<?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 0𝕏koji on Medium]]></title>
        <description><![CDATA[Stories by 0𝕏koji on Medium]]></description>
        <link>https://medium.com/@koji-kanao?source=rss-9004d20e450b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*1YvU02HHRfLrUNwtBPQKKw.jpeg</url>
            <title>Stories by 0𝕏koji on Medium</title>
            <link>https://medium.com/@koji-kanao?source=rss-9004d20e450b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 22:38:50 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@koji-kanao/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[How We Strengthened Dependency Security in Our Monorepo]]></title>
            <link>https://koji-kanao.medium.com/how-we-strengthened-dependency-security-in-our-monorepo-b6514af6fc77?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/b6514af6fc77</guid>
            <category><![CDATA[security]]></category>
            <category><![CDATA[github]]></category>
            <category><![CDATA[npm]]></category>
            <category><![CDATA[github-actions]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Mon, 11 May 2026 06:19:16 GMT</pubDate>
            <atom:updated>2026-05-11T06:19:16.300Z</atom:updated>
            <content:encoded><![CDATA[<h3>Strengthening Dependency Security in a Monorepo</h3><p>In response to the recent increase in supply chain attacks in the npm ecosystem, we revisited how dependencies are managed across our monorepo. This article summarizes the security improvements we implemented.</p><h3>Background</h3><p>Looking at recent attack patterns, there has been a noticeable rise in cases where malicious code is introduced into packages immediately after publication. However, in many cases, these packages are <strong>detected and removed within 24–48 hours</strong>.</p><p>This leads to a couple of important observations:</p><ul><li><strong>Using newly published packages immediately carries higher risk</strong></li><li><strong>Simply waiting for a short period significantly reduces that risk</strong></li></ul><p>While waiting 24–48 hours already provides some protection, we decided to introduce additional buffer time to account for delayed detection and cases that span weekends. As a result, we adopted a <strong>1-week cooldown period</strong>.</p><p>This approach provides:</p><ul><li>A higher likelihood that known attacks have already been mitigated before adoption</li><li>A simple and easy-to-understand operational rule</li></ul><p>Based on this, we introduced several measures centered around adding a cooldown period before adopting new dependencies.</p><h3>1. Introducing a Cooldown Period</h3><h3>Policy</h3><ul><li>Do not use newly published packages immediately</li><li>Only allow packages that are <strong>at least 1 week old</strong></li></ul><h3>Implementation by Tool</h3><h4>Yarn</h4><pre>yarn config set npmMinimalAgeGate 7</pre><p><a href="https://yarnpkg.com/configuration/yarnrc#npmMinimalAgeGate">https://yarnpkg.com/configuration/yarnrc#npmMinimalAgeGate</a></p><p>This feature is only supported in modern Yarn (Berry). As a result, this also became a good opportunity to <strong>move away from Yarn v1</strong>.</p><h4>pnpm</h4><p>Configured as follows:</p><pre># pnpm-workspace.yaml<br>minimumReleaseAge: 10080</pre><p><a href="https://pnpm.io/settings#minimumreleaseage">https://pnpm.io/settings#minimumreleaseage</a><br>uv:</p><p>This has already been applied to our workspace.</p><h4>uv (Python)</h4><pre># pyproject.toml<br>exclude-newer = &quot;1 week&quot;</pre><p><a href="https://docs.astral.sh/uv/reference/settings/#exclude-newer">https://docs.astral.sh/uv/reference/settings/#exclude-newer</a></p><p>ToolConfiguration KeyExample Valuenpmmin-release-age4320 (minutes)pnpmminimumReleaseAge4320 (minutes)YarnnpmMinimalAgeGate”3d” (duration string)BunminimumReleaseAge4320 (minutes in bunfig.toml)</p><p>bun: <a href="https://bun.com/docs/runtime/bunfig#install-minimumreleaseage">https://bun.com/docs/runtime/bunfig#install-minimumreleaseage</a></p><h3>2. Lockfile-based Dependency Management</h3><p>We ensure that all dependency installations use lockfiles:</p><ul><li>yarn.lock</li><li>pnpm-lock.yaml</li></ul><p>This allows us to:</p><ul><li>Prevent unintended version upgrades</li><li>Eliminate differences between local and CI/CD environments</li></ul><p>yarn</p><pre>yarn install --frozen-lockfile</pre><p>pnpm</p><pre>pnpm install --frozen-lockfile</pre><h3>3. Disabling postinstall Scripts</h3><p>While postinstall scripts can be useful, they are also a common entry point for supply chain attacks.</p><p>Our approach:<br>.npmrc</p><pre>ignore-scripts=true</pre><ul><li><strong>Disable postinstall execution by default</strong></li></ul><p>pnpm v10 supports this behavior natively, so no additional configuration was required.</p><h3>4. SHA Pinning in GitHub Actions</h3><p>We also treat GitHub Actions as part of our dependency surface and strengthened version pinning.</p><h3>Before</h3><pre>- uses: actions/checkout@v4</pre><h3>After</h3><pre>- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2</pre><p>By pinning to a <strong>specific commit SHA instead of a tag</strong>, we:</p><ul><li>Reduce the risk of supply chain attacks</li><li>Prevent unintended updates</li></ul><p>This is secure but it requires to update yaml files manually which is very painful.<br>So I’m planning to introduce a tool like <a href="https://github.com/mheap/pin-github-action">pin-github-action</a>.</p><h3>5. Static analysis for GitHub Actions</h3><p>As the last step, introduced <a href="https://github.com/zizmorcore/zizmor">zizmor</a> to detect GitHub Actions’ workflows’ security issues.</p><p><a href="https://docs.zizmor.sh/trophy-case/"><strong>docs.zizmor.sh</strong></a></p><h3>Summary</h3><p>Key takeaways from this effort:</p><ul><li>Introduced a <strong>1-week cooldown period</strong> for new packages</li><li>Enforced <strong>strict dependency pinning via lockfiles</strong></li><li><strong>Restricted postinstall script execution</strong></li><li>Applied <strong>SHA pinning in GitHub Actions</strong></li></ul><p>Among these, the cooldown period is particularly simple yet highly effective, and can be adopted quickly in most projects.</p><h3>Closing Thoughts</h3><p>Supply chain attacks are likely to continue increasing. However, even basic practices such as:</p><ul><li>“Don’t immediately adopt the latest version”</li><li>“Lock down dependencies”</li></ul><p>could reduce risk.</p><p>Hopefully, this helps others managing monorepos think about improving their dependency security.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b6514af6fc77" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Strengthening Dependency Security in a Monorepo]]></title>
            <link>https://koji-kanao.medium.com/strengthening-dependency-security-in-a-monorepo-6cc42201ea8c?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/6cc42201ea8c</guid>
