ragflow v0.19.0 is released! Cross-language search, new Agent code components, and image direct display functions are fully upgraded!

With the rapid development of AI technology and knowledge management, Ragflow, as an open source RAG (Retrieval-Augmented Generation) process management platform, has once again ushered in a strong upgrade - version v0.19.0! This update covers major features such as cross-language search, multi-language environment adaptation, new Agent components, image display innovation, and integration of leading AI models. At the same time, a large number of details have been fixed, improving all aspects of developer and user experience. This article will cut into the core new features, function optimization, community contributions, and application scenarios, and deeply analyze the highlights of this version upgrade, so that you can fully understand the technical charm and future value of Ragflow v0.19.0!

1. Version time and introduction

• •  Version number:  v0.19.0
• •  Release date:  May 26, 2025
• •  Official repository: github.com/infiniflow/ragflow ^[1] 
• •  Positioning:  AI knowledge retrieval and dialogue management platform for RAG method processes, focusing on multi-language, multi-modal, and multi-engine integration, helping to build an efficient intelligent question-and-answer and knowledge service system.

2. Detailed explanation of the core new features of Ragflow v0.19.0

1. Cross-language search support - seamless query experience for multilingual knowledge bases

• •  Scenario pain points:  In real business scenarios, knowledge bases often integrate content in multiple languages ​​such as Chinese and English. Traditional searches are difficult to accurately match information in complex contexts, affecting user experience and information utilization.
• •  This improvement:  Ragflow adds support for cross-language search in the Knowledge module and Chat module, greatly improving the search accuracy and recall effect of multi-language hybrid knowledge bases.
• •  Application advantages:  For example, the Chinese-English bilingual knowledge base can achieve more intelligent cross-language question and answer, making business globalization more convenient.

2. Agent adds Code component - Python and JavaScript scripts support new development heights

• •  Function description:  A new “Code” module specifically for code execution has been added to the Agent component, which supports dynamic running of Python and JavaScript scripts.
• •  Technical value:  Enables developers to embed complex data processing logic, algorithm calculations or business customization operations into the process, no longer limited to fixed interface calls.
• •  Developer benefits:  It provides flexible customization capabilities to achieve smarter and more efficient Agent task orchestration and execution, such as automatic data cleaning, dynamic variable calculation, asynchronous task scheduling, etc.

3. Direct display of pictures in chat and search modules - breaking the "link" barrier and making a great leap forward in visual experience

• •  Change:  Chat and Search modules support direct rendering and display of images in response content, saying goodbye to copying and pasting Image URL references.
• •  Knowledge retrieval test upgrade:  images can be directly presented as search results without being extracted from text.
• •  User experience:  Greatly improves the sense of interactivity and efficiency of information acquisition, especially suitable for scenarios containing multimodal content such as technical documents, report screenshots, product images, etc.

4. Integrate the most cutting-edge large models in the industry - support Claude 4 and ChatGPT o3 reasoning models

• •  Background:  Claude 4 and ChatGPT o3 are both the industry-leading natural language understanding and generation models.
• •  Version support:  Ragflow v0.19.0 natively supports these two advanced models, making it easy for developers to quickly build intelligent question-and-answer, decision-making support and other applications based on the latest large models.
• •  Competitive advantage:  Embrace the latest AI models, speed up product iteration, and enhance intelligent interaction effects.

5. Community contributions are now available: many features are implemented by community developers

• •  Agent tool calling capability:  Supports calling tools in Agent through Generate component, which provides rich expansion capability scenarios. Thanks to notsyncing for his contribution.
• •  Markdown images are displayed in blocks after rendering:  Images in the document can be displayed in blocks, thanks to Woody-Hu.
• •  Document engine OpenSearch support:  Support OpenSearch as the Ragflow document engine to improve document retrieval efficiency, thanks to pyyuhao.

