Easily build knowledge-intensive LLM applications that reason over your data with LOTUS!

LOTUS (LLMs Over Tables of Unstructured and Structured Data) provides a declarative programming model and an optimized query engine for serving powerful reasoning-based query pipelines over structured and unstructured data! We provide a simple and intuitive Pandas-like API, that implements semantic operators.

LOTUS' implements is the semantic operator programming model. Semantic operators as declarative transformations on one or more datasets, parameterized by a natural language expression, that can be implemented by a variety of AI-based algorithms. Semantic operators seamlessly extend the relational model, operating over tables that may contain traditional structured data as well as unstructured fields, such as free-form text. These composable, modular language- based operators allow you to write AI-based pipelines with high-level logic, leaving the rest of the work to the query engine! Each operator can be implemented and optimized in multiple ways, opening a rich space for execution plans, similar to relational operators. To learn more about the semantic operator model, read the full research paper.

LOTUS offers a number of semantic operators in a Pandas-like API, some of which are described below. To learn more about semantic operators provided in LOTUS, check out the full documentation, run the colab tutorial, or you can also refer to these examples.

conda create -n lotus python=3.10 -y
conda activate lotus
pip install lotus-ai

If you're already familiar with Pandas, getting started will be a breeze! Below we provide a simple example program using the semantic join operator. The join, like many semantic operators, are specified by langex (natural language expressions), which the programmer uses to specify the operation. Each langex is parameterized by one or more table columns, denoted in brackets. The join's langex serves as a predicate and is parameterized by a right and left join key.

import pandas as pd
import lotus
from lotus.models import LM

# configure the LM, and remember to export your API key
lm = LM(model="gpt-4o-mini")
lotus.settings.configure(lm=lm)

# create dataframes with course names and skills
courses_data = {
    "Course Name": [
        "History of the Atlantic World",
        "Riemannian Geometry",
        "Operating Systems",
        "Food Science",
        "Compilers",
        "Intro to computer science",
    ]
}
skills_data = {"Skill": ["Math", "Computer Science"]}
courses_df = pd.DataFrame(courses_data)
skills_df = pd.DataFrame(skills_data)

# lotus sem join 
res = courses_df.sem_join(skills_df, "Taking {Course Name} will help me learn {Skill}")
print(res)

There are 3 main model classes in LOTUS:

• LM: The language model class.
  • The LM class is built on top of the LiteLLM library, and supports any model that is supported by LiteLLM. See this page for examples of using models on OpenAI, Ollama, and vLLM. Any provider supported by LiteLLM should work. Check out litellm's documentation for more information.
• RM: The retrieval model class.
  • Any model from SentenceTransformers can be used with the SentenceTransformersRM class, by passing the model name to the model parameter (see an example here). Additionally, LiteLLMRM can be used with any model supported by LiteLLM (see an example here).
• Reranker: The reranker model class.
  • Any CrossEncoder from SentenceTransformers can be used with the CrossEncoderReranker class, by passing the model name to the model parameter (see an example here).

For recent updates related to LOTUS, follow @lianapatel_ on X.

If you use LOTUS or semantic operators in a research paper, please cite this work as follows:

@misc{patel2024lotusenablingsemanticqueries,
      title={LOTUS: Enabling Semantic Queries with LLMs Over Tables of Unstructured and Structured Data},
      author={Liana Patel and Siddharth Jha and Carlos Guestrin and Matei Zaharia},
      year={2024},
      eprint={2407.11418},
      archivePrefix={arXiv},
      primaryClass={cs.DB},
      url={https://arxiv.org/abs/2407.11418},
}