the provider catalog

Ollama

{"meta":{"title":"Ollama · run Nika workflows locally, no key","angle":"The shortest path to local: pull a model once, point the file at it, and nothing leaves the machine.","pick":"the default when the work is private, frequent, or should cost hardware instead of tokens.","note":"Open source (MIT) with its own inference engine and model library.","setup":[{"text":"Install the runtime","code":"brew install ollama"},{"text":"Pull the model the file names","code":"ollama pull qwen3.5:4b"},{"text":"No key: the server answers on localhost:11434."}]},"usage":{"id":"ollama","yaml":"nika: v1\nworkflow:\n  id: local-summary\n  description: \"a summary that never leaves the machine — Ollama serves on localhost:11434\"\n\ntasks:\n  summarize:\n    infer:\n      model: ollama/qwen3.5:4b\n      max_tokens: 200\n      prompt: |\n        Summarize in two sentences what a hash-chained trace\n        gives an auditor that a plain log file cannot.\n\noutputs:\n  summary: ${{ tasks.summarize.output }}\n","file":"ollama.nika.yaml"},"audit":{"clean":true,"models_resolve":true,"tasks":1,"waves":1,"cost_ceiling_usd":null,"cost_floor_usd":0,"pricing":[{"model":"ollama/qwen3.5:4b","input_per_million":null,"output_per_million":null}],"pricing_as_of":"2026-07-07"}}

[{"slug":"llama3-2-3b","id":"llama3.2:3b","aliases":["llama"]},{"slug":"qwen3-5-4b~2","id":"qwen3.5:4b","aliases":["qwen"]}]

The shortest path to local: pull a model once, point the file at it, and nothing leaves the machine. One provider/model line moves a workflow here or away again: the same file runs on every provider the catalog names, and swapping is a one-word diff. Machines read the catalog; the binary answers nika catalog.

2pinned modelsaliases pin exact upstream ids
localfamilythe sovereign default: your hardware, no key
no keynothing to leakthe sovereign path
openai-chatwire dialectthe adapter speaks it

the wire

local · the sovereign default: your hardware, no key

A workflow names this provider in the model line (model: ollama/qwen3.5:4b), and the engine speaks openai-chat on the wire. No key exists on this path: nothing to rotate, nothing to leak.

default: qwen3.5:4b
cheap: llama3.2:3b
dialect: openai-chat
tags: local · open-source

the models

2 pins

Aliases are what the file may write; the engine pins each one to an exact upstream id, so a run stays reproducible after a vendor rotates names.

qwenqwen3.5:4bctx 128k · vision
llamallama3.2:3bctx 128k · vision

in the released catalog

2 rooms

The wire models this provider serves in the released binary's catalog · each one is a room with its seats, exact prices and measured energy.

llama3.2:3bas llama
qwen3.5:4bas qwen

the whole register: /catalog/models · what the released binary knows, vendored at the engine pin.

in a real file

a complete workflow · shown whole

ollama.nika.yaml

1nika: v12workflow:3  id: local-summary4  description: "a summary that never leaves the machine — Ollama serves on localhost:11434"56tasks:7  summarize:8    ◇infer:9      model: ollama/qwen3.5:4b10      max_tokens: 20011      prompt: |12        Summarize in two sentences what a hash-chained trace13        gives an auditor that a plain log file cannot.1415outputs:16  summary: ${{ tasks.summarize.output }}17

check-green against the released engine · 1 task · 1 wave · no closed ceiling: an unpriced model is unpriced, never free · the drift gate re-proves this copy on every test run · open it in the playground →

get set up

1Install the runtime brew install ollama
2Pull the model the file names ollama pull qwen3.5:4b
3No key: the server answers on localhost:11434.

Pick it for the default when the work is private, frequent, or should cost hardware instead of tokens. Open source (MIT) with its own inference engine and model library. Then prove the seat: nika doctor names every provider this binary can reach.

swap it

4 more local seats

Moving off Ollama is editing one line. The rest of the file (tasks, permits, outputs, the trace) does not change. The register orders the catalog local first; beyond it the engine embeds 21 more OpenAI-compatible endpoints.

model: ollama/qwen3.5:4b model: mistral/mistral-large-latestthe whole migration · nothing else moves

cross-references

the words that meet it

[{"slug":"the-pipeline-is-a-file","title":"The pipeline is a file","date":"2026-07-11"},{"slug":"written-by-agents","title":"Written by agents, reviewed by you","date":"2026-07-11"}]

from the blog

where it lives

Try it in the playground, or install and go local first. Read the spec →