InfiniFlow’s cover photo
InfiniFlow

InfiniFlow

Information Services

上海, 上海 345 followers

About us

AI Native Infrastructure

Industry
Information Services
Company size
11-50 employees
Headquarters
上海, 上海
Type
Partnership
Founded
2023

Locations

Employees at InfiniFlow

Updates

  • View organization page for InfiniFlow

    345 followers

    Great AI applications need a strong data foundation. We’re excited to see SereneDB integrated with RAGFlow, giving developers another powerful option for building scalable, self-hosted RAG applications with hybrid search. A strong and open ecosystem is essential to making RAG infrastructure more flexible, production-ready, and future-proof. Great to have SereneDB as part of the growing RAGFlow ecosystem! 🚀 #RAG #AI #OpenSource #RAGFlow #SereneDB #VectorDatabase #AIInfrastructure #EnterpriseAI #LLM #Agents

    When choosing your AI project infrastructure, focus on how 𝐟𝐮𝐭𝐮𝐫𝐞-𝐩𝐫𝐨𝐨𝐟 it is. Let me explain. In the end, every AI product depends on a handful of foundational choices: the model, the framework and the store holding the data. Get the storage layer wrong and it blocks your operations 𝑒𝑥𝑎𝑐𝑡𝑙𝑦 𝑤ℎ𝑒𝑛 𝑦𝑜𝑢 𝑐𝑎𝑛 𝑙𝑒𝑎𝑠𝑡 𝑎𝑓𝑓𝑜𝑟𝑑 𝑖𝑡. Here is why it is important: ◾️ A niche vector store looks fine in a demo, then turns into a migration project once you need to scale. ◾️ Infrastructure built into tools your team already uses needs 𝑛𝑜 𝑠𝑝𝑒𝑐𝑖𝑎𝑙𝑖𝑠𝑡 ℎ𝑖𝑟𝑒 to support it. ◾️ A wider ecosystem means more people finding and fixing problems before you hit them. ◾️ The bigger your product gets, the more that foundation pays off, instead of holding you back. Today, SereneDB integrates with LangChain and 𝐑𝐀𝐆𝐟𝐥𝐨𝐰 by InfiniFlow, two of the most widely used toolkits for building AI products on a company's own data. Plug SereneDB in as the vector store and you are choosing infrastructure that 𝐠𝐫𝐨𝐰𝐬 𝐰𝐢𝐭𝐡 𝐚𝐧 𝐞𝐜𝐨𝐬𝐲𝐬𝐭𝐞𝐦, not one you end up maintaining alone. We are proud to work with the tools so many builders already trust, and these integrations are how we make sure that trust reaches the data layer too. What you get through these integrations: 🔹 𝐇𝐲𝐛𝐫𝐢𝐝 𝐬𝐞𝐚𝐫𝐜𝐡: keyword and semantic matching combined, so exact terms like product codes are never missed 🔹 Fast filtering by category, date or customer, built into the search itself 🔹 One system for both your search index and your application data, n̲o̲ ̲s̲e̲p̲a̲r̲a̲t̲e̲ ̲v̲e̲c̲t̲o̲r̲ ̲d̲a̲t̲a̲b̲a̲s̲e̲ ̲t̲o̲ ̲m̲a̲i̲n̲t̲a̲i̲n̲ 🔹 𝐒𝐞𝐥𝐟-𝐡𝐨𝐬𝐭𝐞𝐝, so your documents and embeddings 𝑛𝑒𝑣𝑒𝑟 𝑙𝑒𝑎𝑣𝑒 𝑦𝑜𝑢𝑟 𝑜𝑤𝑛 𝑖𝑛𝑓𝑟𝑎𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑒 🔹 𝐍𝐨 𝐒𝐐𝐋 to write, the integration handles schema, indexing and queries 𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑦𝑜𝑢 𝑎𝑠𝑘𝑖𝑛𝑔 𝑓𝑜𝑟 𝑡ℎ𝑒𝑚 #RAG #AI #LangChain #RAGflow #OpenSource #DeveloperTools

    • No alternative text description for this image
  • View organization page for InfiniFlow

    345 followers

    ⭐ RAGFlow has reached 90,000 stars on GitHub! Another exciting milestone for RAGFlow — and, more importantly, for the community that continues to build, contribute, experiment, and grow with us. From RAG to context. From documents to knowledge. RAGFlow continues to evolve as an open-source platform for turning complex data into high-quality context and knowledge for AI applications. A huge thank you to every contributor, developer, user, partner, and supporter who has been part of this journey. 90K stars is a milestone we share with all of you. And we’re just getting started. Next stop: 100K. 🚀 🔗 https://lnkd.in/eRumubHP #RAG #AI #OpenSource #GitHub #Agent #AgenticAI

    • No alternative text description for this image
  • 🚀 RAGFlow v0.27.1 is now available — with more integrations, broader model support, and improved reliability. Here’s what’s new: 🌐 More Connectivity Connect more data and search sources with the new Azure DevOps data source connector and You.com and Serply web search providers. 🤖 More Model Support Meet Synthorai, a new model provider, along with additional DeepSeek models. ✨ More Improvements Enhanced retrieval controls, faster metadata filtering, better Chat settings, and improved knowledge base validation. 🛠️ Better Reliability A broad set of fixes across Chat, PDF parsing, Wiki, Knowledge Compilation, OCR, MinerU, data source connectors, and more. More connected. More capable. More reliable. 🚀 Upgrade to RAGFlow v0.27.1 and explore what’s new. 🔗 Release notes: https://lnkd.in/g9fijgEJ #RAG #Agent #AI #LLM