3. Summary of optimization and repair highlights

Ragflow v0.19.0 is not just a “new feature”, it also fixes a lot of detailed bugs in previous versions, improving system stability and user experience, including but not limited to:

• •  Support for batch deletion of files, more efficient management
• •  Visualization of knowledge base document parsing status
• •  Improved file deletion and renaming functions
• •  Optimize API interface to improve search speed and accuracy
• •  Fix permissions and data isolation issues for multi-tenant knowledge sharing
• •  UI interface adjustment, including style unification and interactive details beautification
• •  Improve the stability and user-friendliness of multiple assistants and dataset configurations
• •  Improve OAuth2 and OpenID Connect single sign-on integration and support third-party login

4. In-depth interpretation of functions

1. Technical principles and advantages of cross-language search mechanism

Cross-language search relies on multi-language vector representation to enhance the coverage of query indexes. Ragflow internally calls the underlying multi-language model to uniformly map Chinese and English content to the vector space, thereby achieving seamless matching between Chinese and English. In the past, Chinese input could only retrieve Chinese documents, and English input could only match English documents. Now, bilingual knowledge bases can be cross-dispatched, making it easier to meet the needs of multi-cultural and multi-scenario enterprises.

This feature helps companies break through domestic and international knowledge barriers and achieve global knowledge management.

2. What does Agent Code mean to developers?

In the past, Agent interactions were limited by predefined API tools, which lacked flexibility when faced with complex business logic. The newly added Code component allows Python and JavaScript scripts to be directly embedded, injecting programming capabilities into the process, greatly enhancing the intelligence of the "process robot". For example, format conversion before data is stored, dynamic construction of complex query conditions, and even external asynchronous calls and parallel computing processing can all be completed freely through scripts.

This makes Ragflow not only a retrieval tool, but also a business automation hub.

3. Direct display of image content improves user information acquisition efficiency

Previously, for documents returned by knowledge search and images involved in conversations, the normal operation was to attach image links, and users had to open them separately to view them. The new version uses embedded image rendering technology, and images are directly displayed in chats and search results, which greatly facilitates users to quickly understand the content. In particular, in document interpretation, technical support, and product display, the intuitive display of images is extremely valuable.

At the same time, in the retrieval module, images become direct index blocks, supporting fusion retrieval based on visual data.

4. Integrate Claude 4 and ChatGPT o3 to meet diverse application needs

These two models excel in natural language understanding and generation, respectively, and can handle complex conversations, multi-round interactions, and deep contextual understanding. Ragflow supports plug-and-play of these two new models, helping companies use the latest artificial intelligence technology to quickly build smart assistants, knowledge questions and answers, customer service and other services to stay ahead of the competition.

5. Community power: a model of open source + collaboration

This version highlights the community's activity and contribution depth. From tool calls, engine support to interface rendering, many contributors have devoted themselves to Ragflow. The vitality of open source projects lies in this kind of collaborative innovation. Technology iteration is faster, and innovative functions can be integrated and fed back to users in the first place.

VI. Applicable scenarios and practical suggestions

• •  Knowledge center of multinational enterprises : multilingual data integration, hybrid search
• •  Intelligent customer service system : mixed image and text queries improve answering experience
• •  R&D support platform : Agent code components support script operations, automated testing and data processing
• •  Education and training : Rapid access to multi-modal teaching resources
• •  Data analysis and reporting system : dynamic script processing, multi-source data fusion retrieval

7. How to upgrade and best practices

The upgrade steps are simple and easy to follow, and the documentation is detailed. Users are advised to pay attention to the following aspects:

• • Configure appropriate multilingual models and indexes for cross-language search
• • Reasonable use of code execution functions in Agent components, paying attention to security and performance management
• • Use the image direct display function to enrich the diverse forms of expression of knowledge base content
• • Combined with the latest large model interface, optimized intelligent question and answer effect
• • Clean and rebuild indexes in a timely manner to ensure data consistency