Show HN: Dera – A platform to manage chunks and embeddings for building RAG apps (opens in new tab)

(getdera.com)

25 pointshchua2y ago2 comments

Hi everyone,

my cofounder and I built Dera - a platform to help manage chunks and embeddings.

We built this because of the pain points we experienced while building RAG applications for side projects. The biggest pain point we encountered was that we were constantly trying out different chunking strategies, but there’s no easy way to check how the strategies are performing in terms of retrieval when given the same query.

We tried searching for a tool for this but couldn’t find any (most LLM dev tools focus on prompts management).

We hope this tool will be useful for people building RAG apps.

Thank you!

2 comments

pbronez2y ago

Very cool. I wonder when it makes sense to engineer things at this level vs using something like Azure AI search. [0]

Love to see version control on all the things! Wonder if the version control features would be more robust if implemented in Doltgres.

[0] https://azure.microsoft.com/en-us/products/ai-services/ai-se...

[1] https://github.com/dolthub/doltgresql

skeptrune2y ago

Seems quite cool! I definitely agree that there is a gap in the current RAG solution set for iterating on chunking and embedding strategies. Even the prompt chain for querying can be quite hard to adjust.

I tried out the tool itself and found the "define your schema" portion a bit confusing. I did eventually grasp it, but it took me a minute or two. The example curl requests on the page after I made my project were very helpful.

Thought I would expect that most folks doing RAG would probably prefer python that they could copy and past into a notebook.

Hopefully more people find it helpful!

j / k navigate · click thread line to collapse