Is image generation working in v0.7.7?

[rioiart]

Hi all,

I have been trying without success to get image generation working in librechat. I created an agent and I added the DALL-E-3 tool. I set the environment variables correctly (tested on the command line that image generation works with the API key) but I get a completely non-informative error in the UI (checked the log files, there is nothing even with debug logging enabled)

The log file () contains this:

2025-03-14T09:25:38.881Z debug: [BaseClient] userMessage
{
  messageId: "1f7fd5e8-dba8-4dd8-800a-a02d32157d98",
  parentMessageId: "00000000-0000-0000-0000-000000000000",
  conversationId: "a71fe417-8bf5-42d9-9c48-870dd74c2a5e",
  sender: "User",
  text: "Please generate an image of a funny duck,.",
  isCreatedByUser: true,
  tokenCount: 13,
}
2025-03-14T09:25:38.960Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-03-14T09:25:40.267Z debug: [saveConvo] api/app/clients/BaseClient.js - saveMessageToDatabase #saveConvo
2025-03-14T09:25:40.282Z debug: [AgentController] Request closed

I added more logging in the _call function in the DALLE3.js script, but it seems that the function is never called.

I also tried flux and that also didn't work. Same issue.

Can anyone confirm the status of the image generation feature in v0.7.7? Looking forward to any replies!

[andywang-249]

+1

[SimplyCorbett]

+1

[danny-avila]

[danny-avila]

Currently, the generated images are being returned to the LLM. If your configured LLM doesn't support this, it will result in an error, but it will still show the image.

[rioiart]

Hi Danny, thanks for your reply, could you please clarify what configuration you used to generate this image (which endpoint, which model configured for the agent)?

I was testing with an agent configured with many different models, ranging from sonnet 3.7, 3.5, gpt-4o, gemini, etc. None of them worked, while other tool calling, including MCP did work. Interpreting images also worked.

Based on my investigation into the code I suspect that image generation only works for the assistants API (however I am not a front-end dev, and was using Aider for this).

Looking forward to your reply!