house with AdGuard Home</title><link>https://freedzhost.cloud/homelab-adguard.html</link><guid>https://freedzhost.cloud/homelab-adguard.html</guid><description>One DNS server for every device, phones and TVs included. And the local rewrite that keeps your own services at home.</description></item><item><title>Sleep when idle, wake on demand: the energy recipe</title><link>https://freedzhost.cloud/sleep-and-wake.html</link><guid>https://freedzhost.cloud/sleep-and-wake.html</guid><description>A tower with big graphics cards should not run for nothing. A small always-on machine wakes it in seconds, from anywhere, without opening anything to the internet.</description></item><item><title>A gaming PC that became an AI server, and still plays</title><link>https://freedzhost.cloud/headless-gaming-node.html</link><guid>https://freedzhost.cloud/headless-gaming-node.html</guid><description>No screen on the tower: a virtual display, game streaming encoded by the integrated GPU, and an Android handheld as the console.</description></item><item><title>Free your graphics cards for AI, and still play</title><link>https://freedzhost.cloud/gpu-for-ai.html</link><guid>https://freedzhost.cloud/gpu-for-ai.html</guid><description>Put the desktop on the integrated GPU, give every byte of VRAM to the models, and let Steam borrow a card only while you play.</description></item><item><title>Homelab 3: your own password manager with Vaultwarden</title><link>https://freedzhost.cloud/homelab-vaultwarden.html</link><guid>https://freedzhost.cloud/homelab-vaultwarden.html</guid><description>Replace the password manager you rent. The server never sees your passwords, and the backup that actually restores.</description></item><item><title>Homelab 2: one HTTPS entrance for all your services</title><link>https://freedzhost.cloud/homelab-proxy.html</link><guid>https://freedzhost.cloud/homelab-proxy.html</guid><description>A reverse proxy with free certificates: how this very site went online, in fifteen minutes.</description></item><item><title>Homelab 1: Proxmox, the base under every service</title><link>https://freedzhost.cloud/homelab-proxmox.html</link><guid>https://freedzhost.cloud/homelab-proxmox.html</guid><description>One machine, many isolated servers, snapshots and backups. The foundation of the series that replaces Google and subscriptions.</description></item><item><title>Your first local model with llama.cpp</title><link>https://freedzhost.cloud/first-model.html</link><guid>https://freedzhost.cloud/first-model.html</guid><description>Download an open model, serve it on your GPU, talk to it with the same API as the big clouds.</description></item><item><title>Docker with access to your GPU</title><link>https://freedzhost.cloud/docker-gpu.html</link><guid>https://freedzhost.cloud/docker-gpu.html</guid><description>Install Docker and the NVIDIA Container Toolkit, test it, and avoid the trap after a driver update.</description></item><item><title>Our findings: every tip and measured fact, one line each</title><link>https://freedzhost.cloud/findings.html</link><guid>https://freedzhost.cloud/findings.html</guid><description>Since the first day, every discovery, every trap and every benchmark result we paid for, in one page. Each line is a fact measured on our machines, with the tutorial that explains it.</description></item><item><title>Glossary: the words of this site, in plain language</title><link>https://freedzhost.cloud/glossary.html</link><guid>https://freedzhost.cloud/glossary.html</guid><description>VRAM, container, proxy, DNS, port, token… every technical word used in the tutorials, explained in one or two sentences.</description></item><item><title>Inspirations: the projects and people we learn from</title><link>https://freedzhost.cloud/inspirations.html</link><guid>https://freedzhost.cloud/inspirations.html</guid><description>Nothing here was invented alone. The open-source projects we build on, the authors whose work made us ask better questions, and what each one taught us.</description></item><item><title>Install the NVIDIA driver and check your card</title><link>https://freedzhost.cloud/nvidia-driver.html</link><guid>https://freedzhost.cloud/nvidia-driver.html</guid><description>The one dependency everything else stands on, and how to know it really works.</description></item><item><title>FAQ: the questions everyone asks first</title><link>https://freedzhost.cloud/faq.html</link><guid>https://freedzhost.cloud/faq.html</guid><description>Do I need a big graphics card? Is it really free? Is my data safe? Where do I start? Short, honest answers, each with the tutorial that goes further.</description></item><item><title>Before you start: hardware, honestly</title><link>https://freedzhost.cloud/hardware.html</link><guid>https://freedzhost.cloud/hardware.html</guid><description>What fits on which card, with measured numbers, and what to buy first.</description></item><item><title>Support the project: donations and sponsorship</title><link>https://freedzhost.cloud/support.html</link><guid>https://freedzhost.cloud/support.html</guid><description>Every tutorial here is free, open and measured on real machines. Hardware, electricity and long nights are not. A recurring donation through Liberapay, or a sponsorship for companies, keeps the tutorials coming.</description></item><item><title>Manifesto</title><link>https://freedzhost.cloud/manifesto.html</link><guid>https://freedzhost.cloud/manifesto.html</guid><description>Why run your own AI: control, portability, and honesty about the limits.</description></item></channel></rss>