            <category><![CDATA[github-actions]]></category>
            <category><![CDATA[security]]></category>
            <category><![CDATA[npm]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Wed, 22 Apr 2026 17:32:14 GMT</pubDate>
            <atom:updated>2026-04-22T17:32:14.929Z</atom:updated>
            <content:encoded><![CDATA[<p>In response to the recent increase in supply chain attacks in the npm ecosystem, we revisited how dependencies are managed across our monorepo. This article summarizes the security improvements we implemented.</p><h3>Background</h3><p>Looking at recent attack patterns, there has been a noticeable rise in cases where malicious code is introduced into packages immediately after publication. However, in many cases, these packages are <strong>detected and removed within 24–48 hours</strong>.</p><p>This leads to a couple of important observations:</p><ul><li><strong>Using newly published packages immediately carries higher risk</strong></li><li><strong>Simply waiting for a short period significantly reduces that risk</strong></li></ul><p>While waiting 24–48 hours already provides some protection, we decided to introduce additional buffer time to account for delayed detection and cases that span weekends. As a result, we adopted a <strong>1-week cooldown period</strong>.</p><p>This approach provides:</p><ul><li>A higher likelihood that known attacks have already been mitigated before adoption</li><li>A simple and easy-to-understand operational rule</li></ul><p>Based on this, we introduced several measures centered around adding a cooldown period before adopting new dependencies.</p><h3>1. Introducing a Cooldown Period</h3><h3>Policy</h3><ul><li>Do not use newly published packages immediately</li><li>Only allow packages that are <strong>at least 1 week old</strong></li></ul><h3>Implementation by Tool</h3><h4>Yarn</h4><pre>yarn config set npmMinimalAgeGate 7</pre><p><a href="https://yarnpkg.com/configuration/yarnrc#npmMinimalAgeGate">https://yarnpkg.com/configuration/yarnrc#npmMinimalAgeGate</a></p><p>This feature is only supported in modern Yarn (Berry). As a result, this also became a good opportunity to <strong>move away from Yarn v1</strong>.</p><h4>pnpm</h4><p>Configured as follows:</p><pre># pnpm-workspace.yaml<br>minimumReleaseAge: 10080</pre><p><a href="https://pnpm.io/settings#minimumreleaseage">https://pnpm.io/settings#minimumreleaseage</a><br>uv:</p><p>This has already been applied to our workspace.</p><h4>uv (Python)</h4><pre># pyproject.toml<br>exclude-newer = &quot;1 week&quot;</pre><p><a href="https://docs.astral.sh/uv/reference/settings/#exclude-newer">https://docs.astral.sh/uv/reference/settings/#exclude-newer</a></p><p>ToolConfiguration KeyExample Valuenpmmin-release-age4320 (minutes)pnpmminimumReleaseAge4320 (minutes)YarnnpmMinimalAgeGate”3d” (duration string)BunminimumReleaseAge4320 (minutes in bunfig.toml)</p><p>bun: <a href="https://bun.com/docs/runtime/bunfig#install-minimumreleaseage">https://bun.com/docs/runtime/bunfig#install-minimumreleaseage</a></p><h3>2. Lockfile-based Dependency Management</h3><p>We ensure that all dependency installations use lockfiles:</p><ul><li>yarn.lock</li><li>pnpm-lock.yaml</li></ul><p>This allows us to:</p><ul><li>Prevent unintended version upgrades</li><li>Eliminate differences between local and CI/CD environments</li></ul><p>yarn</p><pre>yarn install --frozen-lockfile</pre><p>pnpm</p><pre>pnpm install --frozen-lockfile</pre><h3>3. Disabling postinstall Scripts</h3><p>While postinstall scripts can be useful, they are also a common entry point for supply chain attacks.</p><p>Our approach:<br>.npmrc</p><pre>ignore-scripts=true</pre><ul><li><strong>Disable postinstall execution by default</strong></li></ul><p>pnpm v10 supports this behavior natively, so no additional configuration was required.</p><h3>4. SHA Pinning in GitHub Actions</h3><p>We also treat GitHub Actions as part of our dependency surface and strengthened version pinning.</p><h3>Before</h3><pre>- uses: actions/checkout@v4</pre><h3>After</h3><pre>- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2</pre><p>By pinning to a <strong>specific commit SHA instead of a tag</strong>, we:</p><ul><li>Reduce the risk of supply chain attacks</li><li>Prevent unintended updates</li></ul><p>This is secure but it requires to update yaml files manually which is very painful.<br>So I’m planning to introduce a tool like <a href="https://github.com/mheap/pin-github-action">pin-github-action</a>.</p><p>I created a CLI tool to make this easy for me.</p><p><a href="https://www.npmjs.com/package/kensaku">https://www.npmjs.com/package/kensaku</a></p><h3>5. Static analysis for GitHub Actions</h3><p>As the last step, introduced <a href="https://github.com/zizmorcore/zizmor">zizmor</a> to detect GitHub Actions’ workflows’ security issues.</p><p><a href="https://docs.zizmor.sh/trophy-case/"><strong>docs.zizmor.sh</strong></a></p><h3>Summary</h3><p>Key takeaways from this effort:</p><ul><li>Introduced a <strong>1-week cooldown period</strong> for new packages</li><li>Enforced <strong>strict dependency pinning via lockfiles</strong></li><li><strong>Restricted postinstall script execution</strong></li><li>Applied <strong>SHA pinning in GitHub Actions</strong></li></ul><p>Among these, the cooldown period is particularly simple yet highly effective, and can be adopted quickly in most projects.</p><h3>Closing Thoughts</h3><p>Supply chain attacks are likely to continue increasing. However, even basic practices such as:</p><ul><li>“Don’t immediately adopt the latest version”</li><li>“Lock down dependencies”</li></ul><p>could reduce risk.</p><p>Hopefully, this helps others managing monorepos think about improving their dependency security.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6cc42201ea8c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Quick Look at Claw-Family]]></title>
            <link>https://koji-kanao.medium.com/a-quick-look-at-claw-family-67e557401dcb?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/67e557401dcb</guid>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[openclaw]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Sat, 21 Mar 2026 01:01:40 GMT</pubDate>
            <atom:updated>2026-03-21T01:01:40.471Z</atom:updated>
