neuron-explorer
Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
翻譯準備中
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證SI Hive(Superintelligence Hive):一個本機儀表板,可平行執行 Claude Code、Codex 與 Gemini CLI 代理程式,具備工作流程、技能與團隊功能。
AI 翻譯
SI Hive (Superintelligence Hive): a local dashboard for running Claude Code, Codex and Gemini CLI agents in parallel, with workflows, skills and team features.
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
A local mission-control framework for AI agents. Website: superintelligencehive.com
SI Hive is a web dashboard that runs on your machine and gives you visibility and control over every Claude Code (and Codex / Gemini CLI) session: live terminals, session boards, prompt queues, schedules, skills, agents, plugins, and more. It works out of the box for one person on one machine, and grows into a team tool by switching on optional modules — login, a shared database, git hosting, ticket tracking, and an LLM API — all chosen in Settings.檢視完整 README ↗A local mission-control framework for AI agents. Website: superintelligencehive.com
SI Hive is a web dashboard that runs on your machine and gives you visibility and control over every Claude Code (and Codex / Gemini CLI) session: live terminals, session boards, prompt queues, schedules, skills, agents, plugins, and more. It works out of the box for one person on one machine, and grows into a team tool by switching on optional modules — login, a shared database, git hosting, ticket tracking, and an LLM API — all chosen in Settings.
It is meant to be forked and extended: providers (git hosts, trackers, login systems, LLM backends) and feature modules plug in through small typed interfaces, so adding "your" service is a contained change — often one you can simply ask an AI to make.
Core (always on, local only)
Optional modules (off until configured)
| Dashboard — what needs you right now, and every live session | Session Board — sessions by status, with the tool each agent is running |
| Projects — local repos with branch, session counts and one-click launch | Prompt History — every prompt, searchable and filterable by project |
| AI Studio: Agents — manage custom subagents | AI Studio: Skills — manage skills and sync them to other AI CLIs |
| AI Studio: Plugins — enable and disable installed plugins | AI Studio: Permissions — launch flags, presets and allow / deny / ask rules |
| Community Marketplace — browse and install community skills, agents and plugins | Workflows — scheduled automations from templates or plain English |
git clone <your-fork-url> hive
cd hive
npm install
npm run build:hive
HIVE_PORT=4747 npx tsx apps/hive/server/index.ts
Open http://localhost:4747. SI Hive discovers Claude Code sessions from ~/.claude/ automatically — no configuration needed. With login off (the default) the server only listens on loopback.
For an installed background service, build the Windows installer (installer\build-installer.cmd) or the macOS bundle (installer/mac/).
The Windows installer asks how SI Hive should run:
To switch an existing service install to login mode, run this from an administrator PowerShell, signed in as the user SI Hive should run as. It carries over the service's HIVE_* settings, stops the service, sets it to manual, and starts SI Hive:
& "C:\Program Files (x86)\SI Hive\node\node.exe" "C:\Program Files (x86)\SI Hive\app\scripts\install-autostart-windows.cjs" --replace-service
To switch back: ...\scripts\uninstall-autostart-windows.cjs --stop --restore-service. For a source checkout, run the same scripts with your own node.
Everything lives in the SI Hive data directory — ~/.hive/ by default, or $HIVE_HOME if set (use this to run a second, isolated instance):
| File | Contents |
|---|---|
config.json | Settings (non-secret): projects, integrations, modules, UI preferences |
credentials.json | Encrypted secrets (tokens, passwords, API keys) — AES-256-GCM |
secret.key | Key for credentials.json (or set HIVE_SECRET_KEY) |
hive.db | Local SQLite database |
Useful environment variables: HIVE_PORT, HIVE_HOME, HIVE_HOST (bind address), HIVE_ALLOWED_HOSTS, HIVE_PUBLIC_URL (external URL for sign-in callbacks), HIVE_SECRET_KEY, HIVE_AUTH_DISABLED=1 (recovery), HIVE_SKIP_BUNDLED_SKILLS=1.
| Extension point | Where | Guide |
|---|---|---|
| Git host / ticket tracker | apps/hive/server/integrations/ | docs/extending/adding-a-provider.md |
| Feature module | apps/hive/server/modules/registry.ts + your module folder | docs/extending/modules.md |
| AI CLI provider | apps/hive/server/providers/ | — |
npm run lint runs scripts/check-generic.mjs, which flags personal home-folder paths and real email addresses. To keep your organization's names, hosts and IDs out too, list them (one regex per line) in scripts/check-generic.local.txt — it is gitignored, so the list itself is never published.
To share one SI Hive between several people:
HIVE_PUBLIC_URL.HIVE_SECRET_KEY, so secrets they store there can be decrypted.HIVE_AUTH_DISABLED=1.apps/hive/server/apps/hive/src/packages/shared/macOS and Windows (native). Linux works for development.
MIT