lqhl

7 ms·

1.
Exploring AI Memory Architectures (Part 2): MemOS Framework(blog.lqhl.me)
9 points by lqhl 1y ago|0 comments
2.
Exploring AI Memory Architectures (Part 3): From Prototype to Blueprint(blog.lqhl.me)
2 points by lqhl 1y ago|0 comments
3.
Exploring AI Memory Architectures (Part 1): A Deep Dive into Memory³(blog.lqhl.me)
2 points by lqhl 1y ago|0 comments
4.
MyScaleDB, a fork of ClickHouse, integrated Tantivy for better full-text search(thenewstack.io)
2 points by lqhl 2y ago|0 comments
5.
by lqhl 2y ago
Oh, we missed this new standard. Our first version must have been implemented before its release. We should definitely consider it now.
6.
by lqhl 2y ago
LLM applications can benefit from Retrieval-Augmented Generation (RAG) in a similar way that humans benefit from search engines like Google. Therefore, I believe RAG cannot be replaced by prompts or fine-tuning. https://myscale.c
7.
by lqhl 2y ago
MyScaleDB utilizes approximate nearest neighbors (ANN) algorithms such as ScaNN, HNSW, and IVF. As a result, it may not achieve a 100% recall rate. However, depending on the search parameters used, it can attain recall rates of up to 95% or
8.
by lqhl 3y ago
Since we developed our internal version before the release of usearch, we integrated hnswlib with faiss's PQ/SQ algorithms. Search performance is typically not a concern in real-world applications, especially for those using LLMs
9.
Filtered Vector Search: A Key Technology in RAG Systems(blog.lqhl.me)
2 points by lqhl 3y ago|0 comments
10.
Ask HN: Should I switch to Grok/Bard/Claude?
2 points by lqhl 3y ago|0 comments
11.
by lqhl 3y ago
I'm not familiar with it. maybe you can submit a PR?
12.
Comparison of Vector Databases(lqhl.github.io)
6 points by lqhl 3y ago|4 comments
13.
by lqhl 3y ago
A comparison matrix of some vector databases. Feel free to submit PRs on https://github.com/lqhl/vectordb-comparison/ to contribute!
14.
by lqhl 3y ago
I work on MyScale ( https://myscale.com ), a fully-managed vector database based on ClickHouse. Some unique features of MyScale: 1. This solution is built on ClickHouse and offers comprehensive SQL support. Our users leverage vect
15.
by lqhl 3y ago
We conducted a benchmark to evaluate the precision, throughput (QPS), insert speed, build speed, and cost-effectiveness of Pinecone, Qdrant, MyScale, Weaviate, and Zilliz (Milvus). This information will be valuable for those seeking to choo
16.
by lqhl 3y ago
Compilance is also important. Like SOC 2
17.
by lqhl 3y ago
As jwells89 mentioned, LLMs can extract the intention from questions and generate better queries for a search engine or database.
18.
by lqhl 3y ago
This idea is a simplified version of Retrieval-Augmented Generation (RAG), and RAG has been studied in various research papers, such as the one available at https://arxiv.org/abs/2005.11401
19.
by lqhl 3y ago
This article suggests that LLMs should use a database as a reference for factual information. Rather than asking LLMs to provide their own answers, it is recommended that they summarize based on the facts extracted from the database. This a
20.
by lqhl 3y ago
That's true and we have tried different configurations for systems that use HNSW. But for the ease of presentation, we only choose the configuration with the highest throughput at precision 98%. Here is a figure in our open-source benc
21.
by lqhl 3y ago
Yes. For MyScale (aws us-east-1), Pinecone (aws us-east-1), Qdrant (aws us-east-1), and Zilliz Cloud (aws us-east-2), we run the clients in the same region as the servers. For Weaviate, the server is in GCP US east, while the client is in a
22.
The most cost-effective Vector Databases(blog.myscale.com)
9 points by lqhl 3y ago|11 comments
23.
by lqhl 3y ago
Over the past few months, we have been working on an exciting project to bridge the gap between high performance vector search and OLAP database. Today, we are thrilled to announce the release of our new end-to-end benchmark of MyScale, whi
24.
by lqhl 3y ago
You can explore MyScale at https://myscale.com/ . It's a SaaS platform built on ClickHouse, offering more sophisticated vector indexing options like HNSW and IVF compared to the open-source version. It also provides a f