# jondurbin/airoboros

Customizable implementation of the self-instruct paper.

Repository: https://github.com/jondurbin/airoboros
Canonical: https://ross.abutalabs.com/products/airoboros
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2024-03-07T16:23:51+00:00

## Health v2 (maintenance only)
Score: 20/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 8, longevity 87
- inputs: {"age_days": 1222, "days_push": 909, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1051, forks 66 (observed 2026-08-28T04:03:23.218689+00:00)

## What it is
Airoboros is a Python library implementing a heavily modified version of the Self-Instruct paper to generate high-quality synthetic instruction datasets using large language models like GPT-4. It supports multiple instructor styles (reasoning, roleplay, etc.), uses Chroma vector similarity for deduplication, and produces datasets for fine-tuning smaller open-source models.

## Use cases
- generate synthetic instruction datasets for fine-tuning LLMs
- create training data without human annotation
- build expert models fine-tuned on task-specific data
- generate orca-style reasoning and math training examples
- deduplicate generated instructions with vector similarity
- fine-tune llama models on high-quality synthetic data

## When to choose
- you need to create instruction-tuning datasets automatically with an LLM
- you want diverse synthetic prompts without human-written seeds
- you want task-specific instructors for different data styles
- you plan to fine-tune smaller open models to mimic larger ones

## When to avoid
- you need a maintained tool with active development
- you want simple prompt templates rather than a full data generation pipeline
- you cannot access OpenAI-compatible completion APIs
- you need human-verified training data

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-training, data-generation, rag, machine-learning
- domain: large-language-models, machine-learning, artificial-intelligence, developer-tools
- platform: python, cli
- tags: self-instruct, synthetic-data, fine-tuning, instruction-tuning, llama, vector-database, dataset-generation

## Member repositories
- jondurbin/airoboros (main) score 20

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:23.218689+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 2026-08-30T07:00:04.636199+00:00, confidence not recorded.
  - readme: https://github.com/jondurbin/airoboros (fetched 2026-08-28T04:03:23.218689+00:00, sha 77d556776c3e)
  - registry_pypi: https://pypi.org/pypi/airoboros/json (fetched 2026-08-29T13:01:34.825179+00:00, sha 0097a7676c7c)
- Data as of 2026-08-30T08:39:29.467469+00:00.