dify version history

- Introduced experimental Dify Agent with Linux sandbox support and web app integration - Added builder UI for creating agents and workflow integration - Includes database migrations and environment variable changes - requires careful upgrade - Current release has security limitations - only use with trusted users

- Redesigned landing/onboarding for a smoother first-run experience. - Introduced `difyctl`, a command-line client for running apps and workflows. - Added support for slow, long-running models and richer human-in-the-loop forms. - Fixed security issues and improved session management across services.

## 🚀 What's New in v1.14.2? v1.14.2 is a patch release focused on security hardening, workflow and knowledge reliability, observability fixes, agent groundwork, and deployment/runtime tuning after v1.14.1. ### 🔐 Security and administration - **Tenant-scoped sensitive endpoints** — strengthened tenant isolation for app trace-config endpoints and FilePreview text extraction. Thanks @xr843 in [#35793](https://github.com/langgenius/dify/pull/35793) and [#35797](https://github.com/langgenius/dify/pull/35797). - **Tool credential safety** — restricted default builtin tool credential updates to workspace admins and owners, and cleaned stale tenant tool credentials during `reset-encrypt-key-pair`. Thanks @NeatGuyCoding and @xr843 in [#36264](https://github.com/langgenius/dify/pull/36264) and [#35843](https://github.com/langgenius/dify/pull/35843). ### 🧩 Workflow, HITL, and app runtime - **Workflow execution reliability** — restored tracing after HITL workflow resume, improved workflow run callback tracking, reduced message-update database roundtrips, fixed memory fetches outside Flask context, and closed base64 file lookup sessions correctly. Thanks @Blackoutta, @CodingO

## 🚀 What's New in v1.14.1? v1.14.1 is a patch release focused on security hardening, workflow and knowledge-base stability, deployment cleanup, and continued UI platform migration after v1.14.0. ### 🔐 Security - **Self-hosted `SECRET_KEY` hardening** — Docker deployments no longer rely on a public default key. When `SECRET_KEY` is left empty, the API generates and persists a runtime key through the configured storage backend, while explicitly configured keys continue to work as before. Thanks @laipz8200 in [#36049](https://github.com/langgenius/dify/pull/36049). - **Internal metrics endpoint protection** — `/threads` and `/db-pool-stat` are hardened to avoid unauthenticated exposure of internal runtime and database-pool details. Thanks @orbisai0security in [#35665](https://github.com/langgenius/dify/pull/35665). - **Account and tool isolation** — fixed an IDOR issue in `GET /account/avatar` and scoped builtin-tool default-credential cleanup to the current tenant. Thanks @NeatGuyCoding and @GareArc in [#35771](https://github.com/langgenius/dify/pull/35771) and [#35887](https://github.com/langgenius/dify/pull/35887). - **Dependency security** — upgraded LiteLLM for CVE

## 🚀 What's New in v1.14.0? ### Collaboration <img width="2632" height="1190" alt="img_v3_02117_67a67218-1ae5-4d68-a45b-3ada592f87fg" src="https://github.com/user-attachments/assets/1d4a3660-8a6a-4b29-9766-4d5b5fd9a51b" /> Collaboration allows workspace members to edit the same workflow together, with synced graph updates, online presence, and shared visibility into who is working where. On self-hosted deployments, collaboration is turned off by default. Enable it by setting: ``` ENABLE_COLLABORATION_MODE = true SERVER_WORKER_CLASS = geventwebsocket.gunicorn.workers.GeventWebSocketWorker NEXT_PUBLIC_SOCKET_URL = your deployment’s WebSocket URL (e.g., wss://dify.example.com) ``` For more details, see [Full Documentation](https://docs.dify.ai/en/use-dify/build/workflow-collaboration) ### Human-in-the-loop (HITL) - **Service API for HITL** — programmatic support for human-in-the-loop flows alongside existing console behavior. ### MCP and plugins - **MCP tool metadata** — refresh after updates so the UI stays in sync. - **MCP server URL** — fix double `/v1` that could break OAuth and authorization (404). - **MCP OAuth discovery** — handle malformed JS

## 🚀 What's New in v1.13.3? Our latest patch release, v1.13.3, focuses on stability and correctness across workflow execution, streaming, and knowledge retrieval. Here's a quick look at the most important updates: ### 🎬 New Features - **Workflow Configuration**: Added variable-reference support for model parameters in LLM, Question Classifier, and Variable Extractor nodes by @scdeng in #33082. ### 🛠️ Bug Fixes - **Streaming Reliability**: Fixed `StreamsBroadcastChannel` replay and concurrency issues to keep frontend/backend event delivery stable by @QuantumGhost in #34030 and #34061. - **Workflow Editor Behavior**: Fixed pasted nodes retaining Loop/Iteration metadata and prevented `HumanInput` nodes from being pasted into invalid containers by @majiayu000 and @hjlarry in #29983 and #34077. - **Runtime Execution**: Restored prompt message transformation logic and corrected `max_retries=0` handling for executor-driven HTTP Request execution by @QuantumGhost, @fatelei, and @kurokobo in #33666, #33619, and #33688. - **Knowledge Retrieval**: Preserved citation metadata in web responses, fixed crashes when dataset icon metadata is missing, corrected hit-count query filteri

