sonsery-flow-community

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

AI workflow automation — bridge web AI (ChatGPT, Claude, Gemini, DeepSeek) with your terminal. No API keys. MIT, open source.

README.md

Sonsery Flow Community - AI Workflow Automation

CI
License: MIT
Python 3.8+
Manifest V3

Sonsery Flow — Turn free web AI into your CLI agent

Sonsery Flow — use free web AI (ChatGPT, Claude, Gemini, DeepSeek, and more)
as a coding agent, right inside your terminal. No API keys, no subscriptions,
no token quotas. Copy a payload in the AI chat → the tool runs it on your
machine → the result is pasted back. Automation loop, end to end.

A tool that automates the workflow between AI Chat (ChatGPT, Claude, Gemini, DeepSeek) and the terminal. Community edition — MIT, open source.


How it works

  1. Copy the payload from AI Chat to the clipboard
  2. watchctx automatically detects it and executes the command
  3. The result is copied back to the clipboard to paste into AI Chat

Setup — step by step

Follow these steps in order. Each one ends with something you can verify
before moving on. Takes about 5 minutes.

Step 1 — Clone the repo

git clone https://github.com/jsonnguyen19/sonsery-flow-community.git
cd sonsery-flow

Step 2 — Run setup

pnpm setup

No pnpm? Install it once: npm install -g pnpm (or see
pnpm.io).

Don't want pnpm? Run the script directly: python3 scripts/setup.py.

This creates the venv/ folder, installs Python runtime deps, and writes the
wrapper scripts run and sync at the repo root.

Step 3 — Test it runs

./run

You should see the clipboard watcher start and the HTTP bridge announce a port
in the 8765–8785 range. Leave this terminal running.

On Windows (cmd/PowerShell) use run.bat instead:

.\run.bat

Open a second terminal for the next steps.

Step 4 — Configure the alias (WSL first)

Setup does NOT write aliases for you — add them yourself so you can type
watchctx from any directory.

WSL / Linux / macOS (bash / zsh) — add to ~/.bashrc or ~/.zshrc,
replacing /path/to/sonsery-flow with your actual clone path:

alias watchctx='/path/to/sonsery-flow/run'
alias watchctx-sync='/path/to/sonsery-flow/sync'
Other shells (Windows PowerShell, cmd, Git Bash) — click to expand

Windows PowerShell — add to $PROFILE:

function watchctx      { & "C:\path\to\sonsery-flow\run.bat" @args }
function watchctx-sync { & "C:\path\to\sonsery-flow\sync.bat" @args }

Git Bash — same as WSL, add to ~/.bashrc:

alias watchctx='/c/path/to/sonsery-flow/run'
alias watchctx-sync='/c/path/to/sonsery-flow/sync'

cmd.exe — no native alias. Use doskey in a startup script, or just call
C:\path\to\sonsery-flow\run.bat directly.

Step 5 — Reload your terminal

For the new aliases to take effect, either open a new terminal or source
the config file:

source ~/.zshrc      # zsh
. ~/.bashrc          # bash

Step 6 — Verify the terminal

Still in the second terminal, run:

watchctx

You should see the same output as Step 3 (watcher + bridge). If it runs,
your terminal setup is done — Ctrl+C to stop, or leave it as your main
watcher terminal.

If watchctx: command not found, the alias didn't load. Re-check Step 4
path and re-run source from Step 5.

Step 7 — Load the Chrome extension

The extension lives in runctx-extension/. To get its exact path (and the
Windows UNC form for WSL), run:

pnpm about

Copy the Extension path it prints. Then:

  1. Open chrome://extensions/ (or edge://extensions/).
  2. Enable Developer mode (top-right toggle).
  3. Click Load unpacked → paste the path
    (on WSL: paste the Windows UNC path — \\wsl.localhost\<Distro>\home\...\runctx-extension — into the Explorer address bar and press Enter first).
  4. Pin the extension (puzzle-piece icon → pin).
  5. Reload the extension once (the refresh icon on its card).

Step 8 — Test with DeepSeek

  1. Open chat.deepseek.com.
  2. Click the pinned Sonsery Flow icon → the side panel opens.
  3. Flip the Automation toggle ON.
  4. Copy this ready-made payload and send it in the chat:
{ "id": 1791016717000, "tool": "shell", "mode": "sequential", "commands": ["pwd"] }

Bump the id to any newer millisecond timestamp (e.g. Date.now()) if the
extension ever reports the payload as already processed.

