Platform
Linux/Other (please specify)
Lemonade Version
10.5.1
GPU / APU Model
AMD RYZEN AI MAX+ 395 w/ Radeon 8060S
Component
stable-diffusion.cpp
Bug Description
Editing of images through OpenWebUI doesn't work.
OWUI Side:
prompt Add a small top hat to the frog's head
image_urls ["/api/v1/files/3283b170-1317-4c9d-ab45-3c558ff27219/content"]
Output
{
"error": "400: [ERROR: Bad Request]"
}
Lemonade Side:
Error 400: POST /api/v1/images/edits
[Server] 400 Bad Request details - Body length: 0, Content-Type: multipart/form-data;
Steps to Reproduce
Start a chat with native tool calling
Ask to generate a frog
Ask to edit a hat on the frog
Edit fails - sometimes attempts a new generation but the edit is what matters here
Expected vs Actual Behavior
Expected is that it works without this failure. I have a feeling that openwebui is using some specific subset of the openai api that isn't fully implemented here.
Log Output
2026-05-19 22:34:37.707 [Info] (Telemetry) === Telemetry ===
2026-05-19 22:34:37.707 [Info] (Telemetry) Input tokens: 5092
2026-05-19 22:34:37.708 [Info] (Telemetry) Output tokens: 89
2026-05-19 22:34:37.708 [Info] (Telemetry) TTFT (s): 5.142
2026-05-19 22:34:37.708 [Info] (Telemetry) TPS: 65.29
2026-05-19 22:34:37.708 [Info] (Telemetry) =================
2026-05-19 22:34:37.714 [Info] (Server) POST /api/v1/images/edits
2026-05-19 22:34:37.714 [Error] (Server) Error 400: POST /api/v1/images/edits
2026-05-19 22:34:37.714 [Error] (Server) 400 Bad Request details - Body length: 0, Content-Type: multipart/form-data; boundary=0eea14c1192e4561847097dbf13e807c
2026-05-19 22:34:37.723 [Info] (Server) Model already loaded: Qwen3.6-35B-A3B-MTP
2026-05-19 22:34:37.723 [Info] (Server) POST /api/v1/chat/completions - Streaming
2026-05-19 22:34:37.748 [Info] (Process) 0.11.665.624 I srv params_from_: Chat format: peg-native
2026-05-19 22:34:37.750 [Info] (Process) 0.11.667.749 I slot get_availabl: id 0 | task -1 | selected slot by LCP similarity, sim_best = 0.994 (> 0.100 thold), f_keep = 1.000
2026-05-19 22:34:37.751 [Info] (Process) 0.11.668.299 I reasoning-budget: activated, budget=2147483647 tokens
2026-05-19 22:34:37.751 [Info] (Process) 0.11.668.301 I reasoning-budget: deactivated (natural end)
2026-05-19 22:34:37.751 [Info] (Process) 0.11.668.349 I slot launch_slot_: id 0 | task 30 | processing task, is_child = 0
Additional Context
Relevant code in openwebui
https://github.com/open-webui/open-webui/blob/3660bc00fd807deced3400a63bfa6db47811a3bb/backend/open_webui/routers/images.py#L787
I've attached the captured request to lemonade from openwebui.
Platform
Linux/Other (please specify)
Lemonade Version
10.5.1
GPU / APU Model
AMD RYZEN AI MAX+ 395 w/ Radeon 8060S
Component
stable-diffusion.cpp
Bug Description
Editing of images through OpenWebUI doesn't work.
OWUI Side:
Lemonade Side:
Steps to Reproduce
Start a chat with native tool calling
Ask to generate a frog
Ask to edit a hat on the frog
Edit fails - sometimes attempts a new generation but the edit is what matters here
Expected vs Actual Behavior
Expected is that it works without this failure. I have a feeling that openwebui is using some specific subset of the openai api that isn't fully implemented here.
Log Output
Additional Context
Relevant code in openwebui
https://github.com/open-webui/open-webui/blob/3660bc00fd807deced3400a63bfa6db47811a3bb/backend/open_webui/routers/images.py#L787
I've attached the captured request to lemonade from openwebui.