Release v1.13.0

Enable streaming responses from the GPT

Substantially update embeddings UI to show embedding status per model

Remove unused connectors

Add admin notifications tab to send emails to users

Added dialog to add and assign bulk users and seats

Encode document IDs as some have '/' in them

Add multi-class predictions to classification models for spreadsheets

Add support for attaching metadata when creating embeddings

Enhanced web scraper for GPT to read PDFs and other documents as well as HTML

Expose an OpenAI-compliant set of endpoints (which proxy to vllm)

Data fabric restrictions apply to new search endpoint

OAI formatted v1/embeddings endpoint

Add swagger docs for OAI-compliant endpoints

Enable GPT streamed responses

Added support for specifying metadata filter for RAG query

Data Fabric vectorstore migration for new search endpoint

New endpoint for retrieving schema metadata for documents

Add AND/OR operator to search filters for enhanced query flexibility

Enable auto recovery for RabbitMQ

Only pull rawText if document is not vectorized

Remove recreate_index from GptGenerateCommand parameters

Handle unexpected documentId returned from LLM

No more NPE in /gpt/retrieval/deleteDocuments if document doesn't exist

DataFabricId needs to be snake-cased on calls to generateUrl in streaming mode

Text rendering issue in Curate for bioRxiv and medRxiv sources

Create embeddings should not look at docs not inside data fabric

Fix error handling for GPT stream requests

Layar notifications (scheduled downtime, new release, etc.).

Import users and assign seats in bulk.

Lack of required role now returns 403 forbidden (vs. 401 unauthorized)

Open the /register endpoint so that users can register correctly

Token timeout and revocation issues

404 error on email confirmation for new users

File extension was not appended when extracting images from PDFs

Allow for both conversation history and documents injection(RAG) in the same prompt

New OpenAI schema endpoint resolver

Add new endpoint for documents/collection deletion, remove overwrite embeddings flag, add validation check

Add the ability to add custom metadata to documents in Weaviate and the ability to filter on it

Add mandatory dataFabricId in retrieval/apis' and /createEmbeddings endpoint

Add new /retrieval/search functionalities.

Add internal endpoints with Weaviate helpers for data-migration.

Send documents in batches to Layar, along with collection stats.

Add ability to stream LLM responses

If document is too long for RAG, skip and error out instead of auto RAG

Re-enable token counts in logged metrics

Fix memory leak in python worker

Remove unneeded aggregation in pod metrics reporting

Move collections stats to separate internal API

Improve error handling

Skip vectorization for new properties and new collections

Health check now properly tied to model server deploys

General bug fixes

Add QSP logistic regression models for IBD

Complete LR model endpoint

Add logic to calculate MW and logP for property prediction

Add support for standalone mode for property prediction

General bug fixes