            <content:encoded><![CDATA[<h4>OpenClaw</h4><p><a href="https://github.com/openclaw/openclaw">GitHub - openclaw/openclaw: Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞</a></p><p>I tried using OpenClaw, but I had some concerns about its security. So I decided to do a bit of research to see if there were any alternative options.<br>I’m testing `zeroclaw` now 💪</p><h4>HiClaw</h4><p><a href="https://github.com/alibaba/hiclaw">GitHub - alibaba/hiclaw: An open-source Collaborative Multi-Agent OS for transparent, human-in-the-loop task coordination via Matrix rooms.</a></p><p><strong>overview</strong><br>HiClaw is an open-source Collaborative Multi-Agent OS for transparent, human-in-the-loop task coordination via Matrix rooms.</p><h4>NemoClaw</h4><p><a href="https://github.com/NVIDIA/NemoClaw">GitHub - NVIDIA/NemoClaw: Run OpenClaw securely inside NVIDIA OpenShell with managed inference</a></p><p><strong>overview</strong><br>NVIDIA NemoClaw is an open source stack that simplifies running OpenClaw always-on assistants safely. It installs the NVIDIA OpenShell runtime, part of the NVIDIA Agent Toolkit, which provides a secure environment for running autonomous agents, with inference routed through the NVIDIA cloud.</p><h4>picoclaw</h4><p><a href="https://github.com/sipeed/picoclaw">GitHub - sipeed/picoclaw: Tiny, Fast, and Deployable anywhere - automate the mundane, unleash your creativity</a></p><p><strong>overview</strong></p><p>🦐 PicoClaw is an ultra-lightweight personal AI Assistant inspired by nanobot, refactored from the ground up in Go through a self-bootstrapping process, where the AI agent itself drove the entire architectural migration and code optimization.</p><h4>nanoclaw</h4><p><a href="https://github.com/qwibitai/nanoclaw">GitHub - qwibitai/nanoclaw: A lightweight alternative to OpenClaw that runs in containers for security. Connects to WhatsApp, Telegram, Slack, Discord, Gmail and other messaging apps,, has memory, scheduled jobs, and runs directly on Anthropic&#39;s Agents SDK</a></p><p><strong>overview</strong><br>A lightweight alternative to OpenClaw that runs in containers for security. Connects to WhatsApp, Telegram, Slack, Discord, Gmail and other messaging apps, has memory, scheduled jobs, and runs directly on Anthropic’s Agents SDK.</p><h4>zeroclaw</h4><p><a href="https://github.com/zeroclaw-labs/zeroclaw">GitHub - zeroclaw-labs/zeroclaw: Fast, small, and fully autonomous AI personal assistant infrastructure, ANY OS, ANY PLATFORM - deploy anywhere, swap anything 🦀</a></p><p><strong>overview</strong><br>Fast, small, and fully autonomous AI assistant infrastructure — deploy anywhere, swap anything 🦀</p><h4>nanobot</h4><p><a href="https://github.com/HKUDS/nanobot">GitHub - HKUDS/nanobot: Lightweight, open-source AI agent for your tools, chats, and workflows.</a></p><p><strong>overview</strong><br>🐈 nanobot is an ultra-lightweight personal AI assistant inspired by OpenClaw.</p><h4>ironclaw</h4><p><a href="https://github.com/nearai/ironclaw">GitHub - nearai/ironclaw: IronClaw is OpenClaw inspired implementation in Rust focused on privacy and security</a></p><p><strong>overview</strong><br>IronClaw is OpenClaw inspired implementation in Rust focused on privacy and security.</p><h4>nullclaw</h4><p><a href="https://github.com/nullclaw/nullclaw">GitHub - nullclaw/nullclaw: Fastest, smallest, and fully autonomous AI assistant infrastructure written in Zig</a></p><p><strong>overview</strong><br>Fastest, smallest, and fully autonomous AI assistant infrastructure written in Zig.</p><h4>moltis</h4><p><a href="https://github.com/moltis-org/moltis">GitHub - moltis-org/moltis: A Rust-native claw you can trust. One binary - sandboxed, secure, auditable. Voice, memory, MCP tools, and multi-channel access built-in.</a></p><p><strong>overview</strong><br>A Rust-native personal AI assistant with voice, memory, sandboxing, MCP tools, and multi-channel access.</p><h4>zclaw</h4><p><a href="https://github.com/tnm/zclaw">GitHub - tnm/zclaw: Your personal AI assistant at all-in 888KiB (~35KB in app code). Running on an ESP32. GPIO, cron, custom tools, memory, and more.</a></p><p><strong>overview</strong><br>Your personal AI assistant at all-in 888KiB (~35KB in app code). Running on an ESP32. GPIO, cron, custom tools, memory, and more.</p><h4>clawlet</h4><p><a href="https://github.com/mosaxiv/clawlet">GitHub - mosaxiv/clawlet: Ultra-lightweight and efficient personal AI assistant</a></p><p><strong>overview</strong><br>Ultra-lightweight and efficient personal AI assistant.</p><h4>rosclaw</h4><p><a href="https://github.com/PlaiPin/rosclaw">GitHub - PlaiPin/rosclaw: ROS2 meets OpenClaw 💗</a></p><p><strong>overview</strong><br>ROS2 meets OpenClaw.</p><h4>geminiclaw</h4><p><a href="https://github.com/e-mon/geminiclaw">GitHub - e-mon/geminiclaw: Personal autonomous agent powered by Gemini CLI. Orchestrates scheduling, persistent memory, MCP tools, and multi-channel messaging.</a></p><p><strong>overview</strong><br>Personal autonomous agent powered by Gemini CLI. Orchestrates scheduling, persistent memory, MCP tools, and multi-channel messaging.</p><h4>ClawX</h4><p><a href="https://github.com/ValueCell-ai/ClawX">GitHub - ValueCell-ai/ClawX: ClawX is a desktop app that provides a graphical interface for OpenClaw AI agents. It turns CLI-based AI orchestration into a desktop experience without using the terminal. China website is https://clawx.com.cn.</a></p><p><strong>overview</strong><br>ClawX is a desktop app that provides a graphical interface for OpenClaw AI agents. It turns CLI-based AI orchestration into a desktop experience without using the terminal.</p><h4>claw-empire</h4><p><a href="https://github.com/GreenSheep01201/claw-empire">GitHub - GreenSheep01201/claw-empire: Command Your AI Agent Empire from the CEO Desk - A local-first AI agent office simulator that orchestrates CLI, OAuth, and API-connected agents (Claude Code, Codex CLI, Gemini CLI, OpenCode, and more) as a virtual autonomous company.</a></p><p><strong>overview</strong><br>Command Your AI Agent Empire from the CEO Desk — A local-first AI agent office simulator that orchestrates CLI, OAuth, and API-connected agents (Claude Code, Codex CLI, Gemini CLI, OpenCode, and more) as a virtual autonomous company.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=67e557401dcb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Prevent Hugging Face Spaces from Sleeping with GitHub Actions + agent-browser]]></title>
            <link>https://koji-kanao.medium.com/prevent-hugging-face-spaces-from-sleeping-with-github-actions-agent-browser-dfdccabbe5d2?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/dfdccabbe5d2</guid>
            <category><![CDATA[hugging-face-spaces]]></category>
            <category><![CDATA[browser-agent]]></category>
            <category><![CDATA[hugging-face]]></category>
            <category><![CDATA[github-actions]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Tue, 17 Feb 2026 06:30:30 GMT</pubDate>
            <atom:updated>2026-02-17T06:30:30.503Z</atom:updated>
