pulse

⚡◉ pulse — Real-time floating dashboard for local AI. Watch live tokens/sec + loaded models straight from Ollama. Zero terminals. Zero fake numbers. Built with Python + PyQt5. PRs welcome! ╰(°▽°)╯

◉ pulse

A floating, cyberpunk-style dashboard that shows your local AI in real time — no terminal, no BS, no fake numbers.

Built with love, Python, and way too much caffeine ☕ (❁´◡`❁)

   ╔══════════════════════════════════════════════════════╗
   ║   ╭──────────────────────────────────────────────╮   ║
   ║   │   ●  AI MODELS: 1                            │   ║
   ║   │                                              │   ║
   ║   │      ┌───────────────────────┐               │   ║
   ║   │      │                       │               │   ║
   ║   │      │          4.6          │  ← real       │   ║
   ║   │      │                       │    tokens/s   │   ║
   ║   │      │    TOKENS / SEC       │               │   ║
   ║   │      └───────────────────────┘               │   ║
   ║   │                                              │   ║
   ║   │        ● ● ● ●   (live pulse)                │   ║
   ║   ╰──────────────────────────────────────────────╯   ║
   ╚══════════════════════════════════════════════════════╝

Live. Real. Yours. (っ◔◡◔)っ ♥

Live tokens/sec

Built-in chat

🌌 What is this?

pulse is a floating desktop widget that sits on top of your screen and shows you, in real time:

🧠 How many AI models you have loaded right now
⚡ Exactly how fast they're generating (real tokens/sec, pulled from Ollama's own logs — the same numbers a terminal would show you)
💬 A built-in chat window so you never need to open a terminal again

It looks like it fell out of Cyberpunk 2077, it costs nothing, and it just works. ^_____^

✨ Why does this exist?

Because right now, if you run local AI (Ollama, LM Studio, llama.cpp…), you either:

Stare at a black terminal watching numbers scroll by 👀 (✋ very 2010s)
Trust some fake "AI monitor" app that just shows random numbers ✋ (✋ very dishonest)
Have no idea how fast your model is running 🤷

We said no to all three. So we built a floating, beautiful, honest dashboard that shows the actual, verified, real numbers. If the log says tg = 4.57 t/s, the circle shows 4.6. No interpretation. No estimation. No hallucination.

"The terminal is where you debug. The dashboard is where you flex." — probably someone cool, maybe us, we don't remember φ(゜▽゜*)♪

🎨 What makes it special

Feature	What it does	Vibe
🔮 Floating always-on-top widget	Sits on your desktop like a HUD from a sci-fi game	`(●'◡'●)`
📊 Real live data	Reads actual `tg = X t/s` from Ollama's log — provably identical to the terminal	🎯
💬 Built-in chat	Talk to Hermes / Llama / Mistral / whatever — no PowerShell required	✨
🎨 4 shapes	Circle, squircle, hex, diamond — swap anytime	💎
🌈 Cyberpunk palette	Cyan / teal / blue glow that actually feels futuristic	🎆
📏 Resizable	Drag from 200px → 600px, fonts scale with it	📐
⚙ Settings panel	Separate floating window — no clutter	🧼
🚫 No terminal ever	Auto-manages Ollama in the background	🙅
🖥 Cross-Windows	Works on Windows 7 / 8 / 10 / 11 — yes, even that one	👴

🧭 How it works — the whole picture

Here's the entire pipeline, drawn with mermaid so your eyes don't bleed reading text:

🔄 Data flow (how a real token/sec number reaches your circle)

flowchart LR
    A[👤 You write<br/>a message] --> B[🖥️ pulse<br/>Chat Window]
    B --> C[📡 POST to Ollama<br/>localhost:11434/api/chat]
    C --> D[🦙 Ollama generate<br/>Hermes3:8b]
    D --> E[⚙️ llama.cpp engine<br/>calculates tg = X t/s]
    E --> F[📝 ollama.log<br/>receives tg lines]
    E --> G[📤 Streams tokens<br/>back to chat window]

    F --> H[👁️ QFileSystemWatcher<br/>notices new line]
    H --> I[📖 LogReader thread<br/>parses tg = X t/s]
    I --> J[🎨 Floating Circle<br/>updates to 4.6]

    G --> B

    style A fill:#00ffc3,color:#000
    style D fill:#0088ff,color:#fff
    style E fill:#0088ff,color:#fff
    style J fill:#00ffc3,color:#000
    style I fill:#4db8ff,color:#000

