#graphrag | Explore Tumblr posts and blogs

govindhtech · 11 months ago

Text

Spanner’s Standard Edition To Create Solid, intelligent apps

Spanner is the globally consistent, always-on database with almost infinite scalability that powers a plethora of contemporary applications in many industries. It is now the cornerstone of many forward-thinking companies’ data-driven transformations. Google is launching Spanner throughout graph today, along with new features like vector search, full-text search, and more. With the help of these advancements, businesses can create intelligent apps that seamlessly integrate several data models and condense their operational deployments onto a single, potent database.

From Spanner versions, you may choose the Standard Edition, Enterprise, and Enterprise Plus editions with the right set of capabilities to fit your needs and budget.

With Spanner Graph, you can easily and succinctly match patterns, navigate relationships, and filter results across interconnected data for popular graph use cases like fraud detection, community discovery, and personalised recommendations.

Building on Google’s many years of search experience, advanced full-text search offers stronger matching and relevance ranking than unstructured text.

Vector search builds on more than a decade of Google research and innovation in approximate closest neighbour algorithms to facilitate semantic information retrieval, the cornerstone of generative AI applications.

Furthermore, in response to Google’s enterprise customers’ increasingly complicated needs regarding latency, cost, and compliance, Google has improved Spanner’s basic infrastructure.

With geo-partitioning, you can store some of your data in particular places to provide quick local access while still deploying globally.

Data sovereignty needs are respected and multi-regional availability attributes are provided via dual-region arrangements.

By dynamically adjusting the size of your Spanner deployment, auto-scaling frees you from having to overprovision and enables you to respond swiftly to changes in traffic patterns.

Presenting the Spanner editions

Google is pleased to present Spanner editions, a new tiered pricing structure that enables you to benefit from these advancements while improving cost transparency and cost-saving options. You can select the Standard, Enterprise, and Enterprise Plus editions from Spanner editions to match your demands and budget with the appropriate collection of features.

The Standard edition is a feature-rich package that comes with additional features like scheduled backups and reverse ETL from BigQuery in addition to all of the current capabilities that are generally available (GA) in single-region settings. The Enterprise edition leverages multi-model features, such as vector search, enhanced full-text search, and the new Spanner Graph, to enhance the Standard edition.

Moreover, it offers improved data security and operational ease with incremental backups and automated autoscaling. The new geo-partitioning feature and multi-region setups with 99.999% availability are included in the Enterprise Plus version for workloads that demand the greatest levels of availability, performance, compliance, and governance.

Let’s examine each of the Spanner editions in more detail

Standard Edition

Everything you adore about Spanner is present in the standard edition, and then some. Access to Spanner’s single-region configurations, which include transparent failover, synchronous replication, and automatic sharding for almost infinite scalability and 99.99% availability, is provided by the Standard edition. All of the features available in the GA today, including Google’s enterprise-grade security and data protection measures, are included in the standard edition.

Furthermore, Google is adding new features to the Standard edition. Google is growing their offering with additional features like reverse ETL and external datasets with BigQuery, in addition to Spanner’s strong connection with BigQuery, which includes features like Data Boost and BigQuery federation. Google is also added scheduled backups to further its data protection capabilities. These improved features in the Standard edition are available to all Spanner customers.

Customers can gain from Standard edition in the following ways:

Migrate your relational workloads to Spanner to save operational toil, minimise costly scheduled and unplanned downtime, and cut the overall cost of ownership.

Shard MySQL or other complicated application-level sharding using Spanner’s automatic sharding instead.

Spanner’s nativePostgreSQL interface lets you modernise your apps without being locked into outdated features.

Elevate your NoSQL workloads with multirow transactions and stable secondary indexes thanks to support for adapters for Cassandra and DynamoDB that are compatible with APIs.

Avoid complex ETL pipelines and data silos by consolidating your data into a single source-of-truth database and running analytics on up-to-date data with Spanner Data Boost and BigQuery federation.

Utilise reverse ETL to service complicated analytics use-cases from Spanner and BigQuery to compute them at high QPS and low latency.

Enterprise Edition

Expanded data modalities are available to you with the Enterprise edition, enabling you to combine your operational deployments and provide your applications with true multi-model access. Spanner Graph, full-text search, and vector search capabilities are available for your data with the Enterprise edition, in addition to the relational and NoSQL functionality that Spanner currently offers.

The best part is that by combining all of this functionality, Spanner’s primary advantages of high availability, scalability, and consistency can be preserved in a genuine, completely interoperable multi-model database. Having all of your data in one location eliminates the need to transfer data between transactional and speciality databases, allowing you to minimise data silos and duplication, streamline data governance, cut down on operational complexity, and eventually minimise total cost of ownership.

