Anchor Links

qdrant-landing/content/articles/binary-quantization-openai.md

@@ -40,7 +40,7 @@ You can also try out these techniques as described in [Binary Quantization OpenA
 
 ## New OpenAI embeddings: performance and changes
 
-As the technology of embedding models has advanced, demand has grown. Users are looking more for powerful and efficient text-embedding models. OpenAI's Ada-003 embeddings offer state-of-the-art performance on a wide range of NLP tasks, including those noted in [MTEB](https://huggingface.co/spaces/mteb/leaderboard) and [MIRACL](https://openai.com/blog/new-embedding-models-and-api-updates). 
+As the technology of [embedding models](/articles/fastembed/) has advanced, demand has grown. Users are looking more for powerful and efficient text-embedding models. OpenAI's Ada-003 embeddings offer state-of-the-art performance on a wide range of NLP tasks, including those noted in [MTEB](https://huggingface.co/spaces/mteb/leaderboard) and [MIRACL](https://openai.com/blog/new-embedding-models-and-api-updates). 
 
 These models include multilingual support in over 100 languages. The transition from text-embedding-ada-002 to text-embedding-3-large has led to a significant jump in performance scores (from 31.4% to 54.9% on MIRACL).
 
@@ -118,7 +118,7 @@ For those exploring the integration of text embedding models with Qdrant, it's c
 
 1. **Model Name**: Signifying the specific text embedding model variant, such as "text-embedding-3-large" or "text-embedding-3-small". This distinction correlates with the model's capacity, with "large" models offering more detailed embeddings at the cost of increased computational resources.
 
-2. **Dimensions**: This refers to the size of the vector embeddings produced by the model. Options range from 512 to 3072 dimensions. Higher dimensions could lead to more precise embeddings but might also increase the search time and memory usage in Qdrant.
+2. **Dimensions**: This refers to the size of the [vector embeddings](/articles/what-are-embeddings/) produced by the model. Options range from 512 to 3072 dimensions. Higher dimensions could lead to more precise embeddings but might also increase the search time and memory usage in Qdrant.
 
 Optimizing these parameters is a balancing act between search accuracy and resource efficiency. Testing across these combinations allows users to identify the configuration that best meets their specific needs, considering the trade-offs between computational resources and the quality of search results.
 

qdrant-landing/content/articles/binary-quantization.md

@@ -20,7 +20,7 @@ keywords:
 
 Qdrant is built to handle typical scaling challenges: high throughput, low latency and efficient indexing. **Binary quantization (BQ)** is our latest attempt to give our customers the edge they need to scale efficiently. This feature is particularly excellent for collections with large vector lengths and a large number of points. 
 
-Our results are dramatic: Using BQ will reduce your memory consumption and improve retrieval speeds by up to 40x.
+Our results are dramatic: Using BQ will reduce your [memory consumption](/articles/memory-consumption/) and improve retrieval speeds by up to 40x.
 
 As is the case with other quantization methods, these benefits come at the cost of recall degradation. However, our implementation lets you balance the tradeoff between speed and recall accuracy at time of search, rather than time of index creation. 
 
@@ -30,7 +30,7 @@ The rest of this article will cover:
 3. Benchmark analysis and usage recommendations
 
 ## What is Binary Quantization?
-Binary quantization (BQ) converts any vector embedding of floating point numbers into a vector of binary or boolean values. This feature is an extension of our past work on [scalar quantization](/articles/scalar-quantization/) where we convert `float32` to `uint8` and then leverage a specific SIMD CPU instruction to perform fast vector comparison. 
+Binary quantization (BQ) converts any [vector embedding](/articles/what-are-embeddings/) of floating point numbers into a vector of binary or boolean values. This feature is an extension of our past work on [scalar quantization](/articles/scalar-quantization/) where we convert `float32` to `uint8` and then leverage a specific SIMD CPU instruction to perform fast vector comparison. 
 
 ![What is binary quantization](/articles_data/binary-quantization/bq-2.png)
 

qdrant-landing/content/articles/chatgpt-plugin.md

@@ -36,7 +36,7 @@ These plugins, designed to enhance the model's performance, serve as modular ext
 that seamlessly interface with the core system. By adding a knowledge base plugin to 
 ChatGPT, we can effectively provide the AI with a curated, trustworthy source of 
 information, ensuring that the generated content is more accurate and relevant. Qdrant 
-may act as a vector database where all the facts will be stored and served to the model 
+may act as a [vector database](/qdrant-vector-database/) where all the facts will be stored and served to the model 
 upon request.
 
