v0.4.0 Release Notes
🦹 Summary
🎉 Refly's strongest update ever! Brand new code component generation/one-click prototype restoration, canvas sharing, custom prompt functionality, greatly improved search speed, and doubled file upload capacity. Building a more powerful AI creation engine to take your creative work efficiency to the next level!
☄️ Improvements for Self-Deployment
- Local support for using Ollama with complete tutorial
- Greatly optimized login experience for local deployment
- Removed subscription and unrelated login components from local deployment, providing a cleaner local deployment experience
- Support for reverse proxy capabilities, painless deployment on cloud services
- Support for local deployment to display and process static resources, including images, files, and other capabilities
- Support for convenient switching of file processing parsers, including pdfjs and marker
- Support for displaying local execution error stacks for easy debugging and problem feedback
🌟 New Features
- Official launch of Code Artifacts (#546)
- Added "Code Artifacts" skill, supporting AI-generated code components with real-time preview of running results
- Support for generation, display, download, and sharing of various mainstream code types including svg, mermaid, html, react
- svg, mermaid, etc. support direct download as images or copying to clipboard for quick sharing
- Code can be used as context for follow-up questions and modifications, or manually modified and optimized with preview of final results
- Mermaid diagrams in Markdown also support image download operation buttons
- Canvas Sharing and Copying (#536, #544, #547)
- All canvases can be shared as public links, viewable in read-only mode
- Support for one-click copying of canvases shared by others, with subsequent modifications and adjustments possible
- Canvas supports export as images
- We welcome everyone to submit canvas use cases, which we will later showcase as official examples with attribution to all Refly users
- Support for Custom Prompts (#533, #539)
- Added "Custom Prompt" skill, allowing arbitrary configuration of system prompts, Temperature, Top P, and other parameters
- Custom prompt execution results support adjusting system prompts and other parameters for re-running
- Support for cloning custom prompt nodes for questions and modifying system prompts before execution
- Custom prompts support follow-up questions with already set prompts, eliminating the need for frequent additions
- Web and Knowledge Base Search Optimization (#523, #533, #538)
- Optimized query preprocessing logic, supporting questions with any canvas context
- Adjusted result generation logic to prioritize answering questions rather than simply referencing search results
- Improved skill execution speed, greatly reducing search freezes or failures
- Enhanced intent understanding capability, significantly optimizing usability for complex questions
- Question Links Support Automatic Parsing (#535, #545)
- Directly paste website links in the input box to ask questions and get answers without importing resources
- After results are generated, you can view link webpage content and import it to the knowledge base with one click
- Supports all skills
- Brand New Website Nodes and Code Nodes (#546)
- Support for adding and rendering custom websites in the canvas, such as Bilibili videos, Medium articles, Github repositories, etc.
- Support for one-click follow-up questions to AI for content summarization of website nodes
- Code nodes can preview SVG, HTML, React, Mermaid, HTML capabilities and support code modifications
- Code nodes can be selected as context with Artifact skills for iterative code optimization
- File Upload Limit Adjustments (#524)
- Significantly increased file upload limits for paid users
- Plus users can upload up to 10MB, Pro users up to 20MB, Max users up to 30MB
- Limits will be gradually relaxed depending on operational conditions
- Support for Custom Node Connection Lines (#522, #529)
- Now you can directly draw lines from free nodes to connect other nodes
- Any connection line can be deleted
- Comments can be written for connection lines
- For AI question nodes, connecting context nodes to question nodes automatically adds context to the question box
- Professional Website Preview Capabilities (#548)
- Support for landing pages to display images/titles and descriptions in social networks
- Support for automatically generating preview images based on canvas status, optimizing canvas listing page display
- Support for Displaying Inference Model Thinking Process (#520, #534)
- Support for streaming display of thinking chains for Claude 3.7 Sonnet Thinking and DeepSeek R1
- Support for collapsing thinking chains
💫 Core Problem Optimization
- Account settings support avatar modification (#556)
- Compact mode turned on by default and automatically maintained in subsequent questions
- Fixed issue of selected skills being lost when re-running skills (#532)
- Fixed lag and even crash issues in canvas due to nested duplicate contexts (#552)
- Memo nodes support text color modification menu (#553)
- Node preview cards support title modification
- Document editor supplemented with multilingual translation (#543)
- Optimized the issue of model names being compressed when too long at the bottom of answer nodes
- Optimized the effect of automatically generating names for empty canvases (#537)
- Fixed collapsing mode follow-up issue; after setting global node collapse or expansion, subsequent nodes all follow this configuration (#546)
- Support for default canvas without automatic layout, fixing node layout chaos issues
- Removed dark mode processing (#531)