Customers can gain from Enterprise edition in the following ways:

By fusing the advantages of SQL and GQL, you can supercharge your applications with Spanner Graph capabilities at nearly infinite scale. This allows analysts and developers to query linked and structured data in a single action.

Use full-text search to modernise your search apps and streamline your operational pipelines. You may leverage sophisticated search features without copying and indexing data into a separate search solution.

By basing fundamental models on contextual, domain-specific, and real-time data with native vector search capability in your operational data, you may enhance the AI-assisted user experience.

Launch robust multi-model applications, such a generative AI product that grounds LLMs using relational and knowledge graph RAG approaches, or a searchable product catalogue that mixes traditional and semantic information retrieval through the combination of full-text and vector search.

Enterprise Plus Edition

Spanner’s most feature-rich and widely available edition is called Enterprise Plus. Enterprise Plus includes new dual-region configurations in addition to our 99.999% multi-region options, which let you take advantage of Spanner’s unparalleled availability while also adhering to your data sovereignty needs. You may run a truly global deployment with Enterprise Plus and still retain low latency and data locality for parts of your data by utilising geo-partitioning.

Customers can gain from Enterprise Plus edition in the following ways:

By using Spanner’s multi-region settings for application deployment, you can achieve real 0-RTO and 0-RPO deployments along with 99.999% availability, eliminating the need to construct and manage complex disaster recovery solutions.

Use Spanner’s dual-region configurations in places like Australia, Germany, India, and Japan to meet data sovereignty needs without sacrificing availability.

Partition your data to make it always available where you need it, latency-free, and manage your worldwide deployment while preserving data locality.

Increased openness and cost efficiency

Google is also introducing a more straightforward pricing approach with the new Spanner editions, shifting to per-replica charging and separating the expenses of compute and data replication to improve efficiency and cost transparency. Each edition has various cost-saving features, and switching to a new edition is meant to be a smooth process.

Customers in a single region can purchase the standard edition for the same price. Instead than paying multiple prices for every function, customers can choose the Enterprise version, which has a single price point, if they want increased multi-model and search capabilities. They can also take advantage of the price-performance improvements like incremental backups and managed autoscaler.

Even more, the Enterprise Plus edition offers lower compute expenses, separates computation and replication, and does not charge for witness replica storage. Finally, since configurable read-only replicas will be priced the same as read-write replicas for each edition, customers employing these replicas will benefit from significant cost savings. Making the switch to Spanner editions is easy, and you can take advantage of all the improvements Spanner has to offer while cutting prices.

The availability of the Spanner versions is scheduled for September 24, 2024. See the overview page for Spanner versions for further details.

Read more on govindhtech.com

#SpannerStandardEdition #CreateSolid #intelligentapps #SpannerGraph #Vectorsearch #dualregionarrangements #BigQuery #datasecurity #data #dataprotection #ETLservice #datagovernance #graphRAG #TECHNEWS #EnterprisePlus #overviewpagefoRSpannerversions #technology #news #govindhtech

1 note · View note

news786hz · 19 days ago

Text

Build GraphRAG applications using Amazon Bedrock Knowledge Bases

0 notes

ai-news · 19 days ago

Link

In this post, we explore how to use Graph-based Retrieval-Augmented Generation (GraphRAG) in Amazon Bedrock Knowledge Bases to build intelligent applications. Unlike traditional vector search, which retrieves documents based on similarity scores, kn #AI #ML #Automation

