Haystack used to be distributed with the package farm-haystack (which continues to distribute 1.0), but yesterday we release Haystack 2.0 with a new package: haystack-ai
Haystack used to be a framework that centered around semantic search and retrieval, first released in 2020. And while the design allowed us to extend to new LLM technologies, it wasn’t an intuitive developer experience for our users. So we made the decision to completely rewrite the fundamentals of the framework. The new design provides a uniform way for _extending_ the framework. And makes it very transparent how AI components interact with eachother.
The announcement article explains our reasons for this major rewrite.
Indeed this tutorial on Haystack is a good one as an example: https://haystack.deepset.ai/tutorials/22_pipeline_with_promp...
It combines a retrieval step followed by a prompt layer which inserts the relevant context into the prompt. You can however change the 'retrieval step' with something that uses a proper embedding model and OpenAI also provides those if you want to. I tend to use lighter (cheaper) OSS models for this step though. PS: There's some functionality in the PromptNode to make sure you don't exceed prompt limit.
Hey, thanks for posting our article. I see a few comments here about memory: This is indeed a challenge which we're working on. And some other topics at hand are making oss models available for our implementation of the agent as well
We got together as a couple of colleagues and made this app just as the Tokyo Olympics wrapped up a few days ago. Hope you enjoy! The data is from:
- 120 years of Olympic History, Kaggle
- Olympics.com
- NBC Olympics.
Data last updated on Aug 9, 2021.
Cumul.io | New York, Leuven, Genk and Remote | Full-Time
Cumul.io is a data visualization and analytics SaaS platform currently serving 200+ customers across the globe. We have offices in Leuven, New York, Genk and the possibility of remote working. Currently, we are hiring for a number of roles in various locations:
This is quite cool, just out of curiosity, are you planning to keep this to exclusively coding topics and skills or to also include other CS related topics too? I have a lesson I built in Jupyter Notebooks about ciphers for example, it uses code to walk through some ideas and concepts of how they work. Would you see these types of lessons as a part of the platform?
You can in most cases invest in your relationship with the person and try create a friendship which can be tiring at first but you just gotta stick through it, but in some cases I understand this is just not possible. If you think you've tried that route enough and go to a colleague/team lead/whoever, it's important to make sure that you don't approach the situation like a 'telling someone off' way, rather 'I think there's a problem and I'd like to fix it' way though and make sure that that message comes across. Especially if everything else in your job seems worth it.
Sorry if this has been mentioned in a previous comment, there are many!: what I find really useful is f.lux for lighting and colour. Especially makes a difference later in the day. I noticed I couldn’t fall asleep after a long d’au staring at the screen and this really helped...
Nice! I was looking for an API that provides air quality as well as temperature data a while ago for a demo project. Used Air Visual in the end but would you know of any alternative? I realize this one here doesn’t include air quality indexes.