Skip to content

Add fallback image generation model support for vertical Agent icon generation #249

Description

@Haohao-end
## Is your feature request related to a problem? Please describe.

When users create vertical Agent intelligent agents through the Agent creation workflow, the system needs to generate an Agent avatar/icon image. Currently, the image generation process may depend on a single image generation model or an external generation capability.

This creates several potential issues:

- If the primary image generation model is unavailable, unstable, or fails to generate an image, Agent creation may be blocked or provide a poor user experience.
- Different vertical Agents may require different visual styles, but the current generation strategy may lack fallback capability.
- Generated Agent icons may introduce safety risks if the image generation model does not have sufficient content filtering or safety guarantees.
- A failure in image generation should not affect the core Agent creation workflow.

Therefore, a more reliable and secure image generation fallback mechanism is needed.

---

## Describe the solution you'd like

Add a fallback mechanism for Agent icon/image generation when creating vertical Agent intelligent agents.

The expected solution:

1. Use the primary image generation model for generating Agent icons by default.
2. If the primary model fails due to:
   - service unavailable
   - timeout
   - generation failure
   - safety filtering failure
   - invalid output format

   automatically switch to an alternative image generation model as a fallback.

3. The fallback image generation model should:
   - provide stable image generation capability
   - support Agent icon/avatar generation scenarios
   - include safety filtering mechanisms
   - avoid generating unsafe, inappropriate, or policy-violating images

4. The image generation workflow should:
   - not block Agent creation when image generation fails
   - record generation model, fallback reason, and generation status for observability
   - support future extension to multiple image generation providers/models

Example flow:

Create Vertical Agent
|
v
Generate Agent Icon
|
v
Primary Image Model
|
+----+----+
| Success |
v |
Return Icon |
|
v
Fallback Image Model
|
v
Safety Check
|
v
Return Safe Icon


---

## Describe alternatives you've considered

1. Use only one image generation model

Pros:
- Simple implementation
- Lower maintenance cost

Cons:
- No availability guarantee
- Higher risk of generation failure affecting Agent creation
- Limited safety and quality fallback capability

---

2. Use manually uploaded Agent icons

Pros:
- Full control over generated images

Cons:
- Poor user experience
- Requires additional user operation
- Cannot automatically generate personalized Agent visuals

---

3. Use predefined icon templates only

Pros:
- Safe and stable
- Easy to maintain

Cons:
- Limited personalization
- Cannot satisfy diverse vertical Agent scenarios

---

## Additional context

This feature is mainly used in the vertical Agent creation workflow.

The Agent icon is a user-facing visual asset and should prioritize:

- availability
- safety
- consistency
- generation quality

The fallback mechanism should be designed as a general image generation capability so that additional image models can be integrated in the future.

Related considerations:

- Need unified image generation API abstraction.
- Need model priority and fallback configuration.
- Need safety validation before returning generated images.
- Need monitoring metrics for generation success rate and fallback frequency.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions