Dify, a visual workflow to build/test LLM applications
github.com
github.com
I am totally fine with closed-source/commercial licenses, but please don't do a "Like Apache 2.0 but not really" type of license. It just confuses everyone.
You can pick from SSPL, BSL, Elastic license among others if you don't want to roll out your own.
> 2. As a contributor, you should agree that: a. The producer can adjust the open-source agreement to be more strict or relaxed as deemed necessary. b. Your contributed code may be used for commercial purposes, including but not limited to its cloud business operations.
This is not very contributor-friendly.
You could consider keeping an open-source core, and extensions for paid features.
You've put it way too politely IMHO. It's a license designed to attract contributors to help build their product for free, but leaves space for them to just re-license the product and sell it when the opportunity arises. It's not like we haven't seen this play already, and it hurts both contributors and users...
But usually you could just go with something like SSPL - which has sufficient protections for the developer (but not the user).
https://devboard.gitsense.com/langgenius?id=4d1dec9067&r=lan...
I created insights for the last 4 weeks and number of new contributors and stars peaked last week. This project has all the signs of a successfully organically grown project.
Haven’t tried it yet - still evaluating autogen.
Maybe, but I don't think so. The number of people engaging with the project is what I would expect from the number of stars received.
https://devboard.gitsense.com/langgenius?r=langgenius%2Fdify...
I've seen projects with 5,000+ stars in a month and only have 20-30 people interacting with the project.
Full disclosure: This is my tool
fyi typo
- Variables should have more types, like an array of objects
- Prompting should incorporate Jinja2/Nunjucks
- For every prompt, I should be able to create many different test examples, along with an answer key, and measure how well it does across many tests
- It should auto-save. I did a lot of prompting work and then clicked another icon. When I came back, all my work was gone. (In fact, I don't see where to save at all! Maybe I'm just missing it.)
When do they become better then just rolling your own custom code?
what we're trying do with Dify currently is to let people put together prototypes quicker and either get to production or fail at a faster rate.
we've seen it being helpful for non-technical folks to collaborate on a project well (e.g. importing documents for knowledge base, creating no-code workflow apps, etc)
everyone is apparently a license pioneer
"Essentially Apache with restrictions"
It turns out to get the most out of LLMs you need to program them.
it's ironic they decided to do this on a post about LLMs. Are they feeling threatened that LLMs are taking their jobs ?