[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80441":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":17,"compositeScore":18,"rankGlobal":10,"rankLanguage":10,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":30,"readmeContent":31,"aiSummary":32,"trendingCount":15,"starSnapshotCount":15,"syncStatus":33,"lastSyncTime":34,"discoverSource":35},80441,"code-stick","MuhammadUsmanGM\u002Fcode-stick","MuhammadUsmanGM","Plug in a USB. Get an offline AI coding agent on any laptop - Windows, macOS, Linux. opencode + Ollama, pre-built for 5 targets, zero install on the host.","https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fcode-stick",null,"TypeScript",57,12,54,0,1,3,3.34,"Other",false,"main",true,[24,25,26,27,28,29],"coding-agent","local-llm","offline-ai","ollama","opencode","usb","2026-06-12 02:04:02","# code-stick\n\n> **Plug in a USB. Get an offline AI coding agent on any laptop.**\n> One stick. Six targets — **Windows (x64 + ARM64), macOS (Apple Silicon + Intel), Linux (x64 + ARM64)**. Zero install on the host.\n\n[![npm](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fv\u002Fcode-stick.svg)](https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Fcode-stick)\n[![license](https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fl\u002Fcode-stick.svg)](.\u002FLICENSE)\n[![node](https:\u002F\u002Fimg.shields.io\u002Fnode\u002Fv\u002Fcode-stick.svg)](https:\u002F\u002Fnodejs.org)\n\n![code-stick demo](.\u002Fpublic\u002Fcode-stick.gif)\n\n```bash\nnpx code-stick install\n```\n\n## Why\n\nYou want an AI coding agent that works on:\n\n- **Airgapped or restricted machines** — banks, defense, hospitals, lab VMs.\n- **Shared \u002F borrowed laptops** — internships, school computers, client sites.\n- **Spotty wifi** — flights, trains, cafes, conferences.\n- **Privacy-sensitive code** — closed-source clients, NDAs, personal projects.\n\nCloud agents won't work. Installing Ollama + a 5 GB model on every machine you touch won't either. code-stick is the in-between: install once on a USB, then run the agent from the stick on whatever laptop is in front of you. Host stays clean.\n\nUnder the hood it's [opencode](https:\u002F\u002Fopencode.ai) (terminal coding agent) talking to [Ollama](https:\u002F\u002Follama.com) (local model runner), both pre-built for every target, both serving from the USB on `127.0.0.1:11434`. Quit the agent, the Ollama process dies, the host has no residue.\n\n## Install + first launch\n\n```bash\nnpx code-stick install            # auto-detects USB; pick a model interactively\n```\n\nPlug the stick into any supported machine, double-click the launcher\nfor that OS:\n\n- `start-windows.bat`\n- `start-mac.command`\n- `start-linux.sh`\n\nopencode runs in the terminal talking to a USB-local Ollama on\n`127.0.0.1:11434`. Quitting opencode kills the Ollama process. Nothing is\nleft behind on the host.\n\n## Models\n\nThe recommended default is **Qwen2.5-Coder 7B** — fits a 32 GB stick, runs\non 8 GB RAM, best all-rounder for code. Other curated options scale up\nto 32B for serious refactors on larger sticks.\n\n| Model               | Size    | Stick \u002F RAM    | Best for                          |\n| ------------------- | ------- | -------------- | --------------------------------- |\n| Qwen2.5-Coder 7B ⭐  | ~4.7 GB | 32 GB \u002F 8 GB   | All-rounder for coding            |\n| Qwen2.5-Coder 14B   | ~9.0 GB | 64 GB \u002F 16 GB  | Stronger reasoning + multi-file   |\n| Qwen2.5-Coder 32B   | ~20 GB  | 128 GB \u002F 32 GB | Top-tier OSS coder, near-frontier |\n| Phi-3 Mini 3.8B     | ~2.2 GB | 32 GB \u002F 4 GB   | Lightweight, low-spec hardware    |\n\nFull curated list, BYO Ollama tag instructions, and per-model context\nwindows: [`docs\u002FMODELS.md`](docs\u002FMODELS.md).\n\n> ⚠ **USB 2.0 sticks are unusable for models >7B.** ~30 MB\u002Fs read means\n> a 20 GB model takes ~11 minutes just to warm up. Use **USB 3.0 or\n> better** for anything bigger than 7B.\n\n> ⚠ **First-prompt latency on large models** can be 30–90 s — Ollama is\n> mmap'ing the weights off the USB. Subsequent prompts are fast once the\n> OS page-caches the blob.\n\n> ⚠ **macOS first-launch:** Gatekeeper will block the unsigned launcher\n> (notarization is on the roadmap). Right-click `start-mac.command` →\n> Open → Open. Details: [`docs\u002FTROUBLESHOOTING.md`](docs\u002FTROUBLESHOOTING.md#macos-gatekeeper-blocks-the-launcher).