            <content:encoded><![CDATA[<p>Hugging Face Spaces are incredibly convenient for hosting demos, tools, and lightweight apps.</p><p><a href="https://huggingface.co/">Hugging Face - The AI community building the future.</a></p><p>However, there is one limitation:</p><p>If a Space is not accessed for 48 hours, it goes to sleep.</p><p>This can be problematic when:<br>• You share a demo URL and it’s slow on first access (cold start)<br>• You host automation tools like n8n<br>• You want your app to always feel “ready”<br>• You want simple uptime monitoring</p><p>To solve this, I built a small automation:</p><p>GitHub Actions accesses my Space daily<br>agent-browser opens the page and takes a screenshot<br>The screenshot is posted to Discord</p><p>This achieves both:<br>• ✅ Preventing sleep<br>• ✅ Visual uptime monitoring</p><p>In this article, I’ll show you exactly how to set it up.</p><p>⸻</p><p>Why use agent-browser?</p><p>You could simply curl the Space URL.</p><p>But I wanted:<br>• A real browser access (not just HTTP ping)<br>• To confirm the page fully renders<br>• To capture a screenshot as proof<br>• To visually verify that nothing is broken</p><p>For that, I used agent-browser, a CLI browser automation tool built on Playwright.</p><p>It works perfectly inside GitHub Actions.</p><p><a href="https://github.com/vercel-labs/agent-browser">GitHub - vercel-labs/agent-browser: Browser automation CLI for AI agents</a></p><p>The GitHub Actions Workflow</p><p>Here is the exact YAML file I use:</p><pre>on:<br>  schedule:<br>    - cron: &quot;0 0 * * *&quot;<br><br>jobs:<br>  access-hugging-face-n8n:<br>    runs-on: ubuntu-latest<br>    steps:<br>      - uses: actions/checkout@v6<br><br>      - name: Install agent-browser<br>        run: npm install -g agent-browser<br><br>      - name: Install Playwright browsers<br>        run: agent-browser install --with-deps<br><br>      - name: Go to the hugging face space<br>        run: |<br>          agent-browser open ${{ secrets.HUGGING_FACE_SPACE_URL }}<br>          agent-browser wait --text &quot;Sign in&quot;<br>          agent-browser screenshot page.png<br>          agent-browser close<br><br>      - name: Notify Discord on done access<br>        if: success()<br>        run: |<br>          curl --fail-with-body -sS -X POST \<br>            -F &#39;payload_json={&quot;content&quot;:&quot;Hugging Face access job done.&quot;}&#39; \<br>            -F &quot;file1=@page.png;type=image/png&quot; \<br>            &quot;${{ secrets.DISCORD_WEBHOOK_URL }}&quot;</pre><p>How It Works</p><ol><li>Scheduled Daily Run</li></ol><p>cron: “0 0 * * *”</p><p>This runs once per day at 00:00 UTC.</p><p>2. Install agent-browser</p><p>npm install -g agent-browser<br>agent-browser install — with-deps</p><p>This installs Playwright and required browser dependencies inside the GitHub Actions runner.</p><p>3. Open the Hugging Face Space</p><p>agent-browser open ${{ secrets.HUGGING_FACE_SPACE_URL }}<br>agent-browser wait — text “Sign in”<br>agent-browser screenshot page.png<br>agent-browser close</p><p>What this does:<br>1. Opens the Space in a real browser<br>2. Waits until “Sign in” text appears (ensures full render)<br>3. Takes a screenshot<br>4. Closes the browser</p><p>The wait — text step prevents taking screenshots before the page is fully loaded.</p><p>You can change the text to something more specific to your app.</p><p>4. Post Screenshot to Discord</p><p>curl -X POST \<br>-F ‘payload_json={“content”:”Hugging Face access job done.”}’ \<br>-F “file1=<a href="https://dev.to/page">@page</a>.png;type=image/png” \<br>“${{ secrets.DISCORD_WEBHOOK_URL }}”</p><p>This sends:<br>• A success message<br>• The screenshot image</p><p>to your Discord channel via webhook.</p><p>Now you have both:<br>• Sleep prevention<br>• Daily visual health check</p><p>⸻</p><p>Setting Up Secrets</p><p>In your GitHub repository:</p><p>Settings → Secrets and variables → Actions</p><p>Add:<br>• HUGGING_FACE_SPACE_URL<br>• DISCORD_WEBHOOK_URL</p><p>These are securely injected into the workflow using:</p><p>${{ secrets.SECRET_NAME }}</p><p>⸻</p><p>Why This Approach Is Nice</p><p>✅ Free to run<br>GitHub Actions free tier is sufficient.</p><p>✅ No server required<br>No EC2, no cron server, no maintenance.</p><p>✅ Real browser check<br>Not just ping — full rendering validation.</p><p>✅ Visual confirmation<br>You instantly see if something broke.</p><p>✅ Easy to extend<br>Add monitoring, metrics, multi-Space support.</p><p>⸻</p><p>Optional Improvements</p><p>You could extend this further:<br>• Fail the job if specific text is missing<br>• Measure page load time<br>• Upload screenshots to S3 / R2<br>• Monitor multiple Spaces in parallel<br>• Add Slack notifications<br>• Store screenshots as GitHub Action artifacts</p><p>You can even turn this into a lightweight uptime monitoring system.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*0rQA4XjQNTwNqO1V.png" /></figure><p>⸻</p><p>Conclusion</p><p>Hugging Face Spaces going to sleep can be inconvenient.</p><p>But with:</p><p>GitHub Actions + agent-browser</p><p>you can:<br>• Automatically keep your Space awake<br>• Capture daily screenshots<br>• Get notified via Discord<br>• Monitor visual health</p><p>All without running your own server.</p><p>Simple, effective, and fully automated.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dfdccabbe5d2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Use Cursor with Local LLM and LM Studio]]></title>
            <link>https://koji-kanao.medium.com/use-cursor-with-local-llm-and-lm-studio-8aad91a0a2df?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/8aad91a0a2df</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[lm-studio]]></category>
            <category><![CDATA[cursor]]></category>
            <category><![CDATA[local-llm]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Tue, 20 Jan 2026 06:45:54 GMT</pubDate>
            <atom:updated>2026-01-20T06:45:54.141Z</atom:updated>
            <content:encoded><![CDATA[<h3>Prerequisites</h3><pre>• Cursor installed<br>• LM Studio installed<br>• ngrok installed<br>• A local machine capable of running an LLMs</pre><p>For this guide, we’ll use the model: zai-org/glm-4.6v-flash</p><h3>Step 1: Install LM Studio</h3><p>Download and install LM Studio from the official website.<br>Once installed, launch the application.<br><a href="https://lmstudio.ai/">https://lmstudio.ai/</a></p><h3>Step 2: Download a Model</h3><p>Inside LM Studio, download the model you want to use.<br>For this article, we use: zai-org/glm-4.6v-flash<br>Make sure the download completes successfully before moving on.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*Is-vPE_O3aRyeJtX.png" /></figure><h3>Step 3: Install ngrok</h3><p>Install ngrok on your system.<br>ngrok allows you to expose your local server to the internet with a public URL.<br><a href="https://ngrok.com/">https://ngrok.com/</a></p><p>If you have Homebrew, you can install ngrok easily</p><pre>brew install ngrok</pre><h3>Step 4: Set Up ngrok</h3><p>Complete the ngrok setup by following the instructions on the official ngrok Setup page.