marly

Data Processor for AI Agents. Search your documents or the web for specific data and get it back in JSON or Markdown in a single tool call.

marly-ai

162

Python

open-extract

Features • What is a Schema? • Use Cases • Getting Started • Documentation

open-extract simplifies the ingestion and processing of unstructured data for those building AI Agents/Agentic Workflows using frameworks such as LangGraph, AG2, and CrewAI.

🚀 Features

📄 Extract Relevant Information Seamlessly: Give your applications the ability to identify and extract relevant data from one or many large documents and websites with just a single API call. Get the content back in JSON or Markdown formats, making it easy to integrate into your workflows.

🔍 Multi-Schema/Multi-Document Support: Extract data based one or many predefined schemas from a variety of document types, without needing a vector database or specifying page numbers.

🔄 Built-in Caching: With built-in caching, previously extracted schemas can be instantly retrieved, enabling rapid repeat extractions without having to reprocess the original documents.

🚫 No Vendor Lock-In: Enjoy complete flexibility with your choice of model provider. Whether using open-source or closed-source models, you’re never tied to a specific vendor, ensuring full control.

🧰 What is a Schema?

A schema is a set of key-value pairs describing what needs to be extracted from a particular document.

📋 Example Schema

{
    "Firm": "The name of the firm",
    "Number of Funds": "The number of funds managed by the firm",
    "Commitment": "The commitment amount in millions of dollars",
    "% of Total Comm": "The percentage of total commitment",
    "Exposure (FMV + Unfunded)": "The exposure including fair market value and unfunded commitments in millions of dollars",
    "% of Total Exposure": "The percentage of total exposure",
    "TVPI": "Total Value to Paid-In multiple",
    "Net IRR": "Net Internal Rate of Return as a percentage"
}

🎯 Use Cases

💼 Financial Report Analysis	📊 Customer Feedback Processing	🔬 Research Assistant	🧠 Legal Contract Parsing
Extract key financial metrics from quarterly PDF reports	Categorize feedback from various document types	Process research papers, extracting methodologies and findings	Extract key legal terms and conditions from contracts

🛠️ Getting Started

Build the Platform

To build the platform from source, run the following command:

./start-oe.sh

Run an example script or notebook

Once the platform is running you can test it out by trying one of our examples

Navigate to the examples folder:
```
cd examples
```
Navigate to the scripts or notebooks folder:
```
cd scripts
```
or
```
cd notebooks/autogen_example
```
Run one of our example scripts:
```
python azure_example.py
```