\n\n## Commands\n\n| Command                            | Description                                                                |\n| ---------------------------------- | -------------------------------------------------------------------------- |\n| `code-stick install`               | Set up code-stick on a USB                                                 |\n| `code-stick start`                 | Start opencode + Ollama from a USB                                         |\n| `code-stick status`                | Show what's installed                                                      |\n| `code-stick doctor`                | Live audit (port + Ollama + opencode + model store)                        |\n| `code-stick prune`                 | Run `ollama prune` on the USB model store to reclaim orphaned blobs        |\n| `code-stick update`                | Refresh launcher scripts and manifest timestamp                            |\n| `code-stick upgrade-engine`        | Re-download Ollama + opencode without nuking the model store               |\n| `code-stick add-model [id]`        | Pull another model onto the stick                                          |\n| `code-stick remove-model [id]`     | Remove a model from the stick                                              |\n| `code-stick add-targets [list]`    | Add OS targets to a stick installed with `--targets` (restore portability) |\n| `code-stick uninstall`             | Wipe code-stick from the stick                                             |\n\nFull flag reference, install modes (Fast vs Direct), and `--targets`\ntrimming: [`docs\u002FCOMMANDS.md`](docs\u002FCOMMANDS.md).\n\n## Requirements\n\n- **USB free space:** 8+ GB only with `--targets host` (see\n  [`docs\u002FSTORAGE.md`](docs\u002FSTORAGE.md)); 16+ GB for full portability with\n  small models; 32+ GB (recommended default + Qwen 7B); 64+ GB (medium), 128+\n  GB (large).\n- **USB format:** exFAT or NTFS. FAT32's 4 GB file limit blocks models\n  above ~4 GB; the installer detects FAT32 and bails clearly.\n- **Install machine:** Node 20+. Target machines need **nothing**.\n- **Target laptop RAM:** rule of thumb is model size × 1.2. See the\n  per-model column in [`docs\u002FMODELS.md`](docs\u002FMODELS.md).\n\n## Status\n\nv0.2.1 — opencode v1.x, per-model `num_ctx` bake. Full install flow on\nWindows (x64 + ARM64), macOS, Linux (x64 + ARM64) — tested manually per\ntarget. Surface Pro \u002F Snapdragon X laptops boot natively (ARM64) with\nx64-via-Prism fallback if only the x64 target is staged. Docker-based\nLinux smoke test in CI. No telemetry, no background HTTP. Bug reports\nare local-only and redacted before write.\n\nKnown gaps: macOS not yet notarized, no Windows\u002FmacOS CI runners.\n\nFile an issue or open a PR if you hit anything:\n[github.com\u002FMuhammadUsmanGM\u002Fcode-stick\u002Fissues](https:\u002F\u002Fgithub.com\u002FMuhammadUsmanGM\u002Fcode-stick\u002Fissues).\n\n## Documentation\n\n- [`docs\u002FSTORAGE.md`](docs\u002FSTORAGE.md) — USB sizing, default 6-target download (~5–7 GB binaries), 8 GB + `--targets host`\n- [`docs\u002FTROUBLESHOOTING.md`](docs\u002FTROUBLESHOOTING.md) — hallucinations, Gatekeeper, MAX_PATH, port conflicts, bug reports\n- [`docs\u002FMODELS.md`](docs\u002FMODELS.md) — full model catalog, BYO Ollama tags, quantization, context windows\n- [`docs\u002FCOMMANDS.md`](docs\u002FCOMMANDS.md) — every subcommand and flag, Fast vs Direct, `--targets`\n- [`docs\u002FARCHITECTURE.md`](docs\u002FARCHITECTURE.md) — stick layout, runtime behavior, process isolation\n- [`docs\u002FSECURITY.md`](docs\u002FSECURITY.md) — trust model, threat boundaries\n- [`docs\u002FTRUST.md`](docs\u002FTRUST.md) — supply-chain provenance\n\n## Development\n\n```bash\nnpm install\nnpm run typecheck     # tsc --noEmit\nnpm test              # vitest run (unit + launcher snapshot tests)\nnpm run build         # tsup → dist\u002Fcli.js\nnpm run smoke:docker  # full end-to-end install + launch in Linux Docker\n```\n\n## License\n\nMIT — Muhammad Usman ([github.com\u002FMuhammadUsmanGM](https:\u002F\u002Fgithub.com\u002FMuhammadUsmanGM))\n","code-stick 是一个通过USB驱动器在任何笔记本电脑上提供离线AI编码助手的项目。其核心功能是将AI编码助手和本地模型运行器Ollama预装在一个USB驱动器中，支持Windows、macOS和Linux多个平台，无需在主机上安装任何软件。该项目特别适用于网络隔离或受限环境、共享或借用的计算机、网络不稳定的地方以及处理敏感代码的场景。使用TypeScript编写，支持多种模型选择，推荐默认模型为Qwen2.5-Coder 7B，适用于大多数编码任务。",2,"2026-06-11 04:00:45","CREATED_QUERY"]