DataFog Instructor SDK

DataFog Instructor is a Python SDK for named entity recognition (NER) using Ollama as the LLM backend. It provides an easy-to-use interface for detecting and classifying entities in text.

Quick Demo

Check out this 20-second demo of DataFog Instructor in action:

This code for this streamlit demo can be found at datafog-ollama-demo

Installation

It is still pre-release since I haven't added testing coverage, so you just need To install the DataFog Instructor SDK, you can use pip:

pip install --pre datafog-instructor

Quick Start

Here's a simple example to get you started with DataFog Instructor:

from datafog_instructor import DataFog

# Initialize DataFog with default settings
datafog = DataFog()

# Detect entities in text
text = "Cisco acquires Hess for $20 billion"
result = datafog.detect_entities(text)

# Print results
for entity in result.entities:
    print(f"Text: {entity.text}, Type: {entity.type.value}")

Configuration

You can customize the DataFog instance using environment variables:

• DATAFOG_LLM_BACKEND: Currently only supports "ollama"
• DATAFOG_LLM_ENDPOINT: The host URL for the Ollama service (default: "http://localhost:11434")
• DATAFOG_LLM_MODEL: The model to use for entity detection (default: "phi3")

Example with custom settings:

import os
os.environ['DATAFOG_LLM_ENDPOINT'] = 'http://custom-ollama-host:11434'
os.environ['DATAFOG_LLM_MODEL'] = 'custom-model'

from datafog_instructor import DataFog

datafog = DataFog()

Features

Detect Entities

Use the detect_entities method to identify and classify named entities in a given text:

text = "Apple Inc. reported $100 billion in revenue for Q4 2023"
result = datafog.detect_entities(text)

for entity in result.entities:
    print(f"Text: {entity.text}, Type: {entity.type.value}")

Manage Entity Types

You can add or remove entity types dynamically:

# Add a new entity type
datafog.add_entity_type("CUSTOM", "Custom Entity")

# Remove an entity type
datafog.remove_entity_type("CUSTOM")

# Get all entity types
entity_types = datafog.get_entity_types()
print(entity_types)

Default Entity Types

The SDK comes with an expanded list of predefined entity types, including:

• Organization Information: ORG, PERSON, TRANSACTION_TYPE, DEAL_STRUCTURE, FINANCIAL_INFO, PRODUCT, LOCATION, DATE, INDUSTRY, ROLE, REGULATORY, SENSITIVE_INFO, CONTACT, ID, STRATEGY, COMPANY, MONEY
• Personal Information: EMAIL, PHONE, SSN, CREDIT_CARD, IP_ADDRESS, URL, AGE, NATIONALITY, JOB_TITLE, EDUCATION
• Location Information: ADDRESS, CITY, STATE, ZIP, COUNTRY, REGION

Error Handling

The SDK includes error handling for various scenarios. If there's an issue with processing the response or an unexpected response format, it will raise a ValueError with details about the error.

Development and Testing

For development purposes, you can install additional dependencies:

pip install datafog-instructor[dev]

This includes tools like pytest, black, flake8, and mypy for testing and code quality.

Documentation

To build the documentation locally:

pip install datafog-instructor[docs]
cd docs
make html

The documentation will be available in the docs/_build/html directory.

Contributing

Contributions to the DataFog Instructor SDK are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License.

Support

If you encounter any problems or have any questions, please open an issue on the GitHub repository or join our Discord community at https://discord.gg/bzDth394R4.

Links

• Homepage: https://datafog.ai
• Documentation: https://docs.datafog.ai
• Twitter: https://twitter.com/datafoginc
• GitHub: https://github.com/datafog/datafog-instructor