  • 🧩 From retrieving chunks to retrieving knowledge. Traditional RAG retrieves fragments based on similarity. RAGFlow Knowledge Compilation changes what RAG can retrieve — turning fragmented data into structured, connected knowledge before retrieval. 📖 Wiki · 🌳 Tree · 🕸️ Graph · ⏱️ Timeline · 📑 Page Index · 💡 Mind Map 🔗 Instead of retrieving isolated chunks, RAG can retrieve knowledge with structure, relationships, hierarchy, and context — giving LLMs and agents a richer foundation for understanding and reasoning. ✨ From fragments to understanding. ⭐ GitHub: https://lnkd.in/gak6YJDe #KnowledgeCompilation #RAG #LLMWiki #AgenticAI

    • No alternative text description for this image
  • View organization page for InfiniFlow

    345 followers

    🚀 RAGFlow v0.27.0 is here. This major release brings three big upgrades: 🧠 Knowledge Compilation Turn documents and datasets into structured, reusable knowledge — including Wiki, Graph, Tree, Page Index, Mind Map, Timeline, and Skills. 🤖 Agentic RAG Four thinking modes — Low · Medium · High · Ultra — for different levels of reasoning. ⚙️ Revamped Model Provider System Easier model configuration and management, with support for more models and providers. Also included: 🔌 SereneDB & GaussDB support 📦 UCloud & Tenki sandbox support 🌐 Querit web search 📄 Mistral OCR 🎙️ FunASR / SenseVoice 🛠️ Major stability and parsing improvements A huge thank you to all RAGFlow contributors! ❤️ From RAG to context. From documents to knowledge. Please check out: https://lnkd.in/g9fijgEJ #AgenticRAG #KnowledgeCompilation #AI #Agent #OpenSource

  • 👏 Querit Contents is now available in RAGFlow. Building on the existing Search integration, RAGFlow users can now retrieve clean full-page content from web pages, PDFs, and papers for RAG workflows. Sign up through RAGFlow to claim exclusive Querit credits.

    • No alternative text description for this image
  • 🚀 Querit Search is now integrated into RAGFlow The integration brings live web search to RAGFlow Agents and Canvas workflows. Highlights 👇  🔎 Dedicated Querit Search node in Canvas  🤖 Built-in search tool for RAGFlow Agents  🎛️ Site, time range, country, and language filters  📄 Structured results for downstream workflow steps Use Querit to add current web context to research agents, monitoring workflows, knowledge assistants, and more. Thanks to the Querit AI team for the contribution! 🙌 Learn more 👉 RAGFlow: https://ragflow.io/ 👉 Querit: https://www.querit.ai/  👉 Integration PR: https://lnkd.in/g2gJAG8v #RAGFlow #Querit #RAG #AIAgents #WebSearch

  • 🚀 RAGFlow v0.26.4 is live! ✨ Highlighting NLP tokenization: 🌐 New language-aware Snowball stemmer supporting 16 languages! 🔗 Dataset language params now integrated across the tokenization pipeline 🇳🇱 Dutch added to the frontend 🐛 Plus under-the-hood bug fixes! #RAGFlow #NLP #AI 👉 https://lnkd.in/gak6YJDe

  • Nailed it! 🔨 Thanks Parth Kalkar for highlighting the crucial difference deep document understanding makes in production RAG pipelines. At InfiniFlow, we believe enterprise AI needs structural context to succeed—complete transparency and zero guesswork. That's why RAGFlow is open-source and constantly evolving to tackle the hardest document extraction challenges out there. Stop ripping tables in half! Check out the original post below to see why ingestion pipelines make or break your AI. 👇

    Blindly chunking your documents by character count destroys context before the model even sees them. Slicing a complex, 100-page enterprise PDF into arbitrary 512-token chunks rips tables in half, orphans headers from their paragraphs, and scrambles financial data. When your AI hallucinates a critical business metric, it isn’t a model problem. It is your garbage ingestion pipeline. If you want to build RAG that actually works in production without hallucinating, you have to stop treating documents like raw text strings. You need *RAGFlow* by InfiniFlow. * The Value:  RAGFlow is an open-source RAG engine built entirely on *deep document understanding*. Instead of blind token-splitting, it uses layout recognition models to understand the actual visual structure of the file (PDFs, Word docs, Excel, PPTs).  It identifies tables, charts, and sections, keeping semantic blocks completely intact during the chunking phase. * The Leverage:  You stop wasting time trying to patch hallucinations with massive, expensive prompts. By feeding your vector database structurally perfect chunks, your retrieval accuracy skyrockets.  Even better, RAGFlow provides traceable citations, so when the model answers, it points the user to the exact visual bounding box in the original document. Complete transparency, zero guesswork. 🔗 I’ve linked the GitHub in the comments below.

    • No alternative text description for this image

Similar pages

Browse jobs