Synt-E: The Protocol for Talking to AIs 🚀

Synt-E is a protocol and a command-line tool for translating natural language into efficient machine commands, designed for interaction with local LLMs (via Ollama). Instead of writing long sentenc[...]

This project is not just an experiment, but a Power Tool for developers and power users who want to optimize their AI workflow, and a prototype of a **Machine-to-Machine (M2M) communication pro[...]

---

🤔 Why Does Synt-E Exist? The Problem

When we talk to an AI, we use colloquial language. This is fine for a chat, but in a professional workflow or an automated system, it is:

• Slow: The AI has to process unnecessary words.
• Expensive: More words = more tokens to compute (or pay for).
• Ambiguous: Human language can be misunderstood.

BEFORE (Natural Language):

"Hi, could you write me a Python script to analyze data from a CSV file?"

AFTER (Synt-E, activated with a shortcut):

task:code lang:python action:analyze_data format:csv

---

🛠️ Why is it Useful in Any Software Project?

In any development environment that integrates LLMs—whether it's a startup, a DevOps team, or an open-source project—efficiency is key. Synt-E offers concrete advantages:

1. 💰 Cost Reduction (Token Saving): A Synt-E prompt can be up to 70% shorter than a normal sentence, resulting in direct savings on costs and computational resources.
2. ⚡ Increased Speed (Latency Reduction): Fewer tokens to process means faster responses, which is essential for real-time applications.
3. 🤖 Reliability and Testability (Fewer Bugs): Synt-E is a standardized protocol. It makes interactions with AIs predictable and easy to test, reducing bugs.

The Fundamental Discovery: The Right Model is Everything

During testing, we discovered that super-trained "assistant" models (like Llama 3 Instruct) are the worst for this task because their instinct to "execute" the command wins over the meta-ins[...]

The best models are those that are more "raw" or "unfiltered," which are more obedient to a strict SYSTEM_PROMPT. Our winner was gpt-oss:20b.

System Limits: When NOT to use Synt-E

Synt-E is a protocol for compressing complexity, not always length.

• Great for: Long, descriptive, and complex sentences. Here, the savings are enormous.
• Useless for: Very short sentences (1-3 words). In this case, the AI has to "invent" the context (task:, topic:), and the output may be longer than the input.

---

💻 Usage Guide: Your Personal Compiler

This tool runs in the background, listening for a series of keyboard shortcuts to give you complete and safe control over the synthesis process.

Installation

1. Install Ollama: Download it from ollama.com and run it.
2. Download a Model: Open the terminal and download the model you want to use (e.g., gpt-oss or qwen3:30b).

   ollama pull gpt-oss:20b

3. Install Python Libraries:

   pip install ollama keyboard pyperclip pywin32 psutil winsound

Startup and Command-Line Options

Start the script from the terminal. You can use different "flags" to customize its behavior.

Basic Startup:

python synt_e.py
(Uses default shortcuts and model)

Startup with a Custom Model:

python synt_e.py --model="qwen3:30b-a3b-unlocked"

Startup with a Custom Shortcut:

python synt_e.py --hotkey="ctrl+shift+x"

Startup in "Append" Mode (does not replace, but adds at the end):

python synt_e.py --append

You can combine as many options as you like!

Daily Workflow (Safe Method)

The workflow has been designed to be 100% reliable, giving you full control.

1. Start the script in a terminal and leave it open in the background.
2. Go to any program (Chrome, VS Code, Notepad...).
3. Select the text you want to synthesize.
4. Copy Manually: Press Ctrl+C.
5. Activate Synthesis: Press the shortcut (default: Ctrl+Alt+S). Your text will be instantly replaced with the Synt-E version!

All Shortcuts (Power User Features)

This tool is more than just a synthesizer. It's a command suite:

• Synthesize (Ctrl+Alt+S): Compiles the text you have copied.
• Undo (Ctrl+Alt+U): You have 10 seconds after a synthesis to press this hotkey and restore the original text. A lifesaver!
• Interrupt AI (Ctrl+Alt+C): If the AI is taking too long for a complex request, press this shortcut to cancel the operation.
• Keyboard Emergency (Ctrl+Alt+Q): In the very rare case that the script locks your keyboard, this is your forced "emergency exit" that resets everything.

Audio Feedback

• Double Beep: Confirms that an operation (synthesis, undo) was successful.
• Single Low Beep: Alerts you to an error (e.g., no text in the clipboard).
• Cancellation Beep: Confirms that the AI operation has been interrupted.