DLLM - D Language 🤖 on 🦙.cpp

[Danny Arends]

On Tuesday, 24 March 2026 at 17:51:20 UTC, Sergey wrote:
> On Tuesday, 24 March 2026 at 17:01:03 UTC, Danny Arends wrote:
>> A minimal, clean D language agent built directly on llama.cpp 
>> via importC. No Python, no bindings, no overhead. Runs a 
>> three-model pipeline (agent, summary, embed) with full CUDA 
>> offloading, multimodal vision via mtmd, RAG, KV-cache 
>> condensation, thinking budget, and an extensible tool system 
>> (auto-registered via user-defined attribute 
>> @Tool("Description") on functions). Tools included cover: file 
>> I/O, web search, date & time, text encoding, Docker sandboxed 
>> code execution, and audio playback.
>>
>> https://github.com/DannyArends/DLLM
>>
>> Let me know what you guys think!
>
> was it vibe-coded?)

Lol, no I made it to learn low level llama.cpp and use importC. 
It was fun figuring out how to generate everything surrounding 
agentic tools by using UDAs with __traits. My normal projects 
generally don't use a lot of meta programming and reflection.

However, the current version can read and "understand" it's own 
codebase with just an 8B model. So you could put it in a vibe 
loop to recursively improve itself. Just remove the limitations 
on writing to the workspace/ folder and see what it'll do.