This patch release fixes several critical regressions and stability issues introduced in v1.13.1, including: - a severe regression in prompt message transformation that caused multiple LLM-plugin invocation failures across LLM-related nodes, including LLM and Question Classifier. - Knowledge Retrieval node execution failures caused by incompatible enum values - improper Weaviate client cleanup --- ## Upgrade Guide > [!IMPORTANT] > If you use custom `CELERY_QUEUES`, make sure `workflow_based_app_execution` is included. > If `ENABLE_API_TOKEN_LAST_USED_UPDATE_TASK=true`, also include `api_token`. > > For background and details, see **⚠️ Important Upgrade Note** and **🔧 Operational Note** above. ### Docker Compose Deployments 1. Back up your customized docker-compose YAML file (optional) ```bash cd docker cp docker-compose.yaml docker-compose.yaml.$(date +%s).bak ``` 2. Get the latest code from the main branch ```bash git checkout main git pull origin main ``` 3. Stop the service. Please execute in the docker directory ```bash docker compose down ``` 4. Back up data ```bash tar -cvf vo

## 🚀 Major Functional Changes ### Data platform and dataset operations - Added **Hologres** as a supported backend option for both vector retrieval and full-text search. - Added Service API endpoints for dataset document downloads: - Batch ZIP download for selected documents. - Signed URL download for single-document original files. ### Workflow and chat experience improvements - **Breaking change:** HITL email content now uses markdown rendering before delivery. Existing email templates that relied on raw/plain behavior may render differently after upgrade. - Draft variables are now user-scoped instead of app-scoped. Historical draft variables will not be available after upgrading. - Added edge context menu support in workflow canvas, including direct edge deletion. - Preserved existing connections when changing node types in workflow editing. - Added configurable send-key behavior (`Enter` vs `Shift+Enter`) for embedded chat input. - Added file payloads in message-end stream responses. - Removed GPT-4-specific hardcoded behavior from default model selection logic. - Added a new `export-app-messages` CLI command to export application messages and rel

## 💥 New Agent x Skills 🥳 `1.14.0-rc1` brings a new agent-building experience: sandboxed runtime, Skill Editor, collaboration, and dynamic variable assembly. preview docs: <https://docs.bash-is-all-you-need.dify.dev/> ## 🌟 Try the Experience Now Build with the new Agent + Skill workflow in minutes. ### 👉 **Experience Now: [Live Demo](https://skills.bash-is-all-you-need.dify.dev/)** Not sure where to begin? Use our ready-made operational templates! ### 🧩 Start from Templates (Recommended) 👉 [Templates](https://marketplace.dify.ai/templates/operations) ## ✨ Highlights - **New Agent runtime** with sandboxed execution and Agent Mode. ([Agent docs](https://docs.bash-is-all-you-need.dify.dev/en/use-dify/nodes/agent) · [Runtime docs](https://docs.bash-is-all-you-need.dify.dev/en/use-dify/build/runtime)) - **Skill Editor** helps you build reusable SOP blocks and call tools inline with `@` (e.g., `@send_email`). ([File system docs](https://docs.bash-is-all-you-need.dify.dev/en/use-dify/build/file-system)) - **Command + Upload File to Sandbox** make runtime operations practical and controllable. ([Command docs](https://docs.bash-is-all-you-need.dify.dev/e

## 🚀 New Features ### Human-in-the-Loop (HITL) We are introducing the **Human Input** node, a major update that transforms how AI and humans collaborate within Dify workflows. **Background** Previously, workflows were binary: either fully automated or fully manual. This created a "trust gap" in high-stakes scenarios where AI speed is needed but human judgment is essential. With HITL, we are making h uman oversight a native part of the workflow architecture, allowing you to embed review steps directly into the execution graph. **Key Capabilities** - **Native Workflow Pausing:** Insert a "Human Input" node to suspend workflow execution at critical decision points. - **Review & Edit:** The node generates a UI where humans can review AI outputs and modify variables (e.g., editing a draft or correcting data) before the process continues. - **Action-Based Routing:** Configure custom buttons (like "Approve," "Reject," or "Escalate") that determine the subsequent path of the workflow. - **Flexible Delivery Methods:** Human input forms can be delivered via **Webapp** or **Email**. In cloud environments, Email delivery availability may depend on plan/feature

## What's Changed * fix(docker): persist IRIS data across container recreation using Durable %SYS by @TomoOkuyama in https://github.com/langgenius/dify/pull/31899 * fix(docker): improve IRIS data persistence with proper Durable %SYS by @TomoOkuyama in https://github.com/langgenius/dify/pull/31901 * fix: base url in client by @hyoban in https://github.com/langgenius/dify/pull/31902 * fix: auto summary env by @zxhlyh in https://github.com/langgenius/dify/pull/31930 * fix: fix delete_draft_variables_batch cycle forever by @fatelei in https://github.com/langgenius/dify/pull/31934 * chore: bump version in docker-compose and package manager to 1.12.1 by @QuantumGhost in https://github.com/langgenius/dify/pull/31947