This ship is taking me far away
This is the first beta release of LightOn Muse API 🎉. We will be following up with regular updates, adding new features and fine-tuning existing ones. Feel free to directly address us any feedback you may have.
We are excited to see what you will build with the LightOn Muse API!
- The API has been completely revamped, all requests in the old format will fail (but API keys are unchanged).
- Updated the base
orion-frmodel, which is now more capable and powerful;
- Extended support for controllable generation with the Create endpoint,
stop_words, and more;
- Added Understand endpoints to help make sense of text: Represent to compute rich contextual embeddings, Analyse to access the likelihood of words in a sentence, or directly use Compare and Select to process multiple sentences.
- The API should now return more explicit error & warnings for invalid requests (WIP).
- Python bindings for the API & demo notebooks (W33-34);
- Behind-the-scene improvements to batching, enabling lower latency and higher throughput (W33-34);
- More structured metadata and warnings (W33-34);
auriga-esalpha release (W34-35).