Quickstart

Ollaya runs open decision models on your own machine. You give a model a state — a message, an email, a ticket, a JSON object — plus a few typed questions, and it returns typed answers with calibrated probabilities in a single forward pass.

Ollaya is pre-release. The commands below describe the first version, which is being built right now. Watch the repository to hear when it ships.

1. Install

On Linux and macOS:

curl -fsSL https://ollaya.cobanov.dev/install.sh | sh

Or run the Docker image — see Download for all options.

2. Run a model

ollaya run pulls the model on first use, loads it, and answers:

ollaya run laya --preset triage "I was charged twice for my subscription this month."

laya is a router: it inspects the language of the state and runs laya:en or laya:multilingual. --preset triage asks a built-in set of support-triage questions.

3. Start the server

ollaya serve

The server listens on 127.0.0.1:11435. Loaded models stay warm for five minutes after their last request (keep_alive), so follow-up calls skip the load.

4. Ask your own questions

curl http://localhost:11435/api/decide \
  -H "Content-Type: application/json" \
  -d '{
    "model": "laya",
    "state": "Hi, I cannot log in since this morning and I have a demo at 3pm.",
    "questions": {
      "topic": {
        "type": "choice",
        "instructions": "What is this message about?",
        "criteria": {
          "billing": "Payments, invoices and refunds",
          "access": "Login, passwords and permissions",
          "other": "Anything else"
        }
      },
      "urgency": {
        "type": "score",
        "instructions": "How urgent is this?",
        "criteria": ["Not urgent", "Normal", "Urgent"]
      },
      "angry": {
        "type": "noul",
        "instructions": "Is the customer angry?"
      }
    }
  }'

Every answer comes back typed: a choice with probabilities per option, a score as an expected value over the levels, and a noul as the probability of true. See the API reference for the full format.

5. Use an existing TypeSafe client

Ollaya also serves TypeSafe's API. Point any TypeSafe SDK at your local server:

export TYPESAFE_BASE_URL=http://localhost:11435

See TypeSafe compatibility.

6. Bake your questions into a model

Write a Modelfile once and run it by name:

ollaya create triage -f Modelfile
ollaya run triage "I was charged twice for my subscription this month."