README.md

metadata

title: RustForHuggingClaw
emoji: 🦀
colorFrom: indigo
colorTo: red
sdk: docker
app_port: 7860
datasets:
  - your-username/your-space-name-data

HuggingClaw 🦀

Deploy OpenClaw on HuggingFace Spaces for free — 2 vCPU · 16 GB RAM · 50 GB storage · always online.

AI provider calls are routed through LiteLLM Proxy so you can plug in any provider (OpenAI, Anthropic, Gemini, Azure, OpenRouter, Ollama, local endpoints …) with just two variables.

---

Required Secrets

Set these in Space → Settings → Repository secrets:

Secret	Example	Description
HF_TOKEN	hf_xxx…	HuggingFace token with read + write access
OPENCLAW_DATASET_REPO	your-name/HuggingClaw-data	Private dataset repo used to persist OpenClaw config
LITELLM_API_KEY	sk-xxx…	Your AI provider API key
LITELLM_MODEL	openai/gpt-4o	Model in LiteLLM format (see table below)

---

Optional Secrets

Secret	Default	Description
LITELLM_API_BASE	(provider default)	Custom API endpoint — required for Azure, OpenRouter, local, etc.
AUTO_CREATE_DATASET	false	Set true to auto-create the dataset repo on first boot
SYNC_INTERVAL	60	Seconds between workspace sync pushes to HuggingFace

---

Model Examples

Provider	LITELLM_MODEL	LITELLM_API_BASE
OpenAI	openai/gpt-4o	(leave empty)
Anthropic	anthropic/claude-3-5-sonnet-20241022	(leave empty)
Google Gemini	gemini/gemini-1.5-pro	(leave empty)
Azure OpenAI	azure/<deployment-name>	https://<resource>.openai.azure.com
OpenRouter	openrouter/openai/gpt-4o	https://openrouter.ai/api/v1
Ollama (local)	ollama/llama3	http://localhost:11434
Any OpenAI-compat	openai/<model-name>	your custom base URL

---

Setup

1. Duplicate this Space on the HuggingFace Space page.
2. Create a private Dataset repo at huggingface.co/new-dataset.
3. Set all Required Secrets above.
4. Edit this README: update the datasets: field to match your dataset repo.
5. The Space will build; OpenClaw will be available on port 7860.

---

Architecture

HF Space (Docker)
├── openclaw-hf-sync  ← Rust binary (pid 1 via tini)
│   ├── on boot  : pull ~/.openclaw from HF dataset
│   ├── every 60s: push ~/.openclaw changes to HF dataset
│   └── on exit  : final push
└── start.sh  ← child process
    ├── LiteLLM Proxy  (127.0.0.1:4000)  ← routes to your AI provider
    └── OpenClaw Gateway  (:7860)  ← points to LiteLLM Proxy