Become an operator

Operators serve models through their own inference endpoint. Start with the operator registration page; sign in to a ph0ny developer account before submitting the form.

Run the maintained blueprint

The public Modal Inference Blueprint contains its current prerequisites, model catalog, configuration, and deployment instructions. It requires Rust, a Modal account with GPU access, Foundry, and cargo-tangle for Tangle deployment. Follow its README for the latest commands.

A local path to a working endpoint is:

git clone https://github.com/tangle-network/modal-inference-blueprint
cd modal-inference-blueprint
python3 models/deploy.py list
python3 models/deploy.py cost --task tts
python3 models/deploy.py deploy --model kokoro-tts --org your-modal-org
cp config/example.toml config/operator.toml
python3 models/deploy.py gen-config --org your-modal-org --task tts >> config/operator.toml
# Review config/operator.toml, including the model endpoint and pricing.
cargo build --release
./target/release/standalone

Verify the local endpoint before registering it:

curl http://localhost:8080/health
curl http://localhost:8080/models

The standalone process is for local testing. The blueprint README covers Tangle registration and running the network operator. Make the endpoint publicly reachable before entering its URL in the ph0ny form.

Register with ph0ny

On Become an Operator, enter the operator name, blueprint type, and public endpoint URL. The form can also record model and GPU details and optionally submit a Tangle registration if you connect a wallet. API registration creates a pending operator; the public listing shows active operators after approval and health checks.

Use the Operator Network to inspect listings. The blueprint's README is the source for supported model IDs and current run commands.