# IBM/Dromedary

Dromedary: towards helpful, ethical and reliable LLMs.

Repository: https://github.com/IBM/Dromedary
Canonical: https://ross.abutalabs.com/products/dromedary
Language: Python
License: GPL-3.0
License Family: copyleft
Last push: 2025-09-18T00:08:16+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 42, release rhythm 35, longevity 87
- inputs: {"age_days": 1218, "days_push": 350, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1137, forks 88 (observed 2026-08-28T04:03:43.839422+00:00)

## What it is
Dromedary is an open-source self-aligned language model trained with minimal human supervision using the principle-driven SELF-ALIGN pipeline. Dromedary-2 applies supervised fine-tuning on LLaMA-2 with diverse user prompts and a principle-driven prompt to produce helpful, ethical, and reliable LLMs.

## Use cases
- train a self-aligned chatbot from a LLaMA base model
- align an LLM with principles using minimal human supervision
- run inference with a self-aligned open-source language model
- fine-tune LLaMA-2 with SFT on diverse instruction datasets
- research principle-driven LLM alignment methods

## When to choose
- you want to reproduce or study the SELF-ALIGN alignment pipeline
- you need an open-source alternative to RLHF-based aligned models
- you are doing research on principle-driven LLM alignment

## When to avoid
- you need a production-ready chatbot with commercial licensing (GPL-3.0 code, CC BY-NC data)
- you lack GPU resources for large model training or inference
- you need the latest actively developed alignment tooling

## Facets
- artifact type: library
- maturity: maintenance
- function: llm-training, machine-learning, prompt-engineering
- domain: large-language-models, artificial-intelligence, machine-learning
- platform: python
- tags: self-alignment, llama, rlaif, sft, neurips-2023, research, gpu, linux

## Member repositories
- IBM/Dromedary (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:43.839422+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 2026-08-30T06:36:13.723335+00:00, confidence not recorded.
  - readme: https://github.com/IBM/Dromedary (fetched 2026-08-28T04:03:43.839422+00:00, sha bf8e5957c523)
- Data as of 2026-08-30T08:39:29.467469+00:00.