 If you’d like to ask ChatGPT questions about your data sources, such as files, notes, or 

qdrant-landing/content/articles/data-privacy.md

@@ -17,7 +17,7 @@ keywords: # Keywords for SEO
   - Enterprise Data Compliance
 ---
 
-Data stored in vector databases is often proprietary to the enterprise and may include sensitive information like customer records, legal contracts, electronic health records (EHR), financial data, and intellectual property. Moreover, strong security measures become critical to safeguarding this data. If the data stored in a vector database is not secured, it may open a vulnerability known as "[embedding inversion attack](https://arxiv.org/abs/2004.00053)," where malicious actors could potentially [reconstruct the original data from the embeddings](https://arxiv.org/pdf/2305.03010) themselves. 
+Data stored in vector databases is often proprietary to the enterprise and may include sensitive information like customer records, legal contracts, electronic health records (EHR), financial data, and intellectual property. Moreover, strong security measures become critical to safeguarding this data. If the data stored in a [vector databases](/qdrant-vector-database/) is not secured, it may open a vulnerability known as "[embedding inversion attack](https://arxiv.org/abs/2004.00053)," where malicious actors could potentially [reconstruct the original data from the embeddings](https://arxiv.org/pdf/2305.03010) themselves. 
 
 Strict compliance regulations govern data stored in vector databases across various industries. For instance, healthcare must comply with HIPAA, which dictates how protected health information (PHI) is stored, transmitted, and secured. Similarly, the financial services industry follows PCI DSS to safeguard sensitive financial data. These regulations require developers to ensure data storage and transmission comply with industry-specific legal frameworks across different regions. **As a result, features that enable data privacy, security and sovereignty are deciding factors when choosing the right vector database.**
 
@@ -234,7 +234,7 @@ Data governance varies by country, especially for global organizations dealing w
 
 To address these needs, the vector database you choose should support deployment and scaling within your controlled infrastructure. [Qdrant Hybrid Cloud](/documentation/hybrid-cloud/) offers this flexibility, along with features like sharding, replicas, JWT authentication, and monitoring.
 
-Qdrant Hybrid Cloud integrates Kubernetes clusters from various environments—cloud, on-premises, or edge—into a unified managed service. This allows organizations to manage Qdrant databases through the Qdrant Cloud UI while keeping the databases within their infrastructure.
+Qdrant Hybrid Cloud integrates Kubernetes clusters from various environments—cloud, on-premises, or edge—into a unified managed service. This allows organizations to manage Qdrant databases through the [Qdrant Cloud](/cloud/) UI while keeping the databases within their infrastructure.
 
 With JWT and RBAC, Qdrant Hybrid Cloud provides a secure, private, and sovereign vector store. Enterprises can scale their AI applications geographically, comply with local laws, and maintain strict data control.
 

qdrant-landing/content/articles/dedicated-service.md

@@ -21,7 +21,7 @@ keywords:
 Ever since the data science community discovered that vector search significantly improves LLM answers,
 various vendors and enthusiasts have been arguing over the proper solutions to store embeddings.
 
-Some say storing them in a specialized engine (aka vector database) is better. Others say that it's enough to use plugins for existing databases.
+Some say storing them in a specialized engine (aka [vector databases](/qdrant-vector-database/)) is better. Others say that it's enough to use plugins for existing databases.
 
 Here are [just](https://nextword.substack.com/p/vector-database-is-not-a-separate) a [few](https://stackoverflow.blog/2023/09/20/do-you-need-a-specialized-vector-database-to-implement-vector-search-well/) of [them](https://www.singlestore.com/blog/why-your-vector-database-should-not-be-a-vector-database/).
 
@@ -72,7 +72,7 @@ Those priorities lead to different architectural decisions that are not reproduc
 
 ###### Having a dedicated vector database requires duplication of data.
 
-By their very nature, vector embeddings are derivatives of the primary source data.
+By their very nature, [vector embeddings](/articles/what-are-embeddings/) are derivatives of the primary source data.
 
