API reference

ollaya serve exposes two APIs on http://localhost:11435:

the native API under /api/*, modelled on Ollama's, for decisions and model management;
the TypeSafe-compatible API under /v1/*, so existing TypeSafe SDKs work unchanged. See TypeSafe compatibility.

Conventions

Request and response bodies are JSON. Send Content-Type: application/json.
Model names use the name:tag format. A missing tag means latest.
Streaming endpoints return newline-delimited JSON (NDJSON): one JSON object per line.
Errors return a non-2xx status code and a body with a single error field:

{ "error": "model 'laya:xl' not found" }

Decide

POST /api/decide

Answers typed questions about a state in a single forward pass. The body is the same as for /v1/systemone, plus an optional keep_alive. The response is TypeSafe's response plus routing and timings.

Request

Field	Type	Description
`model`	string	Model name, e.g. `laya` or `laya:multilingual`
`state`	string or object	The text, email, ticket or JSON to decide about
`questions`	object	Question id → question
`keep_alive`	string	Optional. How long to keep the model loaded after the request (default `5m`)

Questions

Every question has a type, instructions, and — depending on the type — criteria.

Type	`criteria`	Notes
`choice`	Object of option → description	Up to 255 options
`score`	Array of level descriptions, lowest first	2–10 levels
`noul`	Optional `{"true": "…", "false": "…"}`	Binary; the answer is P(true)

Example

curl http://localhost:11435/api/decide \
  -H "Content-Type: application/json" \
  -d '{
    "model": "laya",
    "state": "I was charged twice for my subscription this month.",
    "questions": {
      "department": {
        "type": "choice",
        "instructions": "Which team should handle this?",
        "criteria": {
          "billing": "Payments, invoices and refunds",
          "technical": "Bugs, errors and outages",
          "account": "Login, profile and settings"
        }
      },
      "urgency": {
        "type": "score",
        "instructions": "How urgent is this?",
        "criteria": ["Not urgent", "Normal", "Urgent"]
      },
      "refund": {
        "type": "noul",
        "instructions": "Is the customer asking for a refund?",
        "criteria": { "true": "Asks for money back", "false": "Does not ask for money back" }
      }
    }
  }'

Response

{
  "model": "laya",
  "answers": {
    "department": {
      "type": "choice",
      "choice": "billing",
      "confidence": 0.78,
      "probabilities": { "billing": 0.852, "account": 0.087, "technical": 0.061 }
    },
    "urgency": {
      "type": "score",
      "score": 1.2,
      "confidence": 0.34,
      "legend": { "0": "Not urgent", "1": "Normal", "2": "Urgent" },
      "probabilities": { "0": 0.12, "1": 0.56, "2": 0.32 }
    },
    "refund": { "type": "noul", "noul": 0.91 }
  },
  "routing": { "model": "laya:en" },
  "usage": { "input_tokens": 84, "output_tokens": 0 },
  "total_duration": 41250000,
  "load_duration": 0
}

Field	Description
`model`	The model you asked for
`answers`	Question id → answer (see below)
`routing`	Which checkpoint served the request when `model` is a router such as `laya`
`usage`	`input_tokens` read from the state and questions; `output_tokens` is always `0`
`total_duration`	Time spent on the request, in nanoseconds
`load_duration`	Time spent loading the model, in nanoseconds (`0` when it was already warm)

Answer type	Fields
`choice`	`choice` (the most likely option), `confidence`, `probabilities` per option
`score`	`score` (expected value over level indexes), `confidence`, `legend` (index → level), `probabilities` per index
`noul`	`noul`: probability that the answer is true

confidence is the normalized top probability: (K · p_max − 1) / (K − 1), where K is the number of options or levels. It is 0 when all options are equally likely and 1 when one option has all the probability.

Decision models never generate text, so usage.output_tokens is always 0.

List local models

GET /api/tags

Returns the models on this machine.

{
  "models": [
    { "name": "laya:latest", "size": 1712000000, "modified_at": "2026-10-01T09:30:00Z", "digest": "sha256:…" }
  ]
}

Show model details

POST /api/show

{ "model": "laya:en" }

Returns the model's details: backbone, parameters, context length, precision, baked-in questions, calibration and license.

Pull a model

POST /api/pull

{ "model": "laya:multilingual" }

Streams progress as NDJSON:

{"status": "pulling manifest"}
{"status": "pulling sha256:…", "digest": "sha256:…", "total": 650000000, "completed": 120000000}
{"status": "pulling sha256:…", "digest": "sha256:…", "total": 650000000, "completed": 650000000}
{"status": "verifying sha256 digest"}
{"status": "writing manifest"}
{"status": "success"}

Set "stream": false to receive a single response when the pull finishes.

List running models

GET /api/ps

Returns the models loaded in memory and when each will be unloaded (keep_alive).

Delete a model

DELETE /api/delete

{ "model": "my-guardrail" }

Copy a model

POST /api/copy

{ "source": "laya:en", "destination": "my-guardrail" }

Create a model

POST /api/create

Creates a model from a Modelfile.

{ "model": "triage", "modelfile": "FROM laya:en\nQUESTIONS ./questions.json" }

Progress is streamed as NDJSON, ending with {"status": "success"}.

Push a model

POST /api/push

{ "model": "yourname/triage" }

Uploads the model to a registry. Progress is streamed as NDJSON.

Version

GET /api/version

{ "version": "0.1.0" }

TypeSafe-compatible endpoints

Endpoint	Description
`POST /v1/systemone`	Same request and response as TypeSafe's System One endpoint
`POST /v1/decisions`	Alias of `/v1/systemone`
`GET /v1/models`	The models available on this machine

See TypeSafe compatibility for details.