PromptSource

Free plan APISelf-hostedWeb

What PromptSource is

PromptSource is a free, open-source toolkit for developers who create, share, and use natural-language prompts. Its web interface lets developers write prompts in Jinja and inspect their outputs against examples. The app provides Sourcing, Prompted dataset viewer, and Helicopter view modes. Prompts are kept in structured files and can be applied to examples from Hugging Face Datasets. The API provides Template, Metadata, DatasetTemplates, and TemplateCollection classes for working with prompts and related information. The repository describes the P3 collection as containing about 2,000 English prompts for more than 170 English datasets as of January 20, 2022. Prompt metadata can include evaluation metrics, whether a prompt shows answer choices, and whether it represents the original task. Developers can install the package with pip and launch the local app using Streamlit after setup. Hosted versions allow browsing, but the BigScience-hosted app disables Sourcing mode. Some datasets must be downloaded manually. The package specifies Python >=3.7,<4.0, and stable prompt creation through the interface currently requires Python 3.7.

Who it suits

PromptSource is suited to developers building or exploring natural-language prompts for datasets. It also provides an API and structured prompt files for users working with prompts in code.

What it does well

Free and licensed under Apache-2.0
Web-based prompt editor previews outputs on examples
Includes an API for prompts and metadata
P3 covers more than 170 English datasets

What to know first

BigScience hosted app disables Sourcing mode
Some datasets require manual downloads
Stable interface-based prompt creation requires Python 3.7

Limits

The hosted BigScience app does not provide Sourcing mode. Some datasets require manual downloads into ~/.cache/promptsource by default, although PROMPTSOURCE_MANUAL_DATASET_DIR can change that location.

Getting started

Install the package with pip install promptsource
Use Python 3.7 for stable prompt creation through the interface
Complete any required manual dataset downloads
Launch the local interface with streamlit run promptsource/app.py
Use the hosted version to browse existing prompts

Verdict

PromptSource combines a prompt editor, dataset views, an API, and a substantial English prompt collection. Developers should note its Python requirement and the limits of the hosted Sourcing mode.

Questions

How much does PromptSource cost?

PromptSource is free, and its repository is open source under the Apache-2.0 license.

Which platforms can run PromptSource?

It is available through an API, self-hosted installation, and web interface.

What programming language does the editor use for prompts?

Prompt templates are written in Jinja and stored in standalone structured files.

Does PromptSource include an existing prompt library?

The P3 collection contains about 2,000 English prompts for more than 170 English datasets, as described for January 20, 2022.

Can PromptSource use Hugging Face datasets?

Yes. Prompts can be applied to examples from the Hugging Face Datasets library.

How do I run PromptSource locally?

Install it with pip install promptsource, complete setup, then run streamlit run promptsource/app.py.

Facts from PromptSource's own pages

API: The API provides Template, Metadata, DatasetTemplates, and TemplateCollection classes to store, manipulate, and use prompts and metadata. · github.com, 9 Oct 2026
App modes: The app has Sourcing, Prompted dataset viewer, and Helicopter view modes. · github.com, 9 Oct 2026
Dataset integration: Prompts can be applied to examples from the Hugging Face Datasets library. · github.com, 9 Oct 2026
Hosted access: The README links to hosted versions for browsing existing prompts and notes that the BigScience hosted app disables Sourcing mode. · github.com, 9 Oct 2026
Hosted version: The README links to a hosted version for browsing existing prompts and says its Sourcing mode is disabled. · github.com, 9 Oct 2026
Installation: The README gives pip install promptsource as an installation command for users who do not intend to create new prompts. · github.com, 9 Oct 2026
Intended audience: The package metadata lists developers as the intended audience. · github.com, 9 Oct 2026
Intended contributors: The contribution guide says a way to grow P3 is to write prompts for new datasets and submit them in a pull request. · github.com, 9 Oct 2026
License: The public repository is licensed under Apache-2.0. · github.com, 9 Oct 2026
Local app: The README says to launch the app locally with streamlit run promptsource/app.py after completing setup. · github.com, 9 Oct 2026
Local install: The README gives pip install promptsource as an installation option and documents launching the app locally with Streamlit. · github.com, 9 Oct 2026
Manual datasets: Some datasets require manual download into ~/.cache/promptsource by default, and the location can be overridden with PROMPTSOURCE_MANUAL_DATASET_DIR. · github.com, 9 Oct 2026
Manual dataset setup: Some datasets require manual download into ~/.cache/promptsource by default, and the path can be changed with PROMPTSOURCE_MANUAL_DATASET_DIR. · github.com, 9 Oct 2026
Prompt collection: The README describes a collection of about 2,000 English prompts for more than 170 English datasets as of January 20, 2022. · github.com, 9 Oct 2026
Prompt editor: Its web-based GUI lets developers write prompts in a templating language and view outputs on examples. · github.com, 9 Oct 2026
Prompt format: Prompts are stored in standalone structured files and written in Jinja. · github.com, 9 Oct 2026
Prompt library: The repository describes a collection called P3, with about 2,000 English prompts for more than 170 English datasets as of January 20, 2022. · github.com, 9 Oct 2026
Prompt metadata: Prompt metadata can include the original task flag, whether answer choices appear in the prompt, and evaluation metrics. · github.com, 9 Oct 2026
Purpose: PromptSource is a toolkit for creating, sharing, and using natural language prompts. · github.com, 9 Oct 2026
Runtime requirement: The package metadata specifies Python >=3.7,<4.0, and the README says creating prompts through the interface currently requires Python 3.7 for stability. · github.com, 9 Oct 2026