Last updated
Graphlit now charges for LLM tokens at the cost of the LLM provider. For example, if you use 10,000 input tokens and 1000 output tokens from OpenAI GPT-4o, Graphlit will convert the token usage to credits at the cost of the LLM provider, without any additional margin added.
Graphlit now charges a flat-rate of 2.5 credits ($0.25/Mtok on Hobby tier) per million input tokens for LLM completions, no matter what LLM is used. This will make the cost of LLM usage cheaper in the Graphlit platform, and more consistent, no matter if you are using OpenAI o1 or Deepseek Coder. This ends up reducing the cost of token-intensive operations by 25% or more, on average.
Graphlit now supports deleting multi-tenant entities from project-scope. For example, if an end-user has created a feed with a unique 'owner ID', you can now call deleteFeed from project scope to delete the user's feed.
We now support the retrieval of multi-tenant conversations from project-scope. For example, if an end-user has created a conversation with a unique 'owner ID', when you query conversations from project scope, the end-user's conversations will be included with the retrieved conversations.
GPLA-3764: Groq custom model not using proper API endpoint
GPLA-3754: Not handling Microsoft emails with null parent folder ID
GPLA-3747: Need retry on Jina embeddings
GPLA-2962: Not crawling website where host (w/o 'www') doesn't match sitemap host
GPLA-3750: Not crawling website when sitemap returns application/rss+xml MIME type
GPLA-3742: DeleteContents/DeleteAllContents have missing arguments in Node.js SDK
GPLA-3757: Error on indexing of PDF was not assigning content.error field.
