# bigscience-workshop/promptsource

Toolkit for creating, sharing and using natural language prompts.

Repository: https://github.com/bigscience-workshop/promptsource
Canonical: https://ross.abutalabs.com/products/promptsource
Language: Python
License: Apache-2.0
License Family: permissive
Topics: natural-language-processing, nlp, machine-learning
Last push: 2023-10-23T17:59:41+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 1932, "days_push": 1045, "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 3031, forks 375 (observed 2026-08-28T04:07:38.914163+00:00)

## What it is
PromptSource is a Python toolkit and repository for creating, sharing, and applying natural language prompts to Hugging Face datasets. It hosts the P3 collection of thousands of community-contributed prompt templates written in Jinja, usable via a simple API.

## Use cases
- apply natural language prompts to NLP datasets
- create and share prompt templates for multitask fine-tuning
- build zero-shot or few-shot evaluation prompts for language models
- browse existing prompts for datasets like SNLI or ag_news
- generate prompt-based training data for models like T0

## When to choose
- you need a large existing pool of English prompt templates for Hugging Face datasets
- you are doing multitask prompt-based fine-tuning research
- you want a structured Jinja-based format for defining dataset prompts

## When to avoid
- you need prompts for production LLM applications rather than dataset-based research
- you require actively maintained tooling or recent Python version support
- you need multilingual prompt coverage beyond the English-focused P3 collection

## Facets
- artifact type: library
- maturity: maintenance
- function: nlp, prompt-engineering, data-science
- domain: machine-learning, large-language-models
- platform: python
- tags: prompting, prompt-templates, huggingface-datasets, p3, jinja-templates, crowdsourced-prompts, natural-language-processing

## Member repositories
- bigscience-workshop/promptsource (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:38.914163+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 2026-08-30T07:29:48.750986+00:00, confidence not recorded.
  - readme: https://github.com/bigscience-workshop/promptsource (fetched 2026-08-28T04:07:38.914163+00:00, sha 48ba7312bdb9)
  - registry_pypi: https://pypi.org/pypi/promptsource/json (fetched 2026-08-29T09:44:38.862802+00:00, sha 3be80bb3d977)
- Data as of 2026-08-30T08:39:29.467469+00:00.