Just wanted to share Embedbase. Embedbase is an open-source API to easily use OpenAI embeddings in production.
Started out as an internal API for an Obsidian plugin I developed called AVA. We use it in AVA to find similar notes (see small demo), but you can potentially use it to do anything that OpenAI embeddings can do.
Example of a ChatGPT powered for your documentation:
You can have a look at the GitHub repo here:
Would love to get some feedback!
PS: try out the ChatGPT powered documentation, there are tutorials to build it for yourself
Thanks for sharing. Will you be adding support for Weaviate?
Yes . Curious what use case would you need Weaviate for rather than Pinecone?
Thanks for the response. Mostly because Weaviate is open-source, and it also has more features.
Just shipped a playground to try https://embedbase.xyz
It supports & uses Supabase as vector database FYI now
Also shipped a JS SDK a few weeks ago
You can ask questions to the GPT4-powered documentation FYI
If you need any help with embeddings or GPT-4 contact me (even if you don’t use embedbase )
a peek into the future
Based on a lot of feedback, we have started integrating the gpt interface directly into our application and this is just the first step to help you ship ChatGPT-powered apps faster. Try the new GPT-4 dashboard here.
We’ve seen how difficult it is for people to get started with local Embedbase using Supabase, so we decided to implement Postgres, so you can start iterating right away.
other small goodies
Embedbase JS SDK now supports metadata & listing datasets. Imagine throwing unstructured data into embedbase and get all links and everything when searching ala perplexity.ai?
PS: If you feel like it, drop a️ Embedbase.
Super excited to finally open-source the embedbase dashboard.
Check it out here embedbase/dashboard at main · different-ai/embedbase · GitHub
Added a data table
Right in the open source dashboard
Try it out
sharing this week’s updates w/ you
you’re reading the embedbase weekly newsletter, where we share our progress towards building the platform that allows you to easily connect your data to llms (like chatgpt!)
read along to see what we’ve shipped since last week
click here to
a new one-liner to run embedbase
pip install embedbase && embedbase run
shipped a getting started colab
In addition to that we shipped a google colab (it’s like codesandbox but for python) that allows you to try embedbase using a local embedder and db, running both the api and and python sdk.
- we’ll make it easy to add data without code
- a drop-in js component to create a “chat to your data”
- much more!
Very cool, following your progress! What features it already has?
Less than 2 min to go from an idea to a ChatGPT bot you can show to your friends & colleagues.
New feature: Publish your Embedbase Playground
You can now easily share your Embedbase Playground with the public.
a) select data from Embedbase
b) choose a system prompt
c) click on share & copy the link
Embedbase wants to help you ship something in a few minutes, that scales with your needs.
This is just one little step towards helping you ship faster.
hey openai folks, been shipping lately,
sharing this week’s updates w/ you
you’re reading the embedbase weekly newsletter, where we share our progress towards building a platform that helps you to easily connect your data to llms (like chatgpt!)
> npm uninstall openai
you no longer need to learn multiple APIs, from context creation to generation, embedbase lets you do it all in one unified and dead simple API.
Ask Questions to your Notion Tables with Zapier
we integrated the no-code automation tool Zapier into Embedbase.
we wrote a tutorial that let you ask questions to your Notion here
- option to have your context built partly with internet results
- support for multiple LLMs on the hosted version
- support for agents as first class citizens
it’s 100% open source (everything) btw embedbase/dashboard at main · different-ai/embedbase · GitHub
your feedback means a lot