Bukunmi2108/aristotle-model
0
Aristotle Model Service
Dockerized llama.cpp server for Aristotle's OpenAI-compatible model endpoint.
Current Model
- Repo:
nvidia/NVIDIA-Nemotron-3-Nano-4B-GGUF - File:
NVIDIA-Nemotron3-Nano-4B-Q4_K_M.gguf - Context:
32768 - Parallel slots:
1 - Threads:
2
The image bakes the GGUF into /models, so cold starts load the local file instead of downloading it again.
Local Build
docker build -t aristotle-model ./modelLocal Run
docker run --rm -p 8200:7860 aristotle-modelCheck:
curl http://localhost:8200/v1/modelsSmoke test:
MODEL_BASE_URL=http://localhost:8200 ./model/scripts/smoke.shHosted
https://bukunmi2108-aristotle-model.hf.spaceOpenAI-compatible endpoints:
GET /v1/models
POST /v1/chat/completionsThe model exposes thinking in reasoning_content, which Aristotle's agent runtime should preserve separately from final assistant content.