<br>This typically includes:<br>• Creating an ngrok account<br>• Authenticating your local installation with an auth token</p><pre>ngrok config add-authtoken {your_token}</pre><h3>3 Step 5: Start the Local Server in LM Studio</h3><pre>1.Open LM Studio<br>2.Enable Developer Mode<br>3.Start the local server</pre><p>LM Studio will now serve your local LLM using an OpenAI-compatible API.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*DzGMDX_KAb8Et7yu.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*2ORhcCrZCQzMf8s-.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*iGLs5dEr91-XI8yS.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*QUUupFx9xdUwYu0w.png" /></figure><h3>Step 6: Expose the Local Server with ngrok</h3><p>Open a terminal and run the following command:</p><pre>ngrok http 1234</pre><p>Note: 1234 should match the port used by LM Studio’s local server.<br>After running the command, ngrok will display a public URL like:<br><a href="https://yours.ngrok-free.app/">https://yours.ngrok-free.app</a><br>Copy this URL — you’ll need it for Cursor.</p><p>You’ll see something like this</p><pre>Session Status                online                                                                                                                                                        <br>Account                       your_account (Plan: Free)                                                                                                                         <br>Version                       3.35.0                                                                                                                                                        <br>Region                        United States (us)                                                                                                                                            <br>Latency                       19ms                                                                                                                                                          <br>Web Interface                 <a href="http://127.0.0.1:4040/">http://127.0.0.1:4040</a>                                                                                                                                         <br>Forwarding                    <a href="https://something.ngrok-free.app/">https://something.ngrok-free.app</a> -&gt; <a href="http://localhost:1234/">http://localhost:1234</a>                                                                                                  </pre><pre>Connections                   ttl     opn     rt1     rt5     p50     p90                                                                                                                   <br>                              7       0       0.00    0.00    6.26    263.91                                                                                                                </pre><pre>HTTP Requests                                                                                                                                                                               <br>-------------                                                                                                                                                                               </pre><pre>20:10:37.113 EST POST /v1/chat/completions      200 OK                                                                                                                                      <br>20:06:13.115 EST POST /v1/chat/completions      200 OK                                                                                                                                      <br>20:04:59.112 EST POST /v1/chat/completions      200 OK                                                                                                                                      <br>20:04:42.221 EST POST /v1/chat/completions      200 OK                                                                                                                                      <br>20:03:11.002 EST POST /v1/chat/completions      200 OK                                                                                                                                      <br>20:03:05.636 EST POST /v1/chat/completions      200 OK                                                                                                                                      <br>20:02:22.796 EST POST /v1/chat/completions      200 OK</pre><h3>Step 7: Open Cursor Settings</h3><p>Launch Cursor and navigate to:</p><p>Settings → Models / OpenAI Configuration</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/440/0*HK0FbyQ2puhyBbiT.png" /></figure><h3>Step 8: Configure the OpenAI Base URL</h3><pre>1.Enable OpenAI API Key<br>2.Enter any placeholder value for the API key<br>•In this example, we simply use: 1234<br>3.Paste the ngrok URL into Override OpenAI Base URL<br>4.Append /v1 to the end of the URL</pre><p>Your final URL should look like this:</p><p><a href="https://yours.ngrok-free.app/v1">https://yours.ngrok-free.app/v1</a></p><p>Step 9: Add a Custom Model<br>1. Click Add Custom Model<br>2. Enter a name for your local LLM Example: GLM4.6-local</p><p>⚠️ Windows users:<br>You must enter the exact model name that LM Studio reports internally.<br>For this case, zai-org/glm-4.6v-flash</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/409/0*mtEDoEnIaU-qEu7s.png" /></figure><p>Done! 🎉</p><p>That’s it — the setup is complete.</p><p>You can now open Cursor Chat, enter a prompt, and send it. Cursor will route the request through ngrok to your local LLM running in LM Studio.</p><p>This setup allows you to enjoy Cursor’s powerful coding experience while keeping inference fully local.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/486/0*cKfgO51JdXBckHPb.png" /></figure><p>Final Thoughts</p><p>Using Cursor with a local LLM is a great way to:<br>• Reduce API costs<br>• Improve privacy<br>• Experiment with custom or open-source models</p><p>LM Studio and ngrok make the process surprisingly straightforward. Once configured, it feels almost identical to using a hosted OpenAI model — except everything runs on your own machine.</p><p>Happy hacking! 🚀</p><p><a href="https://buymeacoffee.com/koji">https://buymeacoffee.com/koji</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8aad91a0a2df" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Use Claude Opus 4.5 & Gemini 3 for Free with OpenCode]]></title>
            <link>https://koji-kanao.medium.com/how-to-use-claude-opus-4-5-gemini-3-for-free-with-opencode-b65ae340637d?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/b65ae340637d</guid>
            <category><![CDATA[open-code]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[gemini]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Tue, 13 Jan 2026 15:39:16 GMT</pubDate>
            <atom:updated>2026-01-26T06:37:37.263Z</atom:updated>
