March 15: Support for Podscan search, image similarity search, 'exists' and 'upsert' mutations

New Features

• 💡 Graphlit now supports podcast searches using the Podscan search API with the searchWebmutation. In addition to Tavily and Exa, you can now specify PODSCANas a search service type. It will return the URL to the downloadable audio track of the podcast episode.

• 💡 Graphlit now supports the ingestion of podcast episodes with the web search feed. We will automatically ingest and transcribe the resulting podcast episodes into the Graphlit project, which become available for retrieval and RAG conversations.

• Graphlit now supports direct similarity search based on image embeddings, by providing Base64-encoded imageDataand imageMimeTypeproperties in the ContentFilterobject. We will use your configured image embeddings model, such as Jina, to embed the provided image and then use for vector search to return similar contents.

• Graphlit now supports workflowExists, specificationExistsand feedExistsqueries, which returns whether any workflow, specification or feed exists that matches the provided filter criteria.

• We have added upsert mutations for specifications (upsertSpecification) and workflows (upsertWorkflow), to simplify the creation of specifications and workflows, when you're not sure if they already exist. If the entity already exists by name, it will update the existing entity, else it will create a new entity with that name.

• We have added upsert mutations for labels (upsertLabel) and categories (upsertCategory), to simplify the creation of labels and categories by name, when you're not sure if they already exist.

• ⚡ We have changed the default number of retrieved entities to 100 (used to be 1000), to increase retrieval performance. You can assign the limitproperty on filter objects to raise this default.