🏗️ Architecture (the guts)

graph TB
    subgraph "🎨 Frontend — PyQt5"
        UI[Floating Dashboard<br/>circle / hex / squircle / diamond]
        SET[Settings Window<br/>size · shape · refresh]
        CHAT[Chat Window<br/>streams responses]
    end

    subgraph "⚙️ Background Workers — QThread"
        LR[LogReader<br/>parses tg values]
        MR[ModelsReader<br/>counts loaded models]
        CW[ChatWorker<br/>streams Ollama chat]
    end

    subgraph "🦙 Ollama Layer"
        OM[OllamaManager<br/>spawns / kills serve]
        OL[ollama serve<br/>hidden process]
        LOG[(ollama.log<br/>real tg lines)]
    end

    subgraph "🌐 HTTP API"
        API[localhost:11434<br/>/api/ps · /api/tags · /api/chat]
    end

    UI --> LR
    UI --> MR
    UI --> CW
    CHAT --> CW
    LR --> LOG
    MR --> API
    CW --> API
    OM --> OL
    OL --> LOG
    OL --> API

    style UI fill:#00ffc3,color:#000
    style CHAT fill:#00ffc3,color:#000
    style SET fill:#00ffc3,color:#000
    style LOG fill:#0088ff,color:#fff
    style OL fill:#0088ff,color:#fff

🎬 User journey (from zero to glowing circle)

journey
    title Your first 30 seconds with pulse
    section Install
      Clone repo: 5: You
      pip install PyQt5: 4: You
      Double-click shortcut: 5: You
    section Boot
      App kills old Ollama: 3: App
      App spawns hidden serve: 4: App
      Circle appears: 5: App
    section Use
      Click chat bubble: 5: You
      Type "hello Hermes": 5: You
      Watch tokens/sec dance: 5: You, Circle
    section Flex
      Show friends: 5: You
      Make a PR: 5: You

🌱 Contribution pipeline (how you join in)

flowchart LR
    A[🍴 Fork repo] --> B[🌿 Create branch<br/>feature/your-idea]
    B --> C[💻 Code your magic]
    C --> D[✅ Test it works]
    D --> E[📤 Push + open PR]
    E --> F{👀 Review}
    F -->|Approved| G[🎉 Merged!]
    F -->|Needs work| C
    G --> H[🕺 Celebrate]
    H --> I[🍕 Eat pizza]

    style A fill:#00ffc3,color:#000
    style G fill:#00ffc3,color:#000
    style I fill:#ff69b4,color:#000

🚀 Quick start (2 options ,Double-click .bat or Manual launch )

Option A — One-click launch (Windows)

Download the repo
Double-click run.bat
Enjoy ⚡

The launcher will:

Automatically find your Python installation
Install PyQt5 if you don't have it
Start the dashboard silently (no terminal window)

Option B — Manual launch

# 1. Clone
git clone https://github.com/zainosta/pulse.git
cd pulse

# 2. Install the one dependency
pip install PyQt5

# 3. Run it
python pulse.pyw

That's it. No config files. No environment variables. No terminal window stays open. ╰(*°▽°*)╯

Prereq: Ollama installed + at least one model pulled:
ollama pull hermes3:8b

🎮 How to use it

What you click	What happens
💬 Chat bubble	Opens a floating chat window — talk to any of your installed models
⚙ Gear	Opens the settings window (size, shape, refresh rate)
✕ Close	Quits the app and stops the Ollama it started (clean exit)
Drag the circle	Move it anywhere on your desktop
Bottom-right grip	Resize it freely

🎨 Shapes gallery

   ╭─────────╮       ╭─────────╮       ╱╲             ╱╲
  ╱           ╲     │           │     ╱  ╲           ╱  ╲
 │   CIRCLE    │    │  SQUIRCLE │    ╱    ╲         ╱    ╲
  ╲           ╱     │           │    ╲    ╱         ╲    ╱
   ╰─────────╯       ╰─────────╯      ╲  ╱           ╲  ╱
                                       ╲╱             ╲╱
                                     HEX            DIAMOND

🗺️ Roadmap — where we're going next

The future is bright and we need your brain to get there (ﾉ◕ヮ◕)ﾉ*:･ﾟ✧

🎯 Version 1.1 — Polish & Performance

GPU utilization graph under the tokens counter
Custom color themes (pastel, vaporwave, matrix, mono)
History sparkline (last 60 seconds of tokens/sec)
Multi-monitor support
Tray icon with quick actions

🌟 Version 1.2 — Multi-Backend

🚀 Version 2.0 — Community Dream

Plugin system — write your own panels in Python
Themes marketplace — download community themes
macOS + Linux builds
Web dashboard (same data, in browser)
Mobile companion app (yes, really)
Voice assistant integration (Alexa / Google)
Streaming overlay mode (for Twitch/YouTube)

💫 Version 3.0 — The Wild Ideas

Multi-PC mesh — see all your rigs on one dashboard
Token cost calculator — real money saved vs cloud APIs
AI training monitor (watch loss curves float by)
AR mode (overlay on your wall via webcam)

Have a wilder idea? Open an issue titled 💡 [IDEA] your wild idea — we read everything.

