Quickstart

Install the CLI, deploy a LoRA adapter, and call it in about five minutes end to end. Prefer raw HTTP? Skip to the REST deploy flow.

1. Install the CLI

The installers download the prebuilt lorivo binary and add it to your PATH.

$ curl -fsSL https://lorivo.dev/install.sh | sh

2. Get an API key

Sign in to lorivo.dev and create a key from the API keys card on your dashboard. The full key is shown once. Copy it now. Keys look like ad_9WTqoA….

3. Log in

lorivo login <api-key>

This verifies the key and stores it in ~/.adapters/config.json (mode 0600). For headless use (CI, scripts) you can skip login entirely and set ADAPTERS_ADMIN_KEY. See CLI configuration.

4. Prepare your adapter

lorivo deploy expects a directory with two files:

my-lora/
my-lora/
├── adapter_config.json
└── adapter_model.safetensors
  • LoRA rank 1–64, declared in adapter_config.json
  • base_model_name_or_path matches the base model family (Qwen3)
  • At most 10 files and 2 GiB total

5. Deploy it

lorivo deploy ./my-lora --name my-lora
output
Validated ./my-lora - rank 8, 2 file(s), 1.7 MiB
Registered my-lora
Uploaded v1 in 2.1s (sha256 3f9e1c2a4b5d…)
Deployed my-lora v1 (LOADED)


Live at:  https://lorivo.dev/v1/responses
Model:    qwen3-4b:my-lora

Re-running the command uploads and deploys the next version (v2, v3, …). Roll back any time with lorivo rollback.

6. Call it

Your adapter is served at qwen3-4b:my-lora through the same OpenAI-compatible endpoints as the base model:

curl https://lorivo.dev/v1/chat/completions \
  -H "Authorization: Bearer ad_…" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-4b:my-lora",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

The /v1/responses endpoint is a drop-in alternative. See Inference API for all four endpoints.