🦜 langchain-tzafon

An integration package connecting Tzafon and LangChain.

langchain-tzafon provides two powerful integrations:

• ChatTzafon: A LangChain chat model for Tzafon's AI models (chat completions with streaming support)
• TzafonLoader: A Document Loader using Tzafon's headless browser infrastructure

---

✨ Features

• Chat Completions: Access Tzafon's AI models via LangChain's chat model interface
• Streaming Support: Real-time token streaming for chat responses
• Headless Browser Rendering: Powered by Tzafon's cloud-based browser instances
• JavaScript Support: Naturally handles SPAs and dynamically loaded content
• Sync & Async Support: Features both synchronous and asynchronous APIs
• Seamless Integration: Fully compatible with LangChain's interfaces

---

🚀 Installation

pip install langchain-tzafon

Note: This package requires Playwright for connecting to the remote browser.

---

🔑 Configuration

To use this package, you need a Tzafon API Key.

1. Sign up or log in at tzafon.ai to get your API key.
2. Set it as an environment variable (recommended):

export TZAFON_API_KEY="your_api_key_here"

Alternatively, you can pass the API key directly when initializing the loader.

---

📖 Usage

ChatTzafon - Chat Completions

Use Tzafon's AI models for chat completions:

from langchain_tzafon import ChatTzafon

# Initialize the chat model
chat = ChatTzafon(model="tzafon.sm-1")

# Simple invocation
response = chat.invoke("Hello, how are you?")
print(response.content)

ChatTzafon - Streaming

Stream responses token by token:

from langchain_tzafon import ChatTzafon

chat = ChatTzafon(model="tzafon.sm-1", temperature=0.8)

for chunk in chat.stream("Write a haiku about coding"):
    print(chunk.content, end="", flush=True)

ChatTzafon - With Messages

Use structured messages for conversations:

from langchain_tzafon import ChatTzafon
from langchain_core.messages import HumanMessage, SystemMessage

chat = ChatTzafon()
messages = [
    SystemMessage(content="You are a helpful assistant."),
    HumanMessage(content="What is the capital of France?"),
]
response = chat.invoke(messages)
print(response.content)

TzafonLoader - Text Extraction

Load web pages using Tzafon's headless browser:

from langchain_tzafon import TzafonLoader

loader = TzafonLoader(urls=["https://example.com"])
documents = loader.load()

for doc in documents:
    print(f"Content from {doc.metadata['url']}:")
    print(doc.page_content[:200])

TzafonLoader - Async Loading

For better performance when handling multiple URLs:

import asyncio
from langchain_tzafon import TzafonLoader

async def main():
    loader = TzafonLoader(urls=[
        "https://example.com",
        "https://tzafon.ai"
    ])
    
    async for doc in loader.alazy_load():
        print(f"Loaded {doc.metadata['url']}")

if __name__ == "__main__":
    asyncio.run(main())

---

🛠️ API Reference

ChatTzafon

Argument	Type	Description
model	str	Tzafon model ID. Defaults to "tzafon.sm-1". Options: tzafon.sm-1, tzafon.northstar.cua.sft.
temperature	float	Sampling temperature (0-1). Defaults to 0.7.
max_tokens	Optional[int]	Maximum tokens to generate.
stop	Optional[List[str]]	Stop sequences.
api_key	Optional[str]	Tzafon API key. Defaults to TZAFON_API_KEY env var.

TzafonLoader

Argument	Type	Description
urls	str | List[str]	A single URL or a list of URLs to load.
api_key	Optional[str]	Your Tzafon API key. Defaults to TZAFON_API_KEY env var.
text_content	bool	If True (default), extracts visible text. If False, returns raw HTML.

---

📄 License

This project is licensed under the MIT License.