Replies: 1 comment
-
|
Hi @stefhof! I'm Dosu and I’m helping the dify team. There’s currently no way for an Agent in Dify to pass uploaded files (like images) directly to a Workflow Tool. This is a known limitation: if a workflow tool requires a file-type input (such as file, files, or system_files), it cannot be invoked by an Agent. The agent-to-tool logic will skip or block tools with required file inputs to prevent errors, and the codebase explicitly omits file-type parameters from the tool schema presented to the LLM in the Agent context. As a result, there’s no configuration or prompt engineering workaround to enable this as of now [source]. A suggested workaround is to make the file input parameter non-required, but this usually doesn’t meet practical needs and doesn’t fully solve the problem [source]. Some users have used indirect approaches, like returning a download link via an HTTP node or external service, but this is not a direct file pass and may not fit all use cases [source]. There are ongoing discussions and feature requests for supporting file passing from Agent to Workflow Tool, but as of December 2025, this is not yet implemented. If this feature is critical for your workflow, it’s worth keeping an eye on future releases and community updates [source]. To reply, just mention @dosu. How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Self Checks
1. Is this request related to a challenge you're experiencing? Tell me about your story.
I have an Agent application that handles customer conversations. The Agent has Vision enabled and can receive image uploads from users. I created a Workflow called "Image_Analyzer" that analyzes photos using an LLM with Vision capabilities
This Workflow is published as a Tool and added to the Agent's toolset. The Workflow's Start node has a File-type input variable (user_image) configured to accept images.
The Problem:
When a user uploads an image in the Agent conversation, the Agent attempts to call the Image_Analyzer tool, but it cannot pass the actual image file to the tool. Instead, it only passes the accompanying text message.
What I've tried:
My questions:
2. Additional context or comments
The reason I want to use a separate Workflow Tool for image analysis (rather than the Agent's native Vision) is to have a dedicated, specialized prompt for image assessment and to potentially log/store the analysis results separately.
Beta Was this translation helpful? Give feedback.
All reactions