# lotus-data/lotus

Optimized Agentic and LLM Bulk Processing Over Your Data

Repository: https://github.com/lotus-data/lotus
Canonical: https://ross.abutalabs.com/products/lotus-data-lotus
Homepage: https://lotus-data.github.io
Language: Python
License: Apache-2.0
License Family: permissive
Topics: ai-data-processing, data, llm, pandas, python, semantic-operators, semantic-search, unstructured-data, llm-data-processing, llm-document-processing, agentic-data-processing
Last push: 2026-07-03T22:55:38+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 90, release rhythm 91, longevity 55
- inputs: {"age_days": 778, "days_push": 61, "days_rel": 61, "gap_med": 18, "n_releases_24m": 8}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1661, forks 151 (observed 2026-08-28T04:05:18.513836+00:00)

## What it is
LOTUS is a Python library providing a Pandas-like API of LLM-powered semantic operators (map, filter, extract, aggregate, top-k) for bulk processing structured and unstructured datasets with agents and LLMs. It optimizes these declarative operations for higher accuracy and lower cost, with reported speedups up to 400x.

## Use cases
- classify or filter thousands of documents with natural language criteria
- extract structured fields from unstructured text at scale
- run an agent over every file in a codebase and summarize results
- build RAG pipelines over a document corpus
- score model outputs row by row with LLM-as-judge evals
- mine agent logs for failure modes
- fan out deep research over many papers and synthesize findings

## When to choose
- you need to apply LLM reasoning across a large dataset or document collection
- you want a Pandas-like declarative API instead of writing per-row LLM prompts
- cost and latency of naive LLM bulk processing are too high
- you want to combine structured and unstructured data in one query pipeline

## When to avoid
- you need simple single-prompt LLM calls without bulk dataset operations
- your data is small enough that direct API calls suffice
- you need a no-code GUI tool rather than a Python library
- you require strict deterministic transformations with no LLM nondeterminism

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-inference, rag, etl, data-science, agent-framework, search-engine
- domain: large-language-models, data-science
- platform: python
- tags: semantic-operators, pandas, llm-data-processing, query-engine, unstructured-data, declarative-programming, data-engineering, natural-language-processing, retrieval-augmented-generation, ai-agents

## Member repositories
- lotus-data/lotus (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:18.513836+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:44:38.063618+00:00, confidence not recorded.
  - readme: https://github.com/lotus-data/lotus (fetched 2026-08-28T04:05:18.513836+00:00, sha 71d37646986a)
  - homepage: https://lotus-data.github.io (fetched 2026-08-29T11:16:58.379307+00:00, sha 2f04d362eac0)
- Data as of 2026-08-30T08:39:29.467469+00:00.