0.1.1

What's Changed

• Auto-adaptive parallelization for Sentence Transformers' .encode() by @bensteinberg in #7 .
  This change allows for specifying how many text chunks should be encoded in parallel by the text similarity model when ingesting documents. This feature automatically defaults to processing 1 chunk at a time if performance worsens.
• Update langchain-core by @bensteinberg in #9

New Contributors

• @bensteinberg made their first contribution in #7

Full Changelog: 0.1.0...0.1.1

0.1.0

This new version of WARC-GPT comes with a new UI:

Screen.Recording.2024-03-20.at.5.09.28.PM.mov

This version also features a refactoring of the app's codebase. The API now separates search (/api/search) from text completion (/api/complete).

Upgrading from 0.3.0:

• Vector stores generated with previous versions of WARC-GPT should still be usable.
• The app's dependencies may need to be re-installed.
• Minor changes have been made to .env which will require minor adjustments. See .env.example for details.

More info on changes:

• PR #6
• Full changelog: 0.0.3...0.1.0

Thanks & acknowledgements:

• @edsu for kickstarting work on streamed text completion
• @rebeccacremona for code review and suggestions

0.0.3

Changelog

• Dependencies update
• Ingestion pipeline bugfix: skips records that do not have HTTP headers early-on
• Minor readability-related edits

Full Changelog: 0.0.2...0.0.3

0.0.2

Changelog

• Better UI for API-level errors.
• Minor bugfixes

Full Changelog: 0.0.1...0.0.2

0.0.1

Initial release

Full Changelog: https://github.com/harvard-lil/warc-gpt/commits/0.0.1