 In the vast majority of cases, embeddings are derived from some other data, such as text, images, or additional information stored in your system. So, in fact, all embeddings you have in your system can be considered transformations of some original source.
 

qdrant-landing/content/articles/discovery-search.md

@@ -103,4 +103,4 @@ This way you can give refreshing recommendations, while still being in control b
 - Discovery search is a powerful tool for controlled exploration in vector spaces.
 Context, positive, and negative vectors guide search parameters and refine results.
 - Real-world applications include multimodal search, diverse recommendations, and context-driven exploration.
-- Ready to experience the power of Qdrant's Discovery search for yourself? [Try a free demo](https://qdrant.tech/contact-us/) now and unlock the full potential of controlled exploration in vector spaces!
+- Ready to experience the power of Qdrant's Discovery search for yourself? [Try a free demo](/contact-us/) now and unlock the full potential of controlled exploration in vector spaces!

qdrant-landing/content/articles/fastembed.md

@@ -139,7 +139,7 @@ If anything changes, you'll see a new version number pop up, like going from 0.0
 
 ## Using FastEmbed with Qdrant
 
-Qdrant is a Vector Store, offering comprehensive, efficient, and scalable [enterprise solutions](https://qdrant.tech/enterprise-solutions/) for modern machine learning and AI applications. Whether you are dealing with billions of data points, require a low latency performant [vector database solution](https://qdrant.tech/qdrant-vector-database/), or specialized quantization methods – [Qdrant is engineered](/documentation/overview/) to meet those demands head-on.
+Qdrant is a Vector Store, offering comprehensive, efficient, and scalable [enterprise solutions](/enterprise-solutions/) for modern machine learning and AI applications. Whether you are dealing with billions of data points, require a low latency performant [vector database solution](/qdrant-vector-database/), or specialized quantization methods – [Qdrant is engineered](/documentation/overview/) to meet those demands head-on.
 
 The fusion of FastEmbed with Qdrant’s vector store capabilities enables a transparent workflow for seamless embedding generation, storage, and retrieval. This simplifies the API design — while still giving you the flexibility to make significant changes e.g. you can use FastEmbed to make your own embedding other than the DefaultEmbedding and use that with Qdrant.
 
@@ -229,7 +229,7 @@ Behind the scenes, we first convert the query_text to the embedding and use tha
 
 By following these steps, you effectively utilize the combined capabilities of FastEmbed and Qdrant, thereby streamlining your embedding generation and retrieval tasks.
 
-Qdrant is designed to handle large-scale datasets with billions of data points. Its architecture employs techniques like [binary quantization](https://qdrant.tech/articles/binary-quantization/) and [scalar quantization](https://qdrant.tech/articles/scalar-quantization/) for efficient storage and retrieval. When you inject FastEmbed’s CPU-first design and lightweight nature into this equation, you end up with a system that can scale seamlessly while maintaining low latency.
+Qdrant is designed to handle large-scale datasets with billions of data points. Its architecture employs techniques like [binary quantization](/articles/binary-quantization/) and [scalar quantization](/articles/scalar-quantization/) for efficient storage and retrieval. When you inject FastEmbed’s CPU-first design and lightweight nature into this equation, you end up with a system that can scale seamlessly while maintaining low latency.
 
 ## Summary
 

qdrant-landing/content/articles/hybrid-search.md

@@ -132,7 +132,7 @@ into account. Those models are usually trained on clickstream data of a real app
 very business-specific. Thus, we'll not cover them right now, as there is a more general approach. We will
 use so-called **cross-encoder models**.
 
-Cross-encoder takes a pair of texts and predicts the similarity of them. Unlike embedding models, 
+Cross-encoder takes a pair of texts and predicts the similarity of them. Unlike [embedding models](/articles/fastembed/), 
 cross-encoders do not compress text into vector, but uses interactions between individual tokens of both
 texts. In general, they are more powerful than both BM25 and vector search, but they are also way slower.
 That makes it feasible to use cross-encoders only for re-ranking of some preselected candidates.

qdrant-landing/content/articles/langchain-integration.md

@@ -31,8 +31,8 @@ provides unified interfaces to different libraries, so you can avoid writing boi
 It has been reported millions of times recently, but let's say that again. ChatGPT-like models struggle with generating factual statements if no context 
 is provided. They have some general knowledge but cannot guarantee to produce a valid answer consistently. Thus, it is better to provide some facts we 
 know are actual, so it can just choose the valid parts and extract them from all the provided contextual data to give a comprehensive answer. [Vector database, 
-such as Qdrant](https://qdrant.tech/), is of great help here, as their ability to perform a [semantic search](https://qdrant.tech/documentation/tutorials/search-beginners/) over a huge knowledge base is crucial to preselect some possibly valid 
-documents, so they can be provided into the LLM. That's also one of the **chains** implemented in [LangChain](https://qdrant.tech/documentation/frameworks/langchain/), which is called `VectorDBQA`. And Qdrant got 
+such as Qdrant](https://qdrant.tech/), is of great help here, as their ability to perform a [semantic search](/documentation/tutorials/search-beginners/) over a huge knowledge base is crucial to preselect some possibly valid 
+documents, so they can be provided into the LLM. That's also one of the **chains** implemented in [LangChain](/documentation/frameworks/langchain/), which is called `VectorDBQA`. And Qdrant got 
 integrated with the library, so it might be used to build it effortlessly.
 
