npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@evandu/super-pi-server

v0.1.281

Published

Super Pi remote execution service, without the desktop runtime

Readme

Super Pi

Super Pi is a cross-platform web control plane and companion for Pi (@earendil-works/pi).

It provides a rich Web UI featuring multi-workspace management, interactive agent chats, file browser, subagent delegation, real-time tool inspection, and extensible themes.


Installation & Usage

Shell completion

Run super_pi completion install once, then open a new terminal. Bash, Zsh, Fish and PowerShell are supported; select explicitly with super_pi completion install zsh. Use super_pi completion bash (or another shell name) to print the registration script instead of modifying a profile. Completion covers commands, subcommands and options. Both super_pi and super-pi are registered.

The workspace directory picker also filters child directories as you type part of a path. Click a result to enter it; Tab completes a single remaining candidate. A partial path is never silently submitted as its parent directory.

Desktop and multiple servers

The full package includes the desktop app, local SSH manager, and the complete browser UI:

npm install -g @evandu/super-pi
super_pi

The full package opens the desktop by default, including after super_pi update. Use super_pi --browser to explicitly open the browser, or super_pi start --no-browser to run only the service. The first launch installs Super Pi.app in ~/Applications on macOS, or creates Desktop and Start Menu shortcuts on Windows. Linux gets an application-menu entry. The desktop remembers its sidebar filter, selected conversation, open tabs and unfinished drafts when quit and reopened. Browser tabs retain independent working sets.

The first desktop launch prepares the Electron runtime automatically, with progress, retries, an alternate mirror, and official checksum verification. It caches the runtime in the user's data directory across npm upgrades; users do not need to configure a mirror. Closing the window keeps the app in the system tray. Settings → Desktop and background provides optional startup at sign-in. Browser access remains available; both interfaces use the same local service.

In Settings → Server connections → SSH managed, name a server and enter its SSH address, port, user, authentication, and desired Super Pi service port. Confirm the host fingerprint, then deploy. The manager installs the server-only package in that user's home directory, installs a private Node.js runtime if needed, and connects through SSH. No extra public service port is required. Automatic remote installation currently targets Linux/macOS on x64/arm64.

Configure providers and OAuth logins once in the local entry. Managed servers use a model-only channel through SSH; upstream API keys and OAuth refresh tokens stay local. The manager must remain online for these model calls. SSH connection, deployment, service health, and model-channel status are visible separately. Upgrade managed servers by redeploying from this panel.

For independently administered servers, the direct-connection tab supports an HTTP(S) address and an owner-issued connection key. Direct connections are stored in that browser; SSH profiles are stored by the local manager. The sidebar groups resources by server name and provides a server filter alongside workspace, repository, and activity filters. Selecting a settings target does not change the active conversation or its file/Git/terminal target.

Remote machines can also install the service-only package manually:

npm install -g @evandu/super-pi-server
super-pi-server start --no-browser

This package includes the web interface and execution features, without Electron or the SSH manager. Existing super_pi start, browser access, and Pi-extension usage remain available.

Method 1: Install as a Pi Plugin (Recommended)

Super Pi is designed to work seamlessly as an official/community extension for Pi:

pi install npm:@evandu/super-pi

Once installed, simply launch your Pi terminal as usual, and type:

/super_pi

This will automatically start the background Super Pi server (port 3080) and open the Web UI in your default browser.

commands available inside pi:

  • /super_pi — Open or launch Super Pi
  • /super_pi stop — Stop the background Super Pi server
  • /super_pi restart — Restart the Super Pi server

To update:

pi update @evandu/super-pi
# or update all Pi extensions:
pi update --extensions

Method 2: Run directly via npx / Node

You can also run Super Pi standalone without Pi:

npx @evandu/super-pi start

Or install globally:

npm install -g @evandu/super-pi
super-pi start   # Start server and open browser
super_pi start   # The underscore spelling is also supported
super-pi stop    # Stop the background server
super-pi restart # Restart the server

If the panel suddenly stops working and the browser only says "connecting", the server may have died. Its log says why:

super-pi logs          # The last 200 lines of the server log
super-pi logs 50       # The last 50 lines
super-pi logs --follow # Keep printing new lines (Ctrl+C to stop)
super-pi logs --path   # Where the log file is
super-pi logs --clear  # Empty the log file

The log lives in <data directory>/logs/server.log (super-pi logs --path prints the exact path; SUPER_PI_DATA_DIR moves the whole data directory). A separate logs/heartbeat.txt is touched every minute while the server lives, so super-pi logs can also tell you when the server was last alive - which is how a crash that left no message at all still gets a time on it.

To start Super Pi automatically when the current user signs in, run:

super_pi autostart enable
super_pi autostart status
super_pi autostart disable

This registers a per-user startup entry on Windows, macOS, or Linux and does not require administrator access. The automatic process starts the local server without opening a browser. Use --port and --host with autostart enable when the server uses a non-default address.

Then visit http://127.0.0.1:3080 in your browser (or use the printed Network URL http://<lan-ip>:3080 for LAN access).

