Runpod
Set up a GPU Pod and expose the decisions API.
Use a GPU Pod for the persistent SelfJev HTTP server. This is a manual setup guide, not an implemented selfjev deploy runpod command. It has not been tested end to end by this project.
1. Configure the Pod
Choose a CUDA/PyTorch template with a 24 GB NVIDIA GPU, 16–32 GB system RAM, and sufficient persistent storage (budget 50 GB). Check the live hourly price before deploying. Configure SSH and set HTTP port 8000 in the Pod template.
Keep the checkout and model cache on a persistent volume, for example under /workspace. Storage and compute can have separate billing and lifecycle rules.
2. Install and serve
Connect using the Pod’s SSH instructions. Install Git LFS and uv if your template does not include them. Check the GPU with nvidia-smi.
cd /workspace
git clone https://github.com/Jwuthri/SelfJev.git
cd SelfJev
git lfs install
git lfs pull --include "weights/selfjev_4b/*"
export HF_HOME=/workspace/huggingface
uv sync --frozen --no-dev --extra serve --extra gpu
export SELFJEV_API_KEYS="replace-with-a-long-random-key"
uv run --no-sync selfjev serve --host 0.0.0.0 --port 8000Run the service under a process supervisor for ongoing use; an interactive SSH process may stop when the session closes. Wait until /health responds successfully inside the Pod.
3. Connect your application
The HTTP proxy exposes the service at the following base URL, with your own Pod id:
https://POD_ID-8000.proxy.runpod.netSet this as the SDK’s base_url and pass your Bearer key. The proxy provides HTTPS; the endpoint is public, so authentication must be configured. The proxy has a 100-second request timeout: long operations should use job submission and polling rather than a blocking inference request.
4. Finish the trial
Terminate the Pod when done. Review any retained volumes separately; stopping compute does not necessarily delete storage or stop its charges.
Provider references: SSH access, HTTP ports, proxy URLs, and timeouts. Provider behavior was checked when writing this guide; use the linked docs for current console details.