Model Hub

Models live at /models/<namespace>/<name>. The namespace is your username or an organization you belong to.

curl -X POST https://inferix.co/api/v0/models \
  -H "Authorization: Bearer $INFERIX_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"name": "my-model", "visibility": "PRIVATE"}'

Publishing a public repository requires a verified email address. Private repositories do not.

Large files do not pass through the API. Ask for a pre-signed URL and upload straight to object storage:

# 1. request a URL
curl -X POST https://inferix.co/api/v0/models/$NS/$NAME/files/upload-url \
  -H "Authorization: Bearer $INFERIX_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"path": "model.safetensors", "size": 4294967296}'

# 2. PUT the bytes at the URL it returns
curl -X PUT --upload-file model.safetensors "$UPLOAD_URL"

Git works too, including Git LFS for weights: git clone https://inferix.co/api/git/models/<ns>/<name>.git. Uploaded files are scanned for malware and for leaked secrets before they are served.

• Public — anyone can read the repository, its files, commits and discussions.
• Private — only you, your organization, and teams the repository is assigned to. Every sub-resource is covered, not just the detail page.
• Gated— public listing, but files require an approved access request. You review requests from the repository's Settings tab.

A README.md at the root is the model card. YAML front-matter drives discovery:

---
license: apache-2.0
tags:
  - text-generation
  - lora
base_model: meta-llama/Llama-3-8B
datasets:
  - squad
---

# My model

What it does, what it was trained on, and how to use it.

base_model and datasetspopulate the model tree and the "Trained on" panel. Datasets that are not mirrored here are shown as plain text rather than as links.

Every push is a commit, browsable under the repository's Commits tab or via GET /api/v0/models/<ns>/<name>/commits. Discussions and pull requests hang off the same repository and follow its visibility.