zijus-tools 0.0.1

Framework-agnostic UI tools for LLM Agents to send rich WebSocket messages to Zijus Chat UI.

🛠️ Zijus Tools

Zijus Tools is a Python library containing pre-built, framework-agnostic tools for LLM Agents. It enables agents to send rich, interactive UI components directly to your frontend over WebSockets.

Designed to work seamlessly with any Agentic Framework—including Agno, AutoGen, Microsoft Agent Framework, Google Agent Development Kit (ADK), LangChain, and more—alongside the Zijus Chat UI. This library abstracts away the complexity of WebSocket context management, allowing your LLM agents to render dynamic interfaces with a single function call.

---

🔗 Important Links

To get the most out of zijus-tools, you will need the frontend client:

• 🖥️ Zijus Chat UI GitHub Repo - See examples of how to integrate the Zijus Chat Client with various Agentic frameworks.
• 🎨 Zijus Chat UI Customization - Learn how to style, theme, and customize the frontend client.

---

📦 Installation

pip install zijus-tools

(Note: Requires Python 3.8+)

---

🚀 Quick Start

Using zijus-tools is a simple two-step process:

1. Register your WebSocket sender in your server app (FastAPI, Django, etc.).
2. Provide the tools to your chosen LLM Agent framework.

Step 1: Configure the Server

Because zijus-tools is web-framework-agnostic, you just need to tell it how to send JSON messages. We use Python's built-in contextvars to ensure thread-safe message delivery for concurrent users.

# main.py (Example using FastAPI)
from fastapi import FastAPI, WebSocket
from zijus_tools import set_websocket_sender # 1. Import the setup function

app = FastAPI()

@app.websocket("/ws")
async def websocket_endpoint(websocket: WebSocket):
    await websocket.accept()

    # 2. Define a simple async wrapper that sends a dict over the websocket
    async def sender(msg: dict):
        await websocket.send_json(msg)

    # 3. Inject the sender for this specific user's connection
    set_websocket_sender(sender)

    # ... initialize and run your Agent (Agno, AutoGen, ADK, etc.) ...
    while True:
        data = await websocket.receive_text()
        # Handle incoming data...

Step 2: Add Tools to your Agent

Because Zijus Tools are standard Python asynchronous functions, you can pass them as tools to any agent framework. The tool automatically handles formatting the UI payload and emitting the message behind the scenes!

Example: Google ADK

from google.adk.agents import LlmAgent
from zijus_tools import SendSlots 

agent = LlmAgent(
    name="root_agent",
    tools=[SendSlots],  # Just drop it in the tools list
    model="gemini-2.5-flash",
    instruction="Whenever the user asks for options, use the SendSlots tool."
)

---

🧰 Available Tools

We are constantly expanding the library of UI components.

SendSlots

Sends a list of selectable slots (buttons or checkboxes) to the chat interface. Perfect for multiple-choice questions, filtering, or guiding the user.

Arguments:

• slots (Any): A list, comma-separated string, or JSON string of options. (e.g., ["Mining", "Cement"] or [{"label": "Yes", "value": "1"}])
• content (str, optional): Text to display above the slots.
• checkbox (bool, optional): If True, renders as multi-select checkboxes. Defaults to False (single-select buttons).
• delay (float, optional): Delay rendering to match natural chat cadence.

---

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request if you have a new UI tool you'd like to add. Ensure that your tools rely on emit_message() from zijus_tools.context to remain framework-agnostic.

📄 License

This project is open-sourced under the MIT License.