#Blockchain #Crypto

0 notes

pulipuli · 3 months ago

Link

看看網頁版全文 ⇨ 雜談：到底要怎麼使用RAGFlow呢？ / TALK: RAGFlow Drained All My Resources https://blog.pulipuli.info/2025/03/talk-ragflow-drained-all-my-resources.html 由於這次RAGFlow看起來又無法順利完成任務了，我還是來記錄一下目前的狀況吧。 ---- # 專注做好RAG的RAGFlow / RAGFlow: Focusing on RAG。 https://ragflow.io/ 在眾多LLM DevOps的方案中，RAGFlow也絕對可以算得上是重量級的那邊。相較於其他方案，RAGFlow一直積極加入各種能夠改進RAG的特殊技術，使得它在RAG的應用方面出類拔萃。 RAGFlow的主要特色包括了： 1. 文件複雜排版分析功能：能夠解讀表格，甚至能分析PDF裡面圖片的文字。 2. 分層摘要RAPTOR。能改善RAG用分段(chunking)切斷資訊的問題。 3. 結合知識圖譜的GraphRAG跟LightRAG。讓回答著重與命名實體，而且還可能找到詞彙之間的隱含關係。 4. 能作為Dify外部知識庫使用。不過，除了第四點之外，要做到前三項功能，目前看起來還有很多問題需要克服。 # 硬體要求 / Hardware Requirements。由於運作RAGFlow會使用OCR來分析文件的排版，記憶體最好是給到16GB之多，硬碟空間也需要準備50GB。這真的是重量級的方案。如果這些準備好的話，要做到分析複雜排版文件的這件事情就不是很難了。只要做到這個程度，RAGFlow就能在回答引用時顯示來源的文件位置。這樣幫助就很大了呢。 # 大量請求的難題 / The Challenge of Numerous Requests。相較於排版分析是RAGFlow組件中的功能，RAPTOR跟Knowledge Graph都要搭配大型語言模型才能解析跟查詢資料。而RAGFlow在處理資料的時候會在短時間內發送大量的API請求，很容易就被rate limit限流。既然直接連接LLM API會因為太多請求而被限流，我就試著改轉接到Dify上，並在API請求的時候加上排隊等候的機制。 Dify裡面雖然可以寫程式碼，但他其實也是在沙盒裡面運作的程式，還是有著不少的限制。其中一個限制就是不能讓我直接修改系統上的檔案。因此如果要在Dify內用程式讀寫資料，用HTTP請求傳送可能是比較好的做法。這些做法花了很多時間調整。調整了老半天，總算能夠讓它正常運作。不過過了一陣子，LLM API連回應沒有反應了。我猜想可能是連接的Gemini API已經超過用量而被禁止吧。 ---- 繼續閱讀 ⇨ 雜談：到底要怎麼使用RAGFlow呢？ / TALK: RAGFlow Drained All My Resources https://blog.pulipuli.info/2025/03/talk-ragflow-drained-all-my-resources.html

#LLM #RAG #RAGFlow #Talk

0 notes

digitalmore · 4 months ago

Text

#IFTTT #DigitalMore.co

0 notes

ericvanderburg · 10 months ago

Text

Architectural Patterns for Enterprise Generative AI Apps: DSFT, RAG, RAFT, and GraphRAG

http://securitytc.com/TCJ35x

0 notes

craigbrownphd · 10 months ago

Text

Enhancing Data Accuracy and Relevance with GraphRAG

Retrieval Augmented Generation (RAG) has revolutionized how we fetch relevant and recent facts from vector databases. However, RAG's capabilities fall short when it comes to connecting facts and understanding the […] The post Enhancing Data Accuracy and Relevance with GraphRAG appeared first on Datafloq. https://datafloq.com/read/enhancing-data-accuracy-relevance-graphrag/?utm_source=dlvr.it&utm_medium=tumblr

#DataScince #ArtificialIntelligence #MachineLearning

0 notes

yzkr · 10 months ago

Text

GitHub - microsoft/graphrag: A modular graph-based Retrieval-Augmented Generation (RAG) system

0 notes

lsblog-2 · 11 months ago

Text

#graphrag #esg sustainability #semantic graph model #esg domains #knowledge graph llm #esg and nlp #graph rag llm

0 notes

govindhtech · 1 year ago

Text

GraphRAG: A New Tool Available On GitHub for Data Finding

Graphrag Microsoft

Microsoft has announced GraphRAG, a novel approach that is intended to outperform conventional Retrieval-Augmented Generation (RAG) methods. This is a huge development for artificial intelligence and data discovery. This development represents a turning point in the field of artificial intelligence and machine learning, providing improved capabilities for businesses that mostly depend on data-driven decision-making.

Overview of Retrieval-Augmented Generation (RAG)

One must first understand the foundations of retrieval-augmented generation in order to fully appreciate the implications of GraphRAG. Retrieval methods and generative models are combined in RAG to enhance AI system performance, especially in information retrieval and natural language processing (NLP) applications. Conventional RAG models produce more accurate and contextually relevant responses by extracting pertinent documents or information from a database.

The Development of RAG

RAG has limits even though it has shown to be useful in many situations. Due to their inability to handle the enormous amount and complexity of data, traditional RAG models frequently result in inaccurate and inefficient retrieval processes. This is where Microsoft’s GraphRAG enters the picture, tackling these issues head-on with a more advanced strategy.

What is GraphRAG?

