langchain-tensorlake 0.1.5

A LangChain tool for parsing documents using Tensorlake's DocumentAI.

langchain-tensorlake

langchain-tensorlake provides seamless integration between Tensorlake and LangChain, enabling you to build sophisticated document processing agents with structured extraction workflows.

This repository contains 1 package with Tensorlake integrations with LangChain:

• langchain-tensorlake

---

Installation

pip install -U langchain-tensorlake

---

Quick Start

1. Set up your environment

You should configure credentials for Tensorlake and OpenAI by setting the following environment variables:

export TENSORLAKE_API_KEY="your-tensorlake-api-key"
export OPENAI_API_KEY = "your-openai-api-key"

Get your Tensorlake API key from the Tensorlake Cloud Console. New users get 100 free credits!

2. Do the necessary imports

from langchain_tensorlake import DocumentParserOptions, document_markdown_tool
from langgraph.prebuilt import create_react_agent
import asyncio
import os

3. Build a Signature Detection Agent

async def main(question):
    # Create the agent with the Tensorlake tool
    agent = create_react_agent(
            model="openai:gpt-4o-mini",
            tools=[document_markdown_tool],
            prompt=(
                """
                I have a document that needs to be parsed. \n\nPlease parse this document and answer the question about it.
                """
            ),
            name="real-estate-agent",
        )
    
    # Run the agent
    result = await agent.ainvoke({"messages": [{"role": "user", "content": question}]})

    # Print the result
    print(result["messages"][-1].content)

4. Example Usage

# Define the path to the document to be parsed
path = "path/to/your/document.pdf"

# Define the question to be asked and create the agent
question = f"What contextual information can you extract about the signatures in my document found at {path}?"

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

---

Customization

You can configure how documents are parsed using DocumentParserOptions, such as:

• chunking_strategy: fragment, page, or section
• detect_tables: enable or disable table extraction
• detect_signatures: flag signature pages
• extract_structured: define a schema for structured output