.jpg)
We’re launching an embedding API as part of JigsawStack AI SDK. We see an opportunity here for RAG applications, especially for applications that have a wide range of different document types, including images, audio, and multiple languages
Most embedding models, even the ones provided by OpenAI, only support one modality, which is text. Language support is another key issue where most models tend to specialize in English or a small set of languages, which could be a huge pain point for larger multilingual models. Having different embedding models for each modality and language doesn’t work for embedding models as they share different vector spaces, which won’t allow for retrieval across languages and modalities.
So, to solve this, we built a model that supports multi-modality like most LLMs do today with PDF, image, audio, and text and with support for over 80+ languages.
Unlike many JigsawStack services, embedding models are hard to test with just our small team. Embedding models are used for a wide range of industries, law, Chinese documents, medical notes, general searches, and the list goes on. We need to test this in the wild in production applications with real documents. This model is in Alpha, and we need your help to get it to Beta production. While this model is in Alpha, it is still hosted on the same infra as every of JigsawStack’s which shares the same security standard and document management policy keeping everything private and deletable.
Technical specs:
Drop me a DM here or email me at [email protected] to get the API. We’re happy to provide the embedding model free of charge to whoever has helped test the model in Alpha and for an additional 4 months free with the full suite of JigsawStack when the model is in production.