[Support] Collectathon - Hoarder

[Collectathon]

 

Application Name: Hoarder

Application Website: https://hoarder.app

Docker Hub: N/A

GitHub: https://github.com/hoarder-app/hoarder

 

A self-hostable bookmark-everything app with a touch of AI for the data hoarders out there.

 

Documentation: https://docs.hoarder.app

 

Support thread for Hoarder docker image.

Edited May 19 by Collectathon
Updated GitHub link

[nirav]

I noticed that CA has 2 of these. One is the actual Hoarder app and the other is a "worker".  Is the Hoarder worker required along side Hoarder container on a single server?  On the app github page - i don't see any mention of the worker. So not sure what it even does. 

[Collectathon]

13 minutes ago, nirav said:

I noticed that CA has 2 of these. One is the actual Hoarder app and the other is a "worker".  Is the Hoarder worker required along side Hoarder container on a single server?  On the app github page - i don't see any mention of the worker. So not sure what it even does.

The hoarder app will save and store your bookmarks. The worker app will then fetch additional information. It will fetch the web page for archival with browserless and automatically create tags with chat/ollama. 

[nirav]

thanks!

[drmetro]

How does this app work and what are its usecases ?

[Collectathon]

1 hour ago, drmetro said:

How does this app work and what are its usecases ?

The main use case for Hoarder is as a "read-it-later" app. You can save interesting articles, tools, or other content they find while browsing on your phone or desktop, and then access and read that content later across devices. The developer built Hoarder as a self-hosted alternative to bookmark managers like Pocket, with inspiration from open-source projects like memos and mymind. They wanted a bookmark manager they could host themselves, with features like link previews and automatic AI-based tagging. You can find more info in the GitHub README.

[Trustwbc]

Hi i installed the "Server" Haorder-App at 192.168.2.207 and the Worker on 192.168.2.211 - the Installation in the same .207 was not possible.

Browserless installed and its running, meilisearch installed and its running.

But:

How does the App know, that the worker is an alternative Adress .211 ?

 

It looks so:

 

 

[Collectathon]

17 hours ago, Trustwbc said:

Hi i installed the "Server" Haorder-App at 192.168.2.207 and the Worker on 192.168.2.211 - the Installation in the same .207 was not possible.

Browserless installed and its running, meilisearch installed and its running.

But:

How does the App know, that the worker is an alternative Adress .211 ?

 

It looks so:

 

 

They access the same database and communicate through redis. The data dir of the worker needs to be the same as the web container. 

Edited May 19 by Collectathon

[schubdog]

Are there any plans to have only a single image with the worker and the app like an All in One package ?

 

Also does Hoarder meanwhile supports import of booksmarks from Chrome etc ?

 

I really like the Idea of giving tags KI based but im still not convinced why i should move from Linkwarden if the rest is still under heavy development

[Collectathon]

4 hours ago, schubdog said:

Are there any plans to have only a single image with the worker and the app like an All in One package ?

 

Also does Hoarder meanwhile supports import of booksmarks from Chrome etc ?

 

I really like the Idea of giving tags KI based but im still not convinced why i should move from Linkwarden if the rest is still under heavy development

There will be an AIO template if the developer creates an AIO docker image. From what I can tell, this is on the roadmap, but it's not a priority. 

 

Yes, Hoarder supports importing bookmarks from Chrome.

https://docs.hoarder.app/import/

 

You can run both simultaneously to try it out. I'm not trying to convince you to switch, I just like the software so I made a template. 

[luisalrp]

Has anyone managed to get it working using LocalAI? What models in LocalAI and variables in Hoarder? Thanks!

[Collectathon]

17 hours ago, luisalrp said:

Has anyone managed to get it working using LocalAI? What models in LocalAI and variables in Hoarder? Thanks!

I haven't used LocalAI before but in theory, you can add/modify the variables below as it should be a drop-in replacement for OpenAI. Other who have actually used it may be able to provide more support.

OPENAI_BASE_URL = http://localhost:8080
OPENAI_API_KEY = sk-XXXXXXXXXXXXXXXXXXXX
INFERENCE_TEXT_MODEL = gpt-4

 

Edited Friday at 06:16 AM by Collectathon

[mbassem]

On 5/24/2024 at 7:15 AM, Collectathon said:

I haven't used LocalAI before but in theory, you can add/modify the variables below as it should be a drop-in replacement for OpenAI. Other who have actually used it may be able to provide more support.

OPENAI_BASE_URL = http://localhost:8080
OPENAI_API_KEY = sk-XXXXXXXXXXXXXXXXXXXX
INFERENCE_TEXT_MODEL = gpt-4

 

 

Hi, Hoarder's developer here.

 

I highly recommend against the `gpt-4` model. `gpt-4` costs $30 / 1M tokens which is extremely expensive. Hoarder defaults to `gpt-3.5-turbo-0125` for text which costs $0.5 / 1M tokens (notice the huge difference!). If you want `gpt-4` level of inference, go for `gpt-4o` which is $5 / 1M tokens still much cheaper than the `gpt-4` model.