Danswer - OpenSource Enterprise Question-Answering AI System

[jagan]

Danswer allows you to ask natural language questions against internal documents and get back reliable answers backed by quotes and references from the source material so that you can always trust what you get back. You can connect to a number of common tools such as Slack, GitHub, Confluence, amongst others.

https://github.com/danswer-ai/danswer

We need an AI tool for Cloudron now!

Check out the Video Demo below:

Features

• Direct QA powered by Generative AI models with answers backed by quotes and source links.
• Intelligent Document Retrieval (Semantic Search/Reranking) using the latest LLMs.
• An AI Helper backed by a custom Deep Learning model to interpret user intent.
• User authentication and document level access management.
• Connectors to Slack, GitHub, GoogleDrive, Confluence, local files, and web scraping, with more to come.
• Management Dashboard to manage connectors and set up features such as live update fetching.
• One line Docker Compose (or Kubernetes) deployment to host Danswer anywhere.

[Kubernetes]

I hope this is the right link?

[Kubernetes]

Looks interesting, will have a look locally first

[Kubernetes]

I did several tries and I have to say it runs far from stable. I was not even able to use it for one single query. Installation with docker compose runs fine and I get the web interfaces, but when I start to configure it, it starts to break. I will stop to work on it.

[robi]

@Kubernetes sounds like the configuration writes files to its code dir, which need to be linked outside to the data dir.

[Kubernetes]

@robi That is not the case, I did install it on a empty cloudron-free VPS

[robi]

@Kubernetes I see, then it's an upstream issue or one might need to pull from master vs whatever release that is less current.

[Kubernetes]

@robi Not sure what you mean, but I did build it from repo source

[robi]

@Kubernetes that answers it.. ty.

[jagan]

This looks very new and recent.
The entire repo appears to be just a few days old.
Just excited by the possibility though, we will keep an eye on this.

[jagan]

Folks, here is a demo by Bookstack on Danswer, and it is very impressive:
Tea Break: 8 Years of BookStack, LLM Connection Demo & Video Hosting

[robi]

@jagan Thanks for sharing.

@girish sounds like a great opportunity to share some of your peertube content with Dan and make sure he knows how self-hosting on Cloudron is easier than he describes.

[luckym]

Just wanted to bump this as the project still seems to be under active development. They now offer a community edition freely available under the MIT Expat license, as well as a paid enterprise edition with additional features. The project was also rebranded under onyx.app.

[girish]

@luckym thanks for the note. Maybe we should close this task and create a new one. Onyx looks completely different from danswer. What do you think @LoudLemur ?

[jdaviescoates]

@girish I reckon just edit the OP to make it about https://github.com/onyx-dot-app/onyx using @LoudLemur's template

[LoudLemur]

@girish I can't see Danswer at the github now that it has rebranded. If @jagan still wants Onyx, a new thread would be a good occasion to support the request with the app request template, and gain some more upvotes.

Is the model and its training hidden for Onyx? Is Onyx's final answer generator OpenAI GPT-4 / GPT-3. 5-turbo?

AnythingLLM is somewhat similar to Onyx, and we have an Application Request for that already. AnythingLLM has a query mode which locks down responses to data you have provided:
https://docs.anythingllm.com/features/chat-modes

AnythingLLM also integrates with Ollama, which Cloudron already supports:
https://docs.anythingllm.com/ollama-connection-troubleshooting

AnythingLLM has a lot more contributors and seems less closed-box.

[girish]

Let's close/lock this thread then. @luckym please open a new thread for onyx! thanks