Never miss a release that matters
AI-powered summaries of every GitHub release.
AI Summaries
Changelogs condensed into clear, actionable insights.
Always Free
Track up to 5 packages at no cost, forever.
Weekly Digest
A curated summary of every release, delivered weekly.
TL;DR
Dify Agent introduces a Linux sandbox (isolated execution environment) for shell-based LLM agents with a new builder, workflow integration, and a dedicated web app experience.
Breaking
- OpenAI API default changed to Responses (modern API standard) to ensure compatibility with GPT-5.6 models.
New
- Dify Agent (Beta): Features a built-in code execution sandbox, a skill system for tool distribution, and an agent roster for workspace management.
- AI Workflow Generation: Enhanced
⌘Kgenerator with context-aware suggestions and parallelized node configuration. - MCP Protocol Upgrade: Supports version 2025-06-18 (Model Context Protocol) with dynamic HTTP header injection for authentication.
- Web App Customization: Added ability to customize chat input placeholder text.
Before You Upgrade
- If using custom OpenAI API keys, manually switch the API type from Chat Completions to Responses to avoid errors with newer models.
TL;DR
Dify now includes Dify Agent (a Linux sandbox-based agent), enabling a new approach to building and deploying AI agents directly within the platform, and integrating them into existing workflows and web applications.
Breaking
- Database Migrations Required: This release includes new database migrations that must be run during the upgrade process.
- Environment Variables Changed: Review and update your
.envfile to reflect changes to required environment variables. - Docker Compose Updates: If you use a customized
docker-compose.yaml, review the changes and carefully re-apply your local customizations.
New
- Dify Agent Builder: Create agents through the UI with customizable prompts, skills, files, tools, and knowledge.
- Workflow Integration: Use Dify Agents directly within Dify Workflows (automated task sequences).
- Web App Publishing: Publish Dify Agents as standalone web applications.
Fixes Worth Knowing
None
Before You Upgrade
- Back up your
docker-compose.yamland.envfiles before updating if you have customized them. - Run database migrations after updating the code to ensure compatibility.
TL;DR
Dify now includes critical security patches to prevent code injection and cross-site scripting (XSS) attacks, especially important if you expose the service to the internet and use external WebApps.
Breaking
- SVG Rendering Disabled: SVGs are no longer rendered in messages as a security precaution. (SVG: Scalable Vector Graphics - image format)
Fixes Worth Knowing
- XSS Vulnerability Resolved: Fixed a security flaw that could allow malicious scripts to run when viewing SVGs.
Before You Upgrade
If you use the Community Edition and expose Dify to the internet with external WebApps, upgrade immediately. Ensure SERVICE_API_URL and FILES_URL are on different domains for optimal security.
TL;DR
Dify now supports more LLMs (like Llama 3.2 and Qwen 2.5) and storage options (Baidu OBS), plus enhancements to the Prompt Editor and tools like podcast generation and Discord integration, improving overall flexibility and functionality.
Breaking
- Docker Compose Refactor: The
docker-compose.yamlfile has been updated; review and migrate any custom changes. - Message Migration: A data migration updates a large volume of messages; review changes in GitHub pull request #9132 before upgrading.
New
- Expanded LLM Support: Added models including Llama 3.2 and Qwen 2.5 for OpenRouter.
- New Storage Option: Baidu OBS (cloud storage) is now supported.
- Prompt Editor Flexibility: The Prompt Editor now accepts arrays of strings and numbers.
Fixes Worth Knowing
- Chat History: Corrected an issue where chat history wasn’t loading correctly.
- Image Links: Fixed broken links to images generated by the QR code tool when using Huawei OBS.
- Model Information: Resolved missing model information in Langfuse LLM spans.
Before You Upgrade
TL;DR
Dify workflows are significantly improved: variable referencing is now simpler using / to select variables directly, but existing apps will need to be recreated.
Breaking
- Workflow Compatibility: Existing Dify apps are incompatible and must be recreated due to data structure changes.
- Database Migration: Requires manual database updates (SQL provided) when upgrading from
0.6.0-preview-workflow.1to ensure data integrity.
New
- Simplified Variables: Variables in LLM (Large Language Model) and other nodes can now be selected using a
/prefix, streamlining workflow creation. - App Creation UX: Improved user experience when creating new Dify applications.
Fixes Worth Knowing
- Several unspecified issues were resolved, improving overall stability.
Before You Upgrade
- Run SQL Migration: If upgrading from
0.6.0-preview-workflow.1, execute the provided SQL command against your PostgreSQL database. - Recreate Apps: All existing Dify applications must be recreated to take advantage of the new workflow features.