            <content:encoded><![CDATA[<p>🧵 TL;DR (Too Long; Didn’t Read)<br>Want to use powerful AI models like Claude Opus 4.5 and Gemini 3 for free? This guide walks you through setting up OpenCode with the Antigravity plugin and Google OAuth. You’ll be up and running with top-tier models in minutes — no paid API keys required.</p><h3>🚀 Getting Started with OpenCode + Antigravity Auth</h3><p><a href="https://opencode.ai/">OpenCode</a> is an open-source AI coding agent that lets you connect to models like Claude, GPT, Gemini, and more. It comes with free models out of the box, or you can bring your own from any provider.</p><p>In this guide, I’ll walk you through setting up OpenCode with Antigravity Auth using Google OAuth.</p><h3>🛠 Step 1: Install OpenCode</h3><p>You can install OpenCode using various package managers like Homebrew, npm, pnpm, yarn, scoop, choco, etc.</p><p>For this guide, I’ll use <strong>bun</strong>:</p><pre>bun install -g opencode-ai</pre><p>More install options: <a href="https://opencode.ai/docs/#install">Installation Docs</a></p><h3>⚙️ Step 2: Configure the Plugin</h3><p>We’ll use <a href="https://github.com/NoeFabris/opencode-antigravity-auth">opencode-antigravity-auth</a>. Follow the installation instructions in the repo. I’ll go with <strong>Option B</strong> here:</p><pre>vim ~/.config/opencode/opencode.json</pre><p>Add the plugin:</p><pre>{<br>  &quot;plugin&quot;: [&quot;opencode-antigravity-auth@beta&quot;]<br>}</pre><h3>🔐 Step 3: Authenticate with Google (Antigravity)</h3><p>Make sure you have a Google account that supports Antigravity.</p><p>Run:</p><pre>opencode auth login</pre><p>You’ll see a provider list like this:</p><pre>&gt; Google<br>  OpenAI<br>  Anthropic<br>  GitHub Copilot<br>  ...</pre><p>Select <strong>Google</strong>, then choose:</p><pre>&gt; OAuth with Google (Antigravity)</pre><p>Your browser will open — sign in with your Google account. Once complete, you should see:</p><pre>•  Multi-account setup complete (1 account(s)).<br>o  Login successful<br>—  Done</pre><h3>🧩 Step 4: Update Your Config</h3><p>Now let’s update the full config file to include model definitions. Open:</p><pre>vim ~/.config/opencode/opencode.json</pre><p>And paste the full configuration block from the plugin’s documentation. (You can keep the version you shared above — no changes needed unless you want to customize models.)</p><pre>{<br>  &quot;$schema&quot;: &quot;https://opencode.ai/config.json&quot;,<br>  &quot;plugin&quot;: [&quot;opencode-antigravity-auth@latest&quot;],<br>  &quot;provider&quot;: {<br>    &quot;google&quot;: {<br>      &quot;models&quot;: {<br>        &quot;antigravity-gemini-3-pro&quot;: {<br>          &quot;name&quot;: &quot;Gemini 3 Pro (Antigravity)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 1048576, &quot;output&quot;: 65535 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] },<br>          &quot;variants&quot;: {<br>            &quot;low&quot;: { &quot;thinkingLevel&quot;: &quot;low&quot; },<br>            &quot;high&quot;: { &quot;thinkingLevel&quot;: &quot;high&quot; }<br>          }<br>        },<br>        &quot;antigravity-gemini-3-flash&quot;: {<br>          &quot;name&quot;: &quot;Gemini 3 Flash (Antigravity)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 1048576, &quot;output&quot;: 65536 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] },<br>          &quot;variants&quot;: {<br>            &quot;minimal&quot;: { &quot;thinkingLevel&quot;: &quot;minimal&quot; },<br>            &quot;low&quot;: { &quot;thinkingLevel&quot;: &quot;low&quot; },<br>            &quot;medium&quot;: { &quot;thinkingLevel&quot;: &quot;medium&quot; },<br>            &quot;high&quot;: { &quot;thinkingLevel&quot;: &quot;high&quot; }<br>          }<br>        },<br>        &quot;antigravity-claude-sonnet-4-5&quot;: {<br>          &quot;name&quot;: &quot;Claude Sonnet 4.5 (Antigravity)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 200000, &quot;output&quot;: 64000 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] }<br>        },<br>        &quot;antigravity-claude-sonnet-4-5-thinking&quot;: {<br>          &quot;name&quot;: &quot;Claude Sonnet 4.5 Thinking (Antigravity)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 200000, &quot;output&quot;: 64000 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] },<br>          &quot;variants&quot;: {<br>            &quot;low&quot;: { &quot;thinkingConfig&quot;: { &quot;thinkingBudget&quot;: 8192 } },<br>            &quot;max&quot;: { &quot;thinkingConfig&quot;: { &quot;thinkingBudget&quot;: 32768 } }<br>          }<br>        },<br>        &quot;antigravity-claude-opus-4-5-thinking&quot;: {<br>          &quot;name&quot;: &quot;Claude Opus 4.5 Thinking (Antigravity)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 200000, &quot;output&quot;: 64000 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] },<br>          &quot;variants&quot;: {<br>            &quot;low&quot;: { &quot;thinkingConfig&quot;: { &quot;thinkingBudget&quot;: 8192 } },<br>            &quot;max&quot;: { &quot;thinkingConfig&quot;: { &quot;thinkingBudget&quot;: 32768 } }<br>          }<br>        },<br>        &quot;gemini-2.5-flash&quot;: {<br>          &quot;name&quot;: &quot;Gemini 2.5 Flash (Gemini CLI)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 1048576, &quot;output&quot;: 65536 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] }<br>        },<br>        &quot;gemini-2.5-pro&quot;: {<br>          &quot;name&quot;: &quot;Gemini 2.5 Pro (Gemini CLI)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 1048576, &quot;output&quot;: 65536 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] }<br>        },<br>        &quot;gemini-3-flash-preview&quot;: {<br>          &quot;name&quot;: &quot;Gemini 3 Flash Preview (Gemini CLI)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 1048576, &quot;output&quot;: 65536 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] }<br>        },<br>        &quot;gemini-3-pro-preview&quot;: {<br>          &quot;name&quot;: &quot;Gemini 3 Pro Preview (Gemini CLI)&quot;,<br>          &quot;limit&quot;: { &quot;context&quot;: 1048576, &quot;output&quot;: 65535 },<br>          &quot;modalities&quot;: { &quot;input&quot;: [&quot;text&quot;, &quot;image&quot;, &quot;pdf&quot;], &quot;output&quot;: [&quot;text&quot;] }<br>        }<br>      }<br>    }<br>  }<br>}</pre><h3>🧠 Step 5: Select Claude Opus 4.5</h3><p>Launch OpenCode:</p><pre>opencode</pre><p>Then press Ctrl + P and choose:</p><pre>Switch model → Claude Opus 4.5 (Antigravity)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/479/0*7v05K760GE638wZK.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*NYgxukJZsUslz8l2.png" /></figure><p>That’s it! You’re now ready to use OpenCode with Antigravity-authenticated models. Happy coding! 🌱</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b65ae340637d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Migrating from asdf to mise without the headaches]]></title>
