[bounty] openwebui integration

[louis030195]

describe the definition of done

• [ ] build openwebui integration
• [ ] launch it on https://www.reddit.com/r/LocalLLaMA/ (just a post, video, or screenshot!)

let's discuss if the code should be here or somewhere else, not familiar with their tech

or if any details to discuss

/bounty 100

[linear[bot]]

MED-206 [bounty] openwebui integration

[algora-pbc[bot]]

💎 $100 bounty • Screenpi.pe

Steps to solve:

1. Start working: Comment /attempt #517 with your implementation plan
2. Submit work: Create a pull request including /claim #517 in the PR body to claim the bounty
3. Receive payment: 100% of the bounty is received 2-5 days post-reward. Make sure you are eligible for payouts

Thank you for contributing to mediar-ai/screenpipe!

Add a bounty • Share on socials

Attempt	Started (GMT+0)	Solution
🟢 @TanGentleman	Oct 17, 2024, 3:28:02 AM	WIP
🟢 @vishalpatil1899	Oct 21, 2024, 12:46:21 AM	WIP

[TanGentleman]

/attempt #517

There's two implementations that seem approachable, integrating with Open WebUI's "Functions," or creating a "Pipe." As far as I know, the difference to the user would be this:

• Option 1: Select your favorite LLM from any provider that supports tool calling, and create a ScreenpipeSearch Tool that can be called.
• Option 2: Select "Screenpipe + LLM" in the model dropdown that accepts the same input, but performs in-between steps, like synthesizing a screenpipe search, before serving the output as usual.

I'm going to aim to implement a simple version of Option 2! It'll run fully locally if all goes well :)

Algora profile	Completed bounties	Tech	Active attempts	Options
@TanGentleman	1 mediar-ai bounty	TypeScript		Cancel attempt

[louis030195]

great :)

[vishalpatil1899]

/attempt #517

Options

• Cancel my attempt

[TanGentleman]

Here's the source code for an Open WebUI pipeline to your screenpipe data:

https://github.com/TanGentleman/screenpipe-python-client/blob/main/full-function-pipe.py

To Do:

• Make a <30 sec video of adding this code directly into Open WebUI Workspace > Functions.
• Add a short .md file for a simple How-To

[louis030195]

@TanGentleman

can you tell me how to try this out? like step by step, i never used openwebui

i'm going to make the video today if possible, do you know any community that would like this?

thinking of people into local AI, e/acc, devtools, productivity

[TanGentleman]

Check out instructions: https://github.com/TanGentleman/screenpipe-python-client/blob/main/INSTRUCTIONS.md

Lots of communities would enjoy this! I'll make some demos over this week :)

[louis030195]

@TanGentleman can you add more details like models to use, how to query, screenshots step by step, i'm technical and it was hard for me i had to discover every step myself so i assume 99.9% people will give up

also it just does not work using gpt-4o at all

another issue i had is that i had a 6 month old openwebui image and it was using it so there was no functions (figured out and deleted old image, pulled new one, edge case)

oh and their save button does not work, i had used the default for grammar (btw no idea what is this, can you customise this settings UI? or its just raw guess?) but saving does not work at all

[louis030195]