Super Pi listens on 0.0.0.0 by default. You can customize the host and port:

super-pi start --host 0.0.0.0 --port 3080
# Restrict to local loopback only:
super-pi start --host 127.0.0.1

Active memory

Super Pi includes Hermes-style active memory. It saves only durable information, and keeps the three kinds of information separate:

  • User profile: long-term preferences shared across workspaces.
  • Workspace facts: repository conventions and environment facts, visible only in that workspace.
  • Skills: validated reusable procedures and pitfalls, stored as Markdown and loaded on demand.

Open Settings → 主动记忆 to inspect, edit, delete, or approve pending changes; the same panel controls the automatic review interval. Primary Agents can use memory, skill_manage, and session_search. Sub Agents can read relevant material but cannot write it.

Composer references

Messages may include @worktree:<name> to attach the matching Git worktree path from the current workspace, or @session:<id-or-title> to attach the recent transcript of another session in the same workspace. These references are expanded into explicit context before the turn is sent to Pi, so an agent repairing a harness can inspect the session where it behaved poorly without switching away from the current session.

After the configured number of completed user turns (10 by default), Super Pi starts a separate, restricted review run. It can only use the three memory tools: it cannot read or modify project files, execute shell commands, or dispatch agents. New facts may be saved automatically; changing or deleting existing facts is staged for approval. A new message or stop cancels that conversation's review.

The feature is on by default. Set SUPER_PI_MEMORY_ENABLED=0 to disable it, or set SUPER_PI_MEMORY_REVIEW_INTERVAL=0 to keep manual memory while disabling automatic review. See the design for storage limits, lifecycle, and security boundaries.


Feishu cloud documents

The primary agent can read and write Feishu (飞书) cloud documents: feishu_doc_read returns a document as Markdown, and feishu_doc_write appends to a document, creates one, or replaces one.

It goes through Feishu's own CLI (lark-cli, the official @larksuite/cli package), so Super Pi does not reimplement the Feishu API and does not hold any Feishu credential: the CLI keeps the application and the user token in the operating system's keychain, and Super Pi only runs commands.

Setup is two browser confirmations, with no developer-console work:

npm i -g @larksuite/cli

Then open Settings → 飞书 and press 连接飞书账号. The CLI registers an application through a browser confirmation (config init --new) and then authorizes your account (auth login); once that is done the agent acts as you, so your own documents need no per-document collaborator grant.

Overwriting a document is the only destructive mode: the agent has to read it first and present the version it read, and Feishu refuses the write when the document has changed since. See the design for the command surface, the connection flow, and what is deliberately not covered.


Remote latency diagnosis

For a remote browser, enable phase tracing on the server for one test turn:

SUPER_PI_TRACE_LATENCY=1 super-pi start

The server writes answer.begin, pi.spawned, prompt.sent, pi.first-record, and pi.settled with elapsed milliseconds. This separates process/configuration startup from model time. Streaming is coalesced to one WebSocket update per 16 ms and rendered once per browser frame; structural events such as tool calls and state snapshots flush pending text first.

Use the output to compare a direct SSH Pi run with Super Pi. A long answer.begin ? pi.spawned indicates process startup or local configuration work; a long pi.spawned ? prompt.sent identifies pre-prompt validation/configuration; a long prompt.sent ? pi.first-record belongs to Pi/model/provider. If these phases are fast but the browser remains slow, inspect the network path between the browser and the server.

Development

# Install dependencies
npm install

# Start development servers (frontend + backend hot-reload)
npm run dev

# Build for production
npm run build

# Run check suites
npm run check

License

MIT

Web tools and terminal

Add a sidebar tool

Tell the Super Pi agent what sidebar tool you want. The built-in development skills guide it to use tool_register and register the complete files globally, so the tool is available from every workspace without a separate server. The setup dialog retains 复制备用提示词 for environments that cannot load built-in skills.

For a tool that needs local computation, include a Python script in the registration and declare its parameter schema. The script receives one JSON object on stdin and must print one JSON object on stdout. Load the built-in browser SDK from the mounted page:

<script src="./super-pi-tool.js"></script>
<script>
  async function calculate(params) {
    const run = await SuperPiTool.run(params);
    return await SuperPiTool.wait(run.id);
  }
</script>

SuperPiTool.run, get, wait, and cancel are relative, tool-scoped APIs. A page should not call /api/exec, guess a tool id, start an HTTP server, or send shell commands. The server enforces the declared schema, timeout, cancellation, and Python process isolation.

Legacy URL bookmarks remain available for opening external HTTP/HTTPS pages, but they are not the mechanism for registering or running a Super Pi tool.

Enable the terminal

The terminal is built in and starts with Super Pi. It is restricted to local and private-network connections.

PowerShell:

npm start

macOS/Linux:

npm start

Select a workspace, then open the Terminal tab. The shell starts in that workspace. Windows uses PowerShell by default; set SUPER_PI_TERMINAL_SHELL to choose another shell. Terminal connections are limited to localhost and private-network addresses, and the browser cannot choose an arbitrary working directory.

See docs/TERMINAL.md for security details and troubleshooting.