            <link>https://koji-kanao.medium.com/migrating-from-asdf-to-mise-without-the-headaches-fad759f33dce?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/fad759f33dce</guid>
            <category><![CDATA[tools]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[mises]]></category>
            <category><![CDATA[cli]]></category>
            <category><![CDATA[asdf]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Sun, 04 Jan 2026 03:18:03 GMT</pubDate>
            <atom:updated>2026-01-04T03:18:03.779Z</atom:updated>
            <content:encoded><![CDATA[<h3>1. Remove asdf plugins and uninstall asdf</h3><pre># remove plugins<br>asdf plugin list | xargs -n 1 asdf plugin remove</pre><pre># uninstall asdf<br>brew uninstall asdf --force<br>brew autoremove</pre><p>clean up .zshrc</p><pre># ASDF configuration<br>. &quot;$BREW_PREFIX/opt/asdf/libexec/asdf.sh&quot;</pre><p>remove files</p><pre>rm -rf ~/.asdf<br>rm -rf ~/.tool-versions</pre><p>Confirmed that asdf has been deleted.</p><pre>brew list asdf</pre><pre># if you can remove asdf successfully, you will see something like this<br>Error: No such keg: /opt/homebrew/Cellar/asdf<br>Execution time: 0.88s</pre><h3>2. Install mise</h3><p><a href="https://github.com/jdx/mise">https://github.com/jdx/mise</a></p><p><a href="https://mise.jdx.dev/getting-started.html">https://mise.jdx.dev/getting-started.html</a></p><pre>brew install mise</pre><pre># if you install mise successfully, you can see this when running version command<br> mise --version<br>              _                                        __              <br>   ____ ___  (_)_______        ___  ____        ____  / /___ _________<br>  / __ `__ \/ / ___/ _ \______/ _ \/ __ \______/ __ \/ / __ `/ ___/ _ \<br> / / / / / / (__  )  __/_____/  __/ / / /_____/ /_/ / / /_/ / /__/  __/<br>/_/ /_/ /_/_/____/\___/      \___/_/ /_/     / .___/_/\__,_/\___/\___/<br>                                            /_/                 by @jdx<br>2025.12.12 macos-arm64 (2025-12-18)<br>Execution time: 0.90s</pre><p>add the following to .zshrc. If you use bash or fish, you need to replace zsh with bash/fish</p><pre>eval &quot;$(mise activate zsh)&quot;</pre><h3>3. Install mise plugins</h3><p>In my case, I need to install nodejs, pnpm, bun, yarn, and python.</p><p>First check available plugins</p><pre>mise plugins ls-remote</pre><p>Also check the core list</p><pre>mise plugins ls --core<br>bun<br>deno<br>elixir<br>erlang<br>go<br>java<br>node<br>python<br>ruby<br>rust<br>swift<br>zig</pre><p>Then check available version for each plugin</p><pre># Check available nodes version<br>mise ls-remote node<br>mise ls-remote pnpm<br>mise ls-remote bun<br>mise ls-remote yarn<br>mise ls-remote python</pre><p>Finally install</p><pre>mise install node@22.21.1<br>mise install yarn@1.22.22<br>mise install pnpm@10.22.0<br>mise install bun@1.3.5<br>mise install python@3.12.12</pre><p>or you can install them via mise.toml</p><ol><li>create mise.toml in your repo folder</li><li>add [tools] to mise.toml For this case, the toml file is like below</li></ol><pre>[tools]<br>node = &quot;22.21.1&quot;<br>python = &quot;3.12.12&quot;<br>yarn = &quot;1.22.22&quot;<br>pnpm = &quot;10.22.0&quot;</pre><ol><li>install versions</li></ol><pre>mise ERROR error parsing config file: ~/Desktop/dev_test/blog/mise.toml<br>mise ERROR Config files in ~/Desktop/dev_test/blog/mise.toml are not trusted.<br>Trust them with `mise trust`. See https://mise.jdx.dev/cli/trust.html for more information.<br>mise ERROR Run with --verbose or MISE_VERBOSE=1 for more information</pre><pre># if you see the above Error message, you need to run the trust command to accept the config file.<br>mise trust</pre><pre>mise install</pre><h3>Check versions</h3><pre>yarn -v<br>1.22.22<br>Execution time: 0.17s  </pre><pre>node -v<br>v22.21.1<br>Execution time: 0.03s    </pre><pre>python --version<br>Python 3.12.12<br>Execution time: 0.04s</pre><p>If you see something like this, probably you are missing adding the eval command to your .zshrc</p><pre>node -v<br>zsh: command not found: node</pre><p>You need to add the following to .zshrc.</p><pre>eval &quot;$(mise activate zsh)&quot;</pre><h3>4. asdf → mise command mapping</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*vSA2vxvTv0tuoQPh0BKyIg.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fad759f33dce" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[gemini is not recognized on Windows ]]></title>
            <link>https://koji-kanao.medium.com/gemini-is-not-recognized-on-windows-7aa21ee4b8a5?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/7aa21ee4b8a5</guid>
            <category><![CDATA[windows]]></category>
            <category><![CDATA[gemini]]></category>
            <category><![CDATA[miss]]></category>
            <category><![CDATA[gemini-cli]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Thu, 10 Jul 2025 06:13:47 GMT</pubDate>
            <atom:updated>2025-07-10T06:13:47.512Z</atom:updated>
            <content:encoded><![CDATA[<p>When I tried to run the Gemini CLI, the command wasn’t recognized on Windows 11, even though I could install Gemini CLI without any errors.</p><pre>$ gemini -v<br>gemini: The term &#39;gemini&#39; is not recognized as a name of a cmdlet, function, script file, or executable program.<br>Check the spelling of the name, or if a path was included, verify that the path is correct and try again.</pre><pre>npm install -g @google/gemini-cli</pre><p>First, I tried to check the path of the Gemini CLI</p><pre>$ which gemini<br>which: The term &#39;which&#39; is not recognized as a name of a cmdlet, function, script file, or executable program.<br>Check the spelling of the name, or if a path was included, verify that the path is correct and try again.</pre><p>Then I noticed why I was facing this issue, because of mise.</p><p><a href="https://github.com/jdx/mise">GitHub - jdx/mise: dev tools, env vars, task runner</a></p><p>I tried to check the path of the Gemini CLI again with mise</p><pre>$ mise which gemini<br>C:\Users\username\scoop\apps\mise\current\mise\installs\node\22.14.0\gemini.cmd</pre><p>When I used the above path, my Warp showed Gemini.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*J52H3y_HF9AqgXFHTyC_Fw.png" /></figure><p>To avoid the hassle of calling the path every time, I defined it as a function in the profile so it can be executed with Gemini.</p><pre>code $HOME\Documents\PowerShell\Microsoft.PowerShell_profile.ps1</pre><p>Added the following to profile</p><pre># Gemini CLI function<br>function Invoke-Gemini {<br>    C:\Users\username\scoop\apps\mise\current\mise\installs\node\22.14.0\gemini.cmd @args<br>}<br><br>Set-Alias -Name gemini -Value Invoke-Gemini</pre><pre># check gemini cli version<br>$ gemini -v<br>0.1.10    </pre><pre># test a simple prompt<br>$ gemini -p &quot;tell me about Gemni CLI&quot;<br>Loaded cached credentials.<br>I am Gemini, a large language model from Google, accessed through this command-line interface.<br><br>My purpose is to assist you with software engineering tasks. I can interact with your local file system to:<br>*   Read, write, and search through your code.<br>*   Execute shell commands to run builds, tests, or other tools.<br>*   Help you with tasks like debugging, refactoring, adding features, or creating new applications.<br><br>I operate by analyzing your requests and the provided project context, then using my available tools to help you achieve your goal.</pre><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7aa21ee4b8a5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Set up DeepWiki MCP Server]]></title>
