1.8.1 1.9.0
langgenius/difySep 3, 2025 → Sep 22, 202519 days between releases
What's New
Dify 1.9.0 adds Knowledge Pipeline and Queue-based Graph Engine, requiring a datasource credential migration.
Breaking Changes
- Existing datasource credentials must be transformed by running `uv run flask transform-datasource-credentials` after upgrade.
New Features
- Knowledge Pipeline provides visual, node-based orchestration for document ingestion and processing.
- Queue-based Graph Engine adds unified scheduling, flexible start points, streaming coordination, and command processing.
- Pipeline templates and DSL import/export improve reuse and collaboration.
- Plugin-based customizable data sources and tools include a Q&A Processor for ingestion.
- Image extraction and retrieval are supported in knowledge pipelines.
Improvements
- Knowledge ingestion is refactored into a modular, extensible pipeline architecture.
- One-click migration converts legacy knowledge bases to the Knowledge Pipeline architecture.
- Queue-based execution makes workflows more robust and controllable.
Upgrade Notes
Run `uv run flask transform-datasource-credentials` after upgrading to convert existing datasource credentials. Use the one-click migration to move legacy knowledge bases to the Knowledge Pipeline and validate pipeline behavior before production rollout. The destructive migration noted for v2.0.0-beta users does not apply when upgrading directly from 1.8.1.
The Two Releases
v1.8.1
Sep 3, 2025· 1 year ago
Dify 1.8.1 focuses on stability, performance, and developer experience, adding DSL export from history, downvote reasons, multimodal filename support, OAuth provider support, and slash commands. It also includes critical database session fixes, an XSS fix, SQLAlchemy 2.0 migration, and Indonesian language support.
Full release notes1.9.0 – Orchestrating Knowledge, Powering Workflows
Sep 22, 2025· 1 year ago
Dify 1.9.0 introduces the Knowledge Pipeline for visual, node-based document ingestion and the Queue-based Graph Engine for more robust workflow execution. The release requires a datasource credential migration and includes templates, DSL import/export, image extraction, test-run debugging, and engine command support.
Full release notes