Server surfaces

Three surfaces sound like servers but accept different clients.

Surface	Job	HTTP workflow API
`nika serve` without `--bind`	Fires the `nika.yaml` project clock	no
`nika serve --bind`	Authenticated resident jobs and schedules	yes
`nika model serve`	OpenAI-compatible local inference provider	no

Candidate OpenAPI routes

Method and path	Contract
`GET /health`	unauthenticated health and supported capabilities
`GET /v1/openapi.json`	authenticated contract document
`POST /v1/check`	compact admission acknowledgement or typed workflow refusal
`POST /v1/jobs`	idempotent admission
`GET /v1/jobs/{id}`	durable job projection
`GET /v1/jobs/{id}/status`	current status
`GET /v1/jobs/{id}/events`	SSE event stream
`POST /v1/jobs/{id}/cancel`	cancellation action; 202 remains pending, 200 returns a completed observation
`GET /v1/jobs/{id}/trace/verify`	typed verification diagnostic
`GET /v1/workflows`	contained names
`GET /v1/workflows/{name}`	path-free metadata
`GET /v1/schedules/{id}`	schedule status
`PUT /v1/schedules/{id}`	CAS-backed schedule apply

The contract has no POST /v1/run, job-list, artifact, schedule-list, schedule-delete, trigger or backfill routes. Remote trace verification currently resolves a typed unavailable verdict because the server has no path-free journal authority; the diagnostic route itself is live.

nika serve --bind 127.0.0.1:18765 \
  --workflows ./workflows \
  --token-file .nika/serve.token

Local application path

import { Nika } from '@supernovae-st/nika'

const nika = new Nika({ cwd: process.cwd() })
const run = await nika.run('workflow.nika')
const result = await run.result()

This path needs no resident server. These examples target the published @supernovae-st/nika@0.120.2 npm package (bundled engine 0.120.2). The standalone engine CLI releases independently.

Local model path

nika model serve --model Qwen/Qwen3-0.6B-GGUF

The engine calls this inference endpoint as a provider. Application code does not submit workflow jobs to it.

Resident server

Operate the default firer and opt-in HTTP door.

Remote client

Configure the matching One SDK transport.

Local models

Run inference on your own machine.

Deployment topologies

Choose a boundary by the job it owns.

Overview

Start

Project

Local API

Runtime

Remote API

Operations

Reference

Server surfaces

Candidate OpenAPI routes

Local application path

Local model path

Resident server

Remote client

Local models

Deployment topologies

​Candidate OpenAPI routes

​Local application path

​Local model path

Resident server

Remote client

Local models

Deployment topologies

Candidate OpenAPI routes

Local application path

Local model path