 ### The Two-Model Approach

qdrant-landing/content/articles/memory-consumption.md

@@ -215,7 +215,7 @@ But let's first see how much RAM we need to serve 1 million vectors and then we
 
 ### Vectors and HNSW graph stored using MMAP
 
-In the third experiment, we tested how well our system performs when vectors and [HNSW](https://qdrant.tech/articles/filtrable-hnsw/) graph are stored using the memory-mapped files.
+In the third experiment, we tested how well our system performs when vectors and [HNSW](/articles/filtrable-hnsw/) graph are stored using the memory-mapped files.
 Create collection with:
 
 ```http
@@ -355,7 +355,7 @@ Which might be an interesting option to serve large datasets with low search lat
 
 ## Conclusion
 
-In this article, we showed that Qdrant has flexibility in terms of RAM usage and can be used to serve large datasets. It provides configurable trade-offs between RAM usage and search speed. If you’re interested to learn more about Qdrant, [book a demo today](https://qdrant.tech/contact-us/)!
+In this article, we showed that Qdrant has flexibility in terms of RAM usage and can be used to serve large datasets. It provides configurable trade-offs between RAM usage and search speed. If you’re interested to learn more about Qdrant, [book a demo today](/contact-us/)!
 
 We are eager to learn more about how you use Qdrant in your projects, what challenges you face, and how we can help you solve them.
 Please feel free to join our [Discord](https://qdrant.to/discord) and share your experience with us!

qdrant-landing/content/articles/multitenancy.md

@@ -20,7 +20,7 @@ keywords:
 
 We are seeing the topics of [multitenancy](/documentation/guides/multiple-partitions/) and [distributed deployment](/documentation/guides/distributed_deployment/#sharding) pop-up daily on our [Discord support channel](https://qdrant.to/discord). This tells us that many of you are looking to scale Qdrant along with the rest of your machine learning setup. 
 
-Whether you are building a bank fraud-detection system, [RAG](https://qdrant.tech/articles/what-is-rag-in-ai/) for e-commerce, or services for the federal government - you will need to leverage a multitenant architecture to scale your product.
+Whether you are building a bank fraud-detection system, [RAG](/articles/what-is-rag-in-ai/) for e-commerce, or services for the federal government - you will need to leverage a multitenant architecture to scale your product.
 In the world of SaaS and enterprise apps, this setup is the norm. It will considerably increase your application's performance and lower your hosting costs. 
 
 ## Multitenancy & custom sharding with Qdrant

qdrant-landing/content/articles/neural-search-tutorial.md

@@ -92,7 +92,7 @@ Transformers is not the only architecture suitable for neural search, but for ou
 
 We will use a model called `all-MiniLM-L6-v2`.
 This model is an all-round model tuned for many use-cases. Trained on a large and diverse dataset of over 1 billion training pairs.
-It is optimized for low memory consumption and fast inference.
+It is optimized for low [memory consumption](/articles/memory-consumption/) and fast inference.
 
 The complete code for data preparation with detailed comments can be found and run in [Colab Notebook](https://colab.research.google.com/drive/1kPktoudAP8Tu8n8l-iVMOQhVmHkWV_L9?usp=sharing).
 

qdrant-landing/content/articles/new-recommendation-api.md

@@ -24,7 +24,7 @@ Here, we'll discuss some internals and show how they may be used in practice.
 ### Recap of the old recommendations API
 
 The previous [Recommendation API](/documentation/concepts/search/#recommendation-api) in Qdrant came with some limitations. First of all, it was required to pass vector IDs for 
-both positive and negative example points. If you wanted to use vector embeddings directly, you had to either create a new point 
+both positive and negative example points. If you wanted to use [vector embeddings](/articles/what-are-embeddings/) directly, you had to either create a new point 
 in a collection or mimic the behaviour of the Recommendation API by using the [Search API](/documentation/concepts/search/#search-api).
 Moreover, in the previous releases of Qdrant, you were always asked to provide at least one positive example. This requirement 
 was based on the algorithm used to combine multiple samples into a single query vector. It was a simple, yet effective approach.