InfiniFlow

RAGFlow v0.26.0 is here! 🚀 We’ve just shipped a major update focused on smoother workflows, easier integrations, and faster performance. What’s new: 🔹 Expanded connectors: Instantly plug into your stack with native support for Outlook, OneDrive, Microsoft Teams, Slack, SharePoint, Salesforce, and Azure Blob Storage. 🔹 Effortless model management: Stop typing model names manually. We now offer auto-populating dropdowns for 10+ providers (Ollama, OpenRouter, vLLM, and more) plus support for multiple API keys. 🔹 GraphRAG optimization: New checkpoint & resume features for community extraction and entity resolution—save time and compute on your heaviest indexing tasks. 🔹 Faster reasoning: We’ve removed <think> text buffering, making your interactions with reasoning models feel instantaneous. 🔹 New models: Added support for the latest Anthropic, SiliconFlow, Voyage 4, MiniMax-M3, and Cohere models. 🔹 Localization (i18n): RAGFlow now includes full Korean and Italian translations. See what’s new and upgrade today: https://lnkd.in/gak6YJDe #RAGFlow #GenerativeAI #GraphRAG #LLM #OpenSource #DataEngineering

🚀 RAGFlow v0.25.6 is officially live! This release brings upgrades to autonomous agent capabilities and retrieval performance. Highlights: 🌐 Autonomous AI Browser: A new component enabling agents to autonomously navigate and interact with web pages. ⚡️ Stabilized RAPTOR (Ψ-RAG): Faster index construction and improved Recall@5/F1 scores, eliminating previous semantic loss. 🛠️ Streamlined Dev Tools: Use the new lightweight @tool decorator for quick Python function registration. 🔌 Smarter API: Send only the latest message to the /chat/completions endpoint—no need to transmit the full chat history! 🇫🇷 Bonjour! The interface is now 100% localized in French. We’ve also included crucial stability improvements and bug fixes for Mistral reasoning models and asynchronous tasks. Upgrade here:👇https://lnkd.in/gak6YJDe #RAGFlow #AI #ArtificialIntelligence #AIAgents #RAG #LLM #MachineLearning #Python #OpenSource #DeveloperTools #Mistral

🚀 RAGFlow v0.25.5 is live! This release focuses heavily on core performance improvements, lighter resource consumption, and streamlined admin workflows. Release highlights: ⚡ Elasticsearch: Accelerates retrieval by removing unnecessary vector fetches during the main search phase, reducing latency by 50–100%. ⚡ Infinity engine: Pushes metadata filters down to the document engine, significantly improving retrieval performance. 🛠️ Sandbox config: Introduces local and SSH provider options directly within the admin interface, removing the need to edit environment variables. 📉 Core performance: Improves server startup speed and overall memory usage. 📊 Observability: Adds Langfuse token usage reporting. 🐍 Infrastructure: Bumps minimum supported Python version to 3.13. Check out the full release notes and upgrade here: https://lnkd.in/gak6YJDe #RAGFlow #opensource #LLM #vectordatabase #AIinfrastructure

🚀 RAGFlow v0.25.4 is live! 🚀 We’re making enterprise RAG more flexible, secure, and cost-efficient. Here are the highlights: 🔌 Generic RESTful Connector: Easily ingest data from any niche or enterprise platform. ☁️ S3 cost savings: New ETag-based incremental sync drastically reduces sync time and AWS egress costs. 🤖 GPT-5.4 support: Now supporting gpt-5.4-mini and gpt-5.4-nano. 📊 Visual Agents: The Code component now displays charts and images directly in the chat. 🔒 Enhanced security: Hardened sandboxes and JWT-principal binding for API messages. ⚡ Performance: Faster ingestion server boot times and optimized resource handling. Check out the full release on GitHub: 👉 https://lnkd.in/gak6YJDe #RAGFlow #OpenSource #AI #EnterpriseAI #LLM #DevOps

RAGFlow v0.25.2 is out! 📦 ✅ Better APIs: Continued migration to RESTful standards. ✅ Faster Queries: Metadata filtering now leverages Elasticsearch, clearing performance bottlenecks. ✅ Sync Integrity: New snapshot mechanism ensures deleted files are tracked across 8+ data sources. ✅ Stability: Fixed metadata visibility and chat output bugs. Upgrade now to experience a smoother, more efficient RAGFlow. Full changelog: https://lnkd.in/gb6kJjuj #RAGFlow #AI #LLMs #IngestionPipeline

🔥 RAGFlow just reached 80,000 GitHub stars! ⭐ A huge thank you to our contributors, users, developers, and community around the world for helping us reach this milestone. It’s incredible to see what people are building with RAGFlow — from enterprise RAG systems to next-generation AI agents and data platforms. 80K stars is not the finish line. It’s just the beginning. 🚀 Thank you for building with us. #OpenSource #AI #RAG #Agents #LLM #Harness

🔥 RAGFlow just reached 80,000 GitHub stars! ⭐ A huge thank you to our contributors, users, developers, and community around the world for helping us reach this milestone. It’s incredible to see what people are building with RAGFlow — from enterprise RAG systems to next-generation AI agents and data platforms. 80K stars is not the finish line. It’s just the beginning. 🚀 Thank you for building with us. #OpenSource #AI #RAG #Agents #LLM #Harness

🚀 RAGFlow in the Agent Era Read: https://lnkd.in/gKaJtTzi We’re seeing a clear shift in agent systems: 👉 LLM + harness = stateless “Brain” 🧠 👉 Data = stateful foundation 🗄️ In this architecture, retrieval is no longer just search—it becomes the core data layer for agents, powering memory, pipelines, and context. This also changes requirements dramatically: 📈 10–100x more queries 🔁 Longer, multi-step retrieval chains ⚠️ New challenges in caching, relevance, and stopping strategies RAGFlow is evolving into a Context Engine for agents to meet these demands. 💬 Curious to hear from others: what’s the biggest challenge you’re facing in agent retrieval today?

RAGFlow v0.25.1 is officially here: 🙌 - 💥 Engineering optimization: Standardizes our web APIs to RESTful conventions, unifying document creation and indexing flows while maintaining backward compatibility. - 💥 RAG optimization: New lazy loading and chunked parsing for large PDFs, significantly reducing memory footprint for massive ingestions. - 💥 Improved flexibility: Adds OpenDataLoader PDF parser backend as a new backend option alongside DeepDOC, MinerU, Docling, TCADP, and PaddleOCR. - 💥 Growing ecosystem: Now supports DeepSeek v4 and adds UCloud as a model provider. Details 👉: https://lnkd.in/gqgRbBXv #RAG #DeepSeek #OpenDataLoader #UCloud

🚀 We’ve just released a new introduction to RAGFlow on YouTube. In this video, our backend engineer Idriss Sbaaoui introduces RAGFlow and walks through: 🧠 Why AI agents hallucinate without proper data grounding ⚠️ Key challenges in building robust RAG systems 🔍 How RAGFlow approaches it differently ✨ Our C-S-R-G pipeline: from ingestion to intelligent generation RAGFlow is built to strengthen the context foundation for AI systems—so agents can reliably retrieve, reason, and act using real data. 🎥 Watch here: https://lnkd.in/gEsM5wHV #AI #LLMs #RAG #Agents