kacperlukawski

6 ms·

1.
Ask HN: What's the best-managed open source repo you've seen on GitHub?
2 points by kacperlukawski 1mo ago|1 comments
2.
Bielik.ai: Community-built, open-source LLMs for Polish and European languages(bielik.ai)
3 points by kacperlukawski 2mo ago|0 comments
3.
Haystack 3.0: Agents with hooks, skills, and built-in introspection(haystack.deepset.ai)
8 points by kacperlukawski 3mo ago|1 comments
4.
New multimodal Gemini embeddings from Google (videos and PDFs supported)(haystack.deepset.ai)
1 points by kacperlukawski 7mo ago|0 comments
5.
by kacperlukawski 10mo ago
Although it's in a different area, I wanted to mention https://calmcode.io/ as an excellent example of a calm learning platform. There is a whole movement around enshittification, and I see potential in this kind of ap
6.
Ask HN: Front end stack for a new app in 2025
1 points by kacperlukawski 2y ago|2 comments
7.
Elasticsearch Hybrid Search in Practice(softwaredoug.com)
3 points by kacperlukawski 2y ago|0 comments
8.
by kacperlukawski 2y ago
Is there a free trial available? Even 24 hours should be enough to say if I like it, but currently I have to pay from the day one. Or did I miss it?
9.
by kacperlukawski 2y ago
I'm always a bit worried if an extension gets permission to do anything it wants with all my files, including deleting them. Is there a way to restrict it and allow it to modify only the files it created?
10.
by kacperlukawski 2y ago
The problem is to scale that properly. If you have millions of documents, that won't scale that well. You are not going to prompt the LLM millions of times, aren't you? Embedding models usually have fewer parameters than the LLMs,
11.
by kacperlukawski 2y ago
Interesting! Does it work based on speech or transcriptions?
12.
by kacperlukawski 2y ago
Why is that an issue? Training the tokenizer seems much more straightforward than training the model as it is based on the statistics of the input data. I guess it may take a while for massive datasets, but is calculating the frequencies im
13.
by kacperlukawski 2y ago
Are there any specific reasons for using BPE, not Unigram, in LLMs? I've been trying to understand the impact of the tokenization algorithm, and Unigram was reported to be a better alternative (e.g., Byte Pair Encoding is Suboptimal fo
14.
by kacperlukawski 3y ago
Yeah, it seems like it got published too early.
15.
OpenAI announces GPT-4.5 Turbo(bing.com)
5 points by kacperlukawski 3y ago|2 comments
16.
by kacperlukawski 3y ago
I'm unsure if there is any comparison of LanceDB and Qdrant available out there, but there shouldn't be any issues with Python 3.12 and qdrant-client compatibility. Windows is also not a problem, as the typical local setup is usua
17.
by kacperlukawski 3y ago
If you will be the only app user, then the Python SDK's local mode might be suitable. However, in the long run, when you decide to publish the app, you rather have to switch to an on-premise or cloud environment. Using Qdrant from the
18.
by kacperlukawski 3y ago
Qdrant here! We're already working on that :D
19.
by kacperlukawski 3y ago
How would you host sentence-transformers model for free? You need it to vectorize each query so that has to be hosted somewhere. Is there any way to do it for free?
20.
by kacperlukawski 3y ago
If you need semantic search locally then it's fine, but serving an embedding model might be still challenging. And if you want to expose it, your laptop might be not enough.
21.
by kacperlukawski 3y ago
This is also an interesting piece of how to do it completely for free: https://news.ycombinator.com/item?id=36693239
22.
by kacperlukawski 3y ago
It's a bit easier in Python if you use tools like https://www.serverless.com/ . I'm not sure if Rust has something similar yet.
23.
by kacperlukawski 3y ago
It's Hugo, with a custom styling. https://gohugo.io/documentation/
24.
Product Quantization in Vector Search(qdrant.tech)
10 points by kacperlukawski 3y ago|0 comments
25.
by kacperlukawski 3y ago
Definitely Qdrant is a great option. It's an open source vector DB, so you can experiment locally without any cost, but if you prefer managed solution Qdrant Cloud is an option: https://cloud.qdrant.io/
26.
Qdrant 1.2 Release(qdrant.tech)
9 points by kacperlukawski 3y ago|0 comments
27.
by kacperlukawski 3y ago
Chroma doesn't seem to be a real DB, it's rather a wrapper around tools like hnswlib, DuckDB or Clickhouse. Qdrant is way more mature - it has its own HNSW implementation with some tweaks to incorporate filtering directly during t
28.
by kacperlukawski 3y ago
Did you run the clients in the same regions as the servers? That may impact the results.
29.
by kacperlukawski 3y ago
There are some other options in between as well. FAISS is a library, so not suited well for production usage unless a single machine is enough. The variety is wider than SaaS vs library. Tools such as Qdrant or Weaviate are Open Source. An
30.
by kacperlukawski 3y ago
I'd love to hear more about your thoughts on the complexity that cannot be, in your opinion, captured by the vector DB. I probably didn't get your point. Disclaimer: I work for Qdrant, and we believe a database should be just a da

More ›