# togethercomputer/OpenChatKit

Repository: https://github.com/togethercomputer/OpenChatKit
Canonical: https://ross.abutalabs.com/products/openchatkit
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2024-04-09T19:09:58+00:00

## Health v2 (maintenance only)
Score: 30/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 91
- inputs: {"age_days": 1280, "days_push": 876, "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 8984, forks 998 (observed 2026-08-28T04:10:26.618218+00:00)

## What it is
OpenChatKit is an open-source toolkit from Together, LAION, and Ontocord.ai for building and fine-tuning chat language models, including instruction-tuned models like GPT-NeoXT-Chat-Base-20B and Pythia-Chat-Base-7B. It bundles training code, inference tooling, a moderation model, and an experimental retrieval-augmented generation system.

## Use cases
- train a custom open-source chatbot model
- fine-tune llama-2 for long context conversations
- run local inference with an instruction-tuned chat model
- add retrieval from custom documents to a chat model
- reproduce the pythia chat base 7b training
- add content moderation to a chatbot

## When to choose
- you want to train or fine-tune your own open chat model
- you need an Apache-2.0 licensed chatbot base without API dependencies
- you want to experiment with retrieval-augmented chat models

## When to avoid
- you just want to call a hosted LLM API without training
- you lack GPU resources for 7B-20B parameter models
- you need a production-ready, actively developed chat framework

## Facets
- artifact type: framework
- maturity: maintenance
- function: llm-training, llm-inference, rag, chatbot, machine-learning
- domain: large-language-models, artificial-intelligence, chatbots, deep-learning
- platform: python, cli
- tags: chat-models, instruction-tuning, moderation-model, retrieval-augmentation, hugging-face, open-source-models, retrieval-augmented-generation, gpu, linux

## Member repositories
- togethercomputer/OpenChatKit (main) score 30

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:26.618218+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 2026-08-29T17:24:23.943388+00:00, confidence not recorded.
  - readme: https://github.com/togethercomputer/OpenChatKit (fetched 2026-08-28T04:10:26.618218+00:00, sha b9f9235d97ac)
- Data as of 2026-08-30T08:39:29.467469+00:00.