            <link>https://koji-kanao.medium.com/how-to-set-up-deepwiki-mcp-server-5fb27aa85c7e?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/5fb27aa85c7e</guid>
            <category><![CDATA[mcp-server]]></category>
            <category><![CDATA[windsurf]]></category>
            <category><![CDATA[mcp-client]]></category>
            <category><![CDATA[cursor]]></category>
            <category><![CDATA[deepwiki]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Wed, 28 May 2025 07:13:51 GMT</pubDate>
            <atom:updated>2025-05-28T07:13:51.569Z</atom:updated>
            <content:encoded><![CDATA[<p><a href="https://docs.devin.ai/work-with-devin/deepwiki-mcp">DeepWiki MCP - Devin Docs</a></p><p>DeepWiki released a remote MCP server. Now we can use DeepWiki from MCP clients such as Claude Desktop, Cursor, Windsurf, VSCode(Cline, Roo code), Tare, etc.</p><h4>Available Tools</h4><p>The DeepWiki MCP server offers three main tools:</p><ol><li>read_wiki_structure - Get a list of documentation topics for a GitHub repository</li><li>read_wiki_contents - View documentation about a GitHub repository</li><li>ask_question - Ask any question about a GitHub repository and get an AI-powered, context-grounded response</li></ol><h3>How to set up DeepWiki</h3><p>The DeepWiki MCP server is a free, rmeove, no-auth and it supports SSE(Sever Sent Events) so we just need to add the URL to use DeepWiki MCP sercer which is very easy to setup.</p><h3>Claude Desktop</h3><pre>{<br>  &quot;mcpServers&quot;: {<br>    &quot;deepwiki&quot;: {<br>      &quot;command&quot;: &quot;npx&quot;,<br>      &quot;args&quot;: [<br>        &quot;mcp-remote&quot;,<br>        &quot;https://mcp.deepwiki.com/sse&quot;<br>      ]<br>    }<br>  }<br>}</pre><h3>Cursor</h3><pre>{<br>  &quot;mcpServers&quot;: {<br>    &quot;deepwiki&quot;: {<br>      &quot;url&quot;: &quot;https://mcp.deepwiki.com/sse&quot;<br>    }<br>  }<br>}</pre><h3>Windsurf</h3><pre>{<br>  &quot;mcpServers&quot;: {<br>    &quot;deepwiki&quot;: {<br>      &quot;serverUrl&quot;: &quot;https://mcp.deepwiki.com/sse&quot;<br>    }<br>  }<br>}</pre><h3>Cline</h3><pre>{<br>  &quot;mcpServers&quot;: {<br>    &quot;deepwiki&quot;: {<br>      &quot;url&quot;: &quot;https://mcp.deepwiki.com/sse&quot;,<br>      &quot;disabled&quot;: false,<br>      &quot;autoApprove&quot;: []<br>    }<br>  }<br>}</pre><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5fb27aa85c7e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Easily Export and Import VSCode Extensions with a Shell Script]]></title>
            <link>https://koji-kanao.medium.com/easily-export-and-import-vscode-extensions-with-a-shell-script-cc597af0af2a?source=rss-9004d20e450b------2</link>
            <guid isPermaLink="false">https://medium.com/p/cc597af0af2a</guid>
            <category><![CDATA[vscode-extension]]></category>
            <category><![CDATA[shell]]></category>
            <category><![CDATA[vscode]]></category>
            <category><![CDATA[productivity]]></category>
            <dc:creator><![CDATA[0koji]]></dc:creator>
            <pubDate>Mon, 10 Feb 2025 19:05:47 GMT</pubDate>
            <atom:updated>2025-02-10T19:05:47.368Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to Export and Import VSCode Extensions Using a Shell Script</h3><p>If you use Visual Studio Code across multiple devices or frequently reinstall your system, manually reinstalling extensions can be tedious. This simple shell script automates the process by exporting and importing your VSCode extensions with a single command.</p><h3>Step 1: Save the Script</h3><p>Copy the following script and save it as vscode-extension-importer-exporter.sh in your home directory:</p><pre>#!/bin/bash<br><br>echo &quot;Do you want to export or import VSCode extensions? (export/import)&quot;<br>read action<br>if [ &quot;$action&quot; == &quot;export&quot; ]; then<br>  echo &quot;Exporting VSCode extensions...&quot;<br>  code --list-extensions &gt; vscode-extensions.txt<br>  echo &quot;Extensions have been exported to vscode-extensions.txt&quot;<br>elif [ &quot;$action&quot; == &quot;import&quot; ]; then<br>  echo &quot;Importing VSCode extensions...&quot;<br>  while IFS= read -r extension; do<br>    code --install-extension &quot;$extension&quot;<br>  done &lt; vscode-extensions.txt<br>  echo &quot;Extensions have been imported from vscode-extensions.txt&quot;<br>else<br>  echo &quot;Invalid action. Please choose &#39;export&#39; or &#39;import&#39;.&quot;<br>fi</pre><h3>Step 2: Make the Script Executable</h3><p>After saving the script, give it execution permission by running:</p><pre>chmod +x vscode-extension-importer-exporter.sh</pre><h3>Step 3: Run the Script</h3><p>To execute the script, use the following command:</p><pre>./vscode-extension-importer-exporter.sh</pre><h3>How It Works</h3><p>When you run the script, you will be prompted to choose between exporting or importing VSCode extensions:</p><ul><li><strong>Exporting:</strong> The script lists all installed VSCode extensions and saves them to a file named vscode-extensions.txt.</li><li><strong>Importing:</strong> The script reads the vscode-extensions.txt file and installs each extension automatically.</li></ul><h4>Example Output of vscode-extensions.txt</h4><pre>ms-azuretools.vscode-docker<br>ms-edgedevtools.vscode-edge-devtools<br>ms-python.autopep8<br>ms-python.debugpy<br>ms-python.isort<br>ms-python.python<br>ms-python.vscode-pylance<br>ms-vscode-remote.remote-containers<br>ms-vscode.makefile-tools<br>ms-vscode.vscode-typescript-next</pre><h3>Why Use This Script?</h3><ul><li>It saves time when setting up a new machine.</li><li>Ensures you have the same extensions across multiple devices.</li><li>Eliminates the need to manually search and install extensions.</li></ul><p>With this simple script, managing VSCode extensions becomes effortless! 🚀</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cc597af0af2a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>