Show HN: SuperDuperDB – Open-source framework for integrating AI with databases (github.com)
Today we are officially launching SuperDuperDB, an open-source framework for integrating AI directly with major databases, including streaming inference, scalable model training, and vector search, with release of v0.1. on GitHub and on ProductHunt.
SuperDuperDB is not a database. It transforms your favorite database into an AI development and deployment environment (𝘮𝘢𝘬𝘪𝘯𝘨 𝘪𝘵 𝘴𝘶𝘱𝘦𝘳-𝘥𝘶𝘱𝘦𝘳).
SuperDuperDB eliminates complex MLOps pipelines, specialized vector databases - and the need to migrate and duplicate data by integrating AI at the data's source, directly on top of your existing data infrastructure. This massively simplifies building and managing AI applications.
SuperDuperDB provides a simple Python interface, but allows experts to drill down to any level of implementation detail such as models weights or training details.
Today’s release comes with the full integration of major SQL databases as well as further MongoDB support: PostgreSQL, MySQL, SQLite, DuckDB, Snowflake, BigQuery, ClickHouse, DataFusion, Druid, Impala, MSSQL, Oracle, pandas, Polars, PySpark, and Trino.
Currently Supported AI: Any model from PyTorch, Sklearn, HuggingFace as well as AI APIs such as OpenAI, Anthrophic, Cohere.
A few useful links: - Our website: https://superduperdb.com - Getting started docs: https://docs.superduperdb.com/docs/category/get-started/ - Our repo on Github: https://github.com/SuperDuperDB/superduperdb
Check the uses-cases that we have already implemented here https://docs.superduperdb.com/docs/category/use-cases as well as apps built by the community here https://github.com/SuperDuperDB/superduper-community-apps and try all of them with Jupyter your browser https://demo.superduperdb.com/
For more information about SuperDuperDB and why we believe it is much needed, read the blog post https://docs.superduperdb.com/blog/superduperdb-the-open-sou...
We are keen to hear your feedback!
All the best, Timo
35 comments
[ 5.4 ms ] story [ 87.1 ms ] threadThe upsides to this are pretty compelling. Any downsides vs traditional MLOps?
For us, the fact it supports Postgres and SQLite out of the box is awesome. It means we can run a model on the server to generate embeddings, sync these to a local SQLite and then run local inference on the same data structure.
Sure you can have a mesh of database, Have your input data in one data technology, run models on a distributed compute, save models on different database tech and manage the jobs on other!
Also, does it support Bard APIs?
Absolutely you can query the embeddings of 1.6M rows, you can try with lance or in memory vector search type. The scalablity will depends on embedding size, machine configurations, etc. Thanks
But I’m an engineer, I read the README and the website, and I still don’t know what Super-duper is.
Is it just a python library? Does it have its own persistence (it must)? It doesn’t appear to be a set of plugins for various DB’s but I could be wrong.
As such I don’t know how I’d use it. It might be helpful to describe the product in more concrete terms.
One of the phrases used in YC is: ACME makes soup taste better. We do it with a seasoning that chefs add to their broth.
Maybe that’s helpful. Explaining a product can be hard!
SuperDupeDB is really an end-to-end AI development and deployment framework wrapping and integrating your existing data infrastructure. It replaces MLOps entirely as it covers inference and model training.
Like taking the data from that source (e.g., SQL), processing them (e.g., pytorch or openai), and storing the results somewhere (e.g., data on Mongo metadata on SQL).
It actually consists of the following: 1. nifty abstractions for Data (e.g., sources, encoders, listeners), Metadata (e.g., vector indexes), Compute (e.g., sync, async, parallel). 2. gluing engine that transparently handles the interaction between components 3. out-of-the-box integrations with established tools (databases, AI models and APIs, compute engines)
This way, you can build customized data layers that sit on top of your database and save you from moving the data to dedicated systems (e.g., vector databases or MLops tools)
For further discussion, feel free to join our slack https://join.slack.com/t/superduperdb/shared_invite/zt-1zuoj...
One of the nice things about langchain is the code examples, making it easy to get simple services up and running. And because it’s a toolkit I can take what I need and leave the rest.
However, the ecosystem around langchain is really exploding, is there some way you can retool what you have to extend langchain with better DB support, rather than build your own thing?
However, achieving goals like "training your LLM" or enabling "real-time inference" requires more than just pipelines. For that, we have invested in enhancing compatibility with databases and facilitating parallel computing.
About your last point, I 'm not sure I fully understand. Do you mean to write a guide for moving lang-chain models to superduper? Or to create superduper wrappers for langchain ? Or to move the core functionalities of superduper to langchain ?
The guide, is something have in our immediate plans. The wrappers are under discussion. The latter I don't think it's possible due to architectural differences. For example, superduper is designed with multi-node environments in mind.
SuperDuperDB fernando@superduperdb.com via sendinblue.com
and "Fernando" is Fernando Guerra, their "Business, Marketing and Growth"
not cool and immediately put me off