🤝 Contributing — WE WANT YOU (literally)

This project only grows if you build it with us. Every contribution counts — even fixing a typo in this README (❁´◡❁)`

Ways to help (pick your flavor)

🎨 Creative	🛠️ Technical	📣 Social
Design new themes	Add a backend	Star this repo
Make promo art	Fix bugs	Share on Twitter/X
Write docs	Improve performance	Stream yourself using it
Suggest shapes	Add tests	Write a blog post
Translate README	Refactor code	Tell a friend
Make demo videos	New features	Post on Reddit

Quick contribution guide

# 1. Fork on GitHub (click the button, top right)
# 2. Clone your fork
git clone https://github.com/zainosta/pulse.git

# 3. Make a branch with a cool name
git checkout -b feature/sparkly-animation

# 4. Do your thing
# 5. Test it works
# 6. Commit with a friendly message
git commit -m "✨ add sparkly animation to pulse dots"

# 7. Push and open a Pull Request
git push origin feature/sparkly-animation

Full guide in CONTRIBUTING.md ← read this first!

🎁 Contributor perks

Your name in the Hall of Fame below
Your custom role emoji next to your GitHub handle
Early access to beta features
A warm fuzzy feeling (◕‿◕)

🏆 Hall of Fame — Our legends

These people made pulse better. Want your name here? Open a PR!

Contributor	Role	Contribution
@zainosta	🧠 Creator · 🦙 Ollama whisperer	Original vision + core app
(your name here)	🎨 Designer	Waiting for you!
(your name here)	⚙️ Backend wizard	Waiting for you!
(your name here)	📖 Docs hero	Waiting for you!
(your name here)	🐛 Bug slayer	Waiting for you!
(your name here)	🎭 Meme lord	Waiting for you!

💬 Join the conversation

🐛 Found a bug? Open an issue
💡 Got an idea? Share it
🎨 Made a theme? Submit it
💌 Just want to chat? Start a discussion

We don't bite. We're friendly. (●'◡'●)

🧪 Tested on real hardware

These are not marketing numbers — these are real readings from contributors:

Machine	Model	Tokens/sec
Windows 11 · RTX 3060	hermes3:8b Q4	~45 t/s
Windows 10 · GTX 1660	hermes3:8b Q4	~22 t/s
Windows 10 · CPU only	hermes3:8b Q4	~4.5 t/s
Windows 7 · Old Xeon	tinyllama	~12 t/s

Post yours when you try it — we'll add you to the table ╰(*°▽°*)╯

📜 License

MIT — do whatever you want with it. Fork it, remix it, sell it, print it on a t-shirt. Just be cool, credit us, and share back if you make something awesome. ☆*: .｡. o(≧▽≦)o .｡.:*☆

See LICENSE for the boring legal stuff.

💖 Shoutouts

Built with:

🐍 Python — the language of the future (and the present)
🎨 PyQt5 — for making pretty floating windows
🦙 Ollama — for making local AI actually usable
☕ Coffee — the real MVP
🌙 3 AM coding sessions — where the magic happens
💜 You — for reading this far (๑˃ᴗ˂)ﻭ

⭐ If this made you smile, star the repo. It costs zero dollars and makes us very happy.

     ╱|、
    (˚ˎ 。7
     |、˜〵
     じしˍ,)ノ

Made with 💚 by humans who love local AI

Last updated: when this README stopped being readable at 4 AM ☕

❤️ Support the Project

If this script saved you time, feel free to buy me a virtual coffee! ☕
Your support motivates me to keep improving this tool.

Method	Details
💰 USDT (TRC20)	`TVmL1R1wYXiM8wP2NfJk5hwQYVMtoJhHU5`
🆔 Binance ID	`500857050`

⚠️ Important: Please make sure to select the TRC20 network when sending USDT. Sending via the wrong network may result in permanent loss of funds.

Thank you for your generosity! 🙌

Name		Name	Last commit message	Last commit date
.github		.github
assets		assets
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
SUPPORT.md		SUPPORT.md
pulse.pyw		pulse.pyw
requirements.txt		requirements.txt
run.bat		run.bat

Latest commit

History

Folders and files

Repository files navigation