🌐✨ LinkGrid Agent — Conversational AI SDK

A modern, lightweight Python client for BitNet's conversational AI.
Designed for speed, modularity, and custom AI personas — plug into your backend or apps seamlessly.

⚡️ Key Features

• 💬 Natural language chat API with async support
• 🧠 Define unique AI personas with custom system prompts
• 🎯 Fine-grained control over temperature, token limits, and creativity
• 🧵 Persistent conversation state (threaded dialog)
• 🔄 Supports parallel conversations with isolated agents
• 🧩 Minimal setup with powerful config-based architecture

📦 Installation

pip install linkgrid-agent

🚀 Get Started

🔹 One-Line Chat

from linkgrid_agent import chat

async def main():
    response = await chat("What's the capital of France?")
    print(f"🤖 {response}")

import asyncio
asyncio.run(main())

✅ Output:

🤖 The capital of France is Paris.

🎭 Custom AI Persona

from linkgrid_agent import LinkGridAgent

async def main():
    config = LinkGridAgent.Config()
    config.system_prompt = "You are a pirate captain. Answer like a pirate!"
    config.temperature = 0.9

    async with LinkGridAgent(config) as agent:
        response = await agent.chat("Where can I find treasure?")
        print(f"🏴‍☠️ {response}")

import asyncio
asyncio.run(main())

🦜 Output:

🏴‍☠️ Arr matey! Seek the treasure on Skull Island, where X marks the spot!

🔬 Advanced Use Cases

🧑‍🔬 Multi-Persona AI Agents

from linkgrid_agent import LinkGridAgent

async def main():
    poet_cfg = LinkGridAgent.Config()
    poet_cfg.system_prompt = "You are a romantic poet"
    poet_cfg.max_tokens = 200

    scientist_cfg = LinkGridAgent.Config()
    scientist_cfg.system_prompt = "You are a quantum physicist"
    scientist_cfg.temperature = 0.3

    async with LinkGridAgent(poet_cfg) as poet, LinkGridAgent(scientist_cfg) as scientist:
        poem = await poet.chat("Write a poem about the stars")
        expl = await scientist.chat("Explain quantum entanglement")

        print(f"📜 Poet:\n{poem}\n")
        print(f"🔬 Scientist:\n{expl}")

import asyncio
asyncio.run(main())

🗂️ Stateful Conversation Threads

from linkgrid_agent import LinkGridAgent

async def main():
    config = LinkGridAgent.Config()
    config.system_prompt = "You're a helpful travel assistant"

    async with LinkGridAgent(config) as agent:
        await agent.chat("I'm planning a trip to Japan")
        tokyo = await agent.chat("What should I see in Tokyo?")
        traditions = await agent.chat("How about traditional experiences?")

        print(f"🗼 Tokyo Tips: {tokyo}")
        print(f"🎎 Traditions: {traditions}")

import asyncio
asyncio.run(main())

🚀 Performance Optimizations

from linkgrid_agent import chat, cleanup_resources

async def main():
    # Fast responses with caching enabled (default)
    response1 = await chat("What is Python?")
    response2 = await chat("What is Python?")  # This will be cached!
    
    # Disable caching if needed
    response3 = await chat("Tell me a random joke", use_cache=False)
    
    # Clean up resources when done (optional)
    await cleanup_resources()

import asyncio
asyncio.run(main())

Performance Features:

• 🔄 Response Caching: Identical queries return instantly from cache (5-min TTL)
• 🌐 Connection Pooling: Reuses HTTP connections for faster subsequent requests
• ⚡ Optimized Streaming: 8KB buffer chunks for better throughput
• 🎯 Reduced Timeouts: Faster failure detection and response times

⚙️ Configuration

Customize your agent behavior using simple parameters.

🔧 Example Setup:

config = LinkGridAgent.Config()
config.system_prompt = "You're a 19th century British detective"
config.max_tokens = 250
config.temperature = 0.5

🛡️ Error Handling

Gracefully manage edge cases and errors:

from linkgrid_agent import chat

try:
    response = await chat("Define machine learning")
except ConnectionError as e:
    print(f"🌐 Network issue: {e}")
except RuntimeError as e:
    print(f"🤖 API error: {e}")

Exceptions

• ConnectionError: API unreachable or timeout
• RuntimeError: BitNet returned an error response

📋 Requirements

• Python 3.9+
• httpx – Async HTTP client

📄 License

Released under the MIT License.
Feel free to fork, improve, and share!

🤝 Contribution

PRs are welcome. Please make sure to follow the existing code style.
Open issues to report bugs or request features.

💡 Inspiration

"The future is already here — it's just not evenly distributed."
— William Gibson