VLM support - Image input for multimodal models

[amirvenus]

Hi,

There are some viable VLMs such as Qwen2.5-VL that I would like to run on android.

Is that supported with this library?

Thanks

[agi-2026]

+1

[bernardladenthin]

I noticed the repository hasn't seen recent activity, so I've taken the liberty of forking it and continuing development at: https://github.com/bernardladenthin/java-llama.cpp

See also #125

[bernardladenthin]

Implemented in the downstream fork at https://github.com/bernardladenthin/java-llama.cpp
in PR #189. The build links the upstream mtmd library into jllama
(CMakeLists.txt:125-145,253-255); ModelParameters exposes setMmproj,
setMmprojUrl, enableMmprojAuto, enableMmprojOffload
(ModelParameters.java:1250-1281); and the new typed image API ships as
ContentPart.text/imageUrl/imageBytes/imageFile, ChatMessage(role, List<ContentPart>),
ChatMessage.userMultimodal(...), and InferenceParameters.setMessages(List<ChatMessage>).
The serializer emits the OAI array-form content that the upstream chat path
already routes through mtmd — no new JNI was required.