Graph Retrieval-Augmented Generation, or GraphRAG for short, is a sophisticated artificial intelligence system that uses graph-based data structures to improve retrieval and generation. By incorporating interactions and connections between data points, GraphRAG creates a more comprehensive and integrated framework for information retrieval, in contrast to classic RAG models that just use textual data.

Crucial Elements of GraphRAG

Graph-Based Data Structures: To capture the complex interactions between various types of information, it makes use of graph databases, which represent data in nodes and edges. This makes it possible to retrieve pertinent data more precisely.

Better Contextual knowledge: It’s contextual knowledge is improved by taking into account the relationships between data points. This results in more accurate and pertinent answers to information retrieval tasks.

Scalability and Efficiency: It is built to effectively manage massive amounts of data. Because of its graph-based methodology, which enables quicker development and retrieval procedures, it is appropriate for enterprise-level applications.

Enhanced Accuracy: By adding graph structures, information is retrieved more accurately and there is a lower chance that inaccurate or irrelevant data would be used in the generative process.

How GraphRAG Operates

It is a multi-step method that combines sophisticated generative models with graph-based retrieval. This is an explanation of how it functions:

Building a Graph Database: Using the data at hand, a graph database must be built as the initial stage. Data points are represented by nodes in this database, while relationships between data points are represented by edges.

Data Retrieval: It searches the graph to find pertinent data in response to a query. Through the graph’s connections, the traversal process enables the system to find related information in addition to immediately pertinent data points.

Contextual Analysis: After the data is retrieved, it is examined in relation to the query. The system uses its knowledge of the connections between data points to deliver a response that is more precise and appropriate for the given situation.

Generation: Lastly, a response is produced by the generative model utilising the data that has been retrieved and examined. The response is more accurate and relevant since the graph-based method makes sure it is based on a larger and more connected dataset.

Benefits of GraphRAG Compared to Conventional RAG

It is a better option for data discovery and information retrieval tasks than classic RAG models since it has various advantages over them. The following are some main advantages:

Richer Contextual Understanding: It offers a deeper contextual understanding by integrating relationships between data items, which results in more pertinent and correct responses.

Enhanced Accuracy: By lowering the possibility of retrieving inaccurate or irrelevant data, the usage of graph-based structures improves the system’s overall accuracy.

Scalability: GraphRAG is appropriate for enterprises with significant and complicated data needs because of its capacity to manage enormous datasets effectively.

Faster Retrieval: The system’s efficiency is increased by the graph traversal process, which makes it possible to retrieve pertinent information more quickly.

Applications of GraphRAG in the Real World

With the release of GraphRAG, a multitude of sectors and applications have new opportunities. Here are a few instances of applications for GraphRAG:

Healthcare: GraphRAG can be used to obtain and evaluate patient data, research findings, and available treatments. This information gives medical personnel the precise knowledge they need to make judgements.

Financial Services: By retrieving and analysing market data, investment opportunities, and financial reports using GraphRAG, financial institutions can make faster and more accurate decisions.

Client care: Support agents can provide better client care by using accurate and contextually relevant information from GraphRAG to better handle customer questions and issues.

Research and Development: To enable more thorough and knowledgeable research findings, researchers can use GraphRAG to collect and analyse scientific literature, patents, and research data.

Prospects and Developments for the Future

GraphRAG is a big advancement in the data retrieval and data discovery space as AI and machine learning technology continue to develop. Given Microsoft’s dedication to developing AI capabilities, it is reasonable to anticipate additional improvements and advancements in the future.

Integration with Other AI Systems: By integrating GraphRAG with other AI tools and systems, a more complete and networked AI ecosystem that makes use of several technologies for better performance can be created.

Enhanced Learning Capabilities: GraphRAG’s learning capabilities could be improved in the future, enabling it to adjust and perform better over time in response to fresh data and user interactions.

Increased Industry use: As GraphRAG’s advantages are more widely acknowledged, we may anticipate increased industry use across a range of sectors, which will result in more creative and effective AI technology applications.

In summary

The release of GraphRAG by Microsoft is a noteworthy turning point in the fields of AI and data discovery. GraphRAG provides improved accuracy, efficiency, and contextual knowledge over standard RAG approaches by utilising graph-based data structures and sophisticated generative models. This cutting-edge technology might revolutionise the way businesses access and use data, creating new opportunities for a variety of uses.

GraphRAG has the potential to significantly influence how data discovery and information retrieval are done in the future as it develops and interacts with other AI technologies. Adopting GraphRAG could be a critical strategic move for companies trying to stay ahead in the quickly evolving AI market, giving them a competitive edge in data-driven innovation and decision-making.