  1. Watch: the payload is copied → watchctx runs it → the result (your current
    working directory) is pasted back into the chat.

If you see the result come back, setup is complete. 🎉


Bonus — a dedicated background Chrome for coding

Recommended. By default Chrome throttles timers, freezes tabs, and sleeps
background windows — this can pause the automation loop when the AI tab loses
focus. Launch Chrome with a separate profile and the flags below so the loop
keeps running even when you minimize or switch away.

Flag Purpose
--user-data-dir="..." Separate profile, isolated from your daily Chrome
--disable-background-timer-throttling Don't slow down timers in background tabs
--disable-backgrounding-occluded-windows Don't deprioritize covered/occluded windows
--disable-renderer-backgrounding Don't lower renderer priority in the background
--intensive-wake-up-throttling-policy=disabled Disable aggressive wake-up throttling

Windows (PowerShell) — creates a Desktop shortcut:

$WshShell = New-Object -ComObject WScript.Shell
$ShortcutPath = "$([Environment]::GetFolderPath('Desktop'))\Chrome Agent.lnk"
$Shortcut = $WshShell.CreateShortcut($ShortcutPath)

$Shortcut.TargetPath = "C:\Program Files\Google\Chrome\Application\chrome.exe"
$Shortcut.Arguments = '--user-data-dir="%LOCALAPPDATA%\Google\Chrome\User Data Agent" --disable-background-timer-throttling --disable-backgrounding-occluded-windows --disable-renderer-backgrounding --intensive-wake-up-throttling-policy=disabled'

$Shortcut.Save()

macOS — launch from Terminal:

open -na "Google Chrome" --args \
  --user-data-dir="$HOME/Library/Application Support/Google/Chrome Agent" \
  --disable-background-timer-throttling \
  --disable-backgrounding-occluded-windows \
  --disable-renderer-backgrounding \
  --intensive-wake-up-throttling-policy=disabled

Linux — save as ~/.local/share/applications/chrome-agent.desktop:

[Desktop Entry]
Name=Chrome Agent
Exec=/usr/bin/google-chrome --user-data-dir=%h/.config/google-chrome-agent --disable-background-timer-throttling --disable-backgrounding-occluded-windows --disable-renderer-backgrounding --intensive-wake-up-throttling-policy=disabled
Type=Application
Terminal=false

Note: For Edge, replace the executable path (chrome.exe → msedge.exe,
Google Chrome → Microsoft Edge) — flag names are identical.

Tip: Install the extension into this dedicated profile once. Keep the
window minimized in the background — the automation loop keeps running.


Commands

Command What it does
pnpm setup End-user setup: venv/ + runtime deps + run/sync wrappers
./run Start watchctx (clipboard watcher + HTTP bridge)
./sync Sync prompts into the extension
pnpm about Print extension path, bridge ports, available commands
pnpm setup:dev Contributor setup (.venv-test/ + dev deps)

Requirements

  • Python 3.8+
  • Chrome 100+ / Edge 100+ (Manifest V3)

No GPU, no Docker, no admin rights needed.


Documentation

Full docs live in docs/community/.

Doc What it covers
README.md This file — overview + setup
CONTRIBUTING.md Contribution guide
CHANGELOG.md Release notes
SECURITY.md Security policy
CODE_OF_CONDUCT.md Community code of conduct
scripts/README.md Every script in scripts/
docs/community/getting-started.md Install the backend + extension, run the first automation loop, day-to-day usage
docs/community/tools.md Payload & response format for shell / read / replace / write
docs/community/tabs.md The 4 side-panel tabs and what each one does
docs/community/adapters.md The 5 Community AI adapters + how to debug/add one
docs/community/architecture.md How the content script, watcher, and bridge fit together

Links


Contributing

Contributor / dev setup (dev venv, lint, format, typecheck, tests) lives in
CONTRIBUTING.md. End-users never need it.

Repo layout:

  • watchctx.py — entrypoint: clipboard watcher + HTTP bridge
  • runctx_core.py — entrypoint: payload processing
  • runctx/ — main package
  • runctx-extension/ — Chrome/Edge extension
  • scripts/ — setup scripts (see scripts/README.md)
  • prompts/ — prompt templates
  • tests/python/ — pytest test suite

See CONTRIBUTING.md for the full guide.

Yorumlar (0)

Sonuc bulunamadi