Neum AI (https://neum.ai) is an ETL platform for LLM data. It helps companies connect and synchronize data into vector stores in real-time. No more chatbots that respond with stale information, with Neum vectors are always accurate and up to date.
Hey PH!
I am David de Matheu and together with @pkevinc, we are the founders of Neum AI.
As we built LLM applications, we kept hitting a similar problem, where getting started building an initial PoC was easy, but scaling those applications and improving their quality was hard. The responses of the prompts were only as good as the context we provided. This was especially painful when the context was based on data sources that changed constantly like a Notion notebook or user generated content like listings.
With Neum AI, we have focused on ensuring that the data the LLM prompts have access to is always up to date. Whether someone made a change to the data source, or a new document was just added, we want to make sure the vector database is always fresh.
Keeping things up to date can be expensive, as re-embedding content can quickly rack up cost (ex. embedding 1TB of data with OpenAI Ada can cost almost $10k) To combat this, Neum AI only update the vectors that have changed based on their source of truth.
Today we support a handful of data integrations as well as vector stores with more getting added every week based on feedback. (Tell us if we are missing something you would need). Neum AI automatically detects the type of data being vectorized to load and chunk it appropriately before embedding it.
Other core features include:
- UI and API interfaces to configure data pipelines.
- BYO-databases both for sources and sinks (vector DBs). We include out of the box DBs as well for fast prototyping.
- Schedule pipelines to run in a variety of configurations including on a timed schedule or automatically based on changes (limited sources supported today).
- Pipeline management to see your pipelines status and latest run information.
- Pipeline querying allows you to quickly search a pipeline's vector store. Alternatively, query directly from the vector store you are using.
If you are connecting data with your LLMs today using Langchain or LlamaIndex to do one-off extractions, give Neum AI a try. It can help you automate those data extractions and ensure that your LLM prompt has always the latest data.
Kevin and I are in the chat ready to answer any questions you might have. We love to also chat on anything LLM related.
To access Neum AI visit: https://neum.ai
Hi @david_de_matheu, this is a great product for startups in the technology industry that rely heavily on up-to-date data. Is the NEUM updated in real-time?
@ricardo_luz Hey Ricardo, we support several syncing options including real time for select sources (we hope to expand to more soon).
Depending on your needs you can enable schedules for syncing or manually trigger re-syncs based on logic on your own application. (ex. Use just uploaded a new document)
Neum AI
Voxme: AI coach, insights, guidance
Neum AI
Expert Remote