neuron-explorer
Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
翻译准备中
尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Chrome Plugin to syntax highlight Spice on GitHub
翻译准备中
取自仓库 README
正在加载完整 README。
查看完整 README ↗A Chrome extension that brings the Spice programming language to GitHub until GitHub Linguist supports it natively:
*.spice files```spice in READMEs, issues, comments, etc.The highlighting uses GitHub's own syntax theme colors, so it matches the selected light/dark theme.
chrome://extensions, enable Developer mode (top right).extension directory.Click the extension icon to open the options: toggle the features, choose the Spice color for the language bar and optionally set a GitHub token.
The language statistics are computed with the GitHub REST API. Without a token, GitHub allows 60 requests per hour
and only public repositories can be analyzed. Responses are cached and revalidated with conditional requests, which
do not count against the limit, so this is usually enough for browsing. For private repositories or heavy use, create
a fine-grained personal access token with read-only
Contents permission. The token is stored locally (chrome.storage.local) and only sent to api.github.com.
extension/shared/tokenizer.js is a lossless tokenizer following the lexer rules of the Spice compiler
(src-host/Spice.g4). The content script finds Spice code in the page and wraps the tokens in place by splitting text
nodes, so GitHub's own markup (word-level diff markers, code navigation) stays intact. Block comments and attribute
lists spanning multiple lines are handled via the lexer state: in file views it is computed from the full file
source, in diffs it is tracked separately for the old and new version of the file. Since GitHub renders lazily and
navigates without page reloads, a MutationObserver re-runs the (idempotent) highlighting whenever the page changes.
The background service worker fetches GET /repos/{owner}/{repo}/languages (the bytes per language GitHub knows) and
the recursive git tree of the default branch, which contains the size of every file. The Spice bytes are computed
like Linguist would:
*.spice files are counted, symlinks and submodules are notnode_modules/, third_party/) and documentation paths (e.g. docs/, examples/) are
excluded, using Linguist's own vendor.yml and documentation.yml patterns.gitattributes is respected: linguist-vendored, linguist-documentation, linguist-generated,
linguist-detectable and linguist-language.spice files mapped to another language via linguist-language (a common workaround to get highlighting) are
moved from that language to Spice, so they are not counted twiceThe languages GitHub displayed by name stay visible, the rest is grouped into "Other", like on GitHub.
Limitations compared to Linguist: content-based detection of generated files is not possible without downloading
every file, nested .gitattributes files are ignored, and for very large repositories the GitHub API truncates the
tree (the Spice entry's tooltip notes this).
npm install # installs jsdom for the DOM tests
npm test # runs the tests (the tokenizer is also checked against ../spice if present, or $SPICE_REPO)
npm run update-linguist # regenerates extension/shared/linguist-data.js from upstream Linguist
The extension has no build step: the files in extension/ are loaded as they are. After a change, click the reload
button of the extension in chrome://extensions and reload the GitHub tab.
Project layout:
| Path | Purpose |
|---|---|
extension/manifest.json | Manifest V3 |
extension/background.js | GitHub API access and caching |
extension/content/highlight.js | Syntax highlighting of files, diffs, snippets and Markdown |
extension/content/languages.js | Language bar in the repository sidebar |
extension/content/main.js | Content script entry point, settings and change observation |
extension/shared/tokenizer.js | Spice tokenizer |
extension/shared/linguist.js | Linguist-like file classification, .gitattributes and merging |
extension/shared/linguist-data.js | Generated: language colors, aliases, vendor/documentation paths |
extension/options/ | Options page |
MIT, see LICENSE.