the provider catalog

llama.cpp

{"meta":{"title":"llama.cpp · Nika workflows on bare GGUF","angle":"The C/C++ engine most local stacks build on: one quantized GGUF, one llama-server, zero dependencies.","pick":"when you want the metal: quantized weights, CPU and GPU offload, nothing else running.","note":"Open source (MIT) · the upstream lives at ggml-org/llama.cpp.","setup":[{"text":"Install the server","code":"brew install llama.cpp"},{"text":"Serve the GGUF you have","code":"llama-server -m model.gguf"},{"text":"The wire model id is free-form. The server answers for whatever it loaded."}]},"usage":{"id":"llamacpp","yaml":"nika: v1\nworkflow:\n  id: gguf-classify\n  description: \"llama-server answers for whatever GGUF you loaded — the model id is free-form\"\n\ntasks:\n  classify:\n    infer:\n      model: llamacpp/default\n      max_tokens: 20\n      temperature: 0\n      prompt: |\n        One word, bug or feature:\n        \"clicking save twice creates two invoices\"\n\noutputs:\n  label: ${{ tasks.classify.output }}\n","file":"llamacpp.nika.yaml"},"audit":{"clean":true,"models_resolve":true,"tasks":1,"waves":1,"cost_ceiling_usd":null,"cost_floor_usd":0,"pricing":[{"model":"llamacpp/default","input_per_million":null,"output_per_million":null}],"pricing_as_of":"2026-07-07"}}

[{"slug":"default","id":"default","aliases":["default"]}]

The C/C++ engine most local stacks build on: one quantized GGUF, one llama-server, zero dependencies. 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.

1pinned modelaliases 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: llamacpp/default), and the engine speaks openai-chat on the wire. No key exists on this path: nothing to rotate, nothing to leak. The provider id also answers to llama-cpp.

default: default
dialect: openai-chat
tags: local · open-source

the models

1 pin

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.

defaultdefaultctx 128k · vision

in the released catalog

1 room

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.

defaultas default

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

in a real file

a complete workflow · shown whole

llamacpp.nika.yaml

1nika: v12workflow:3  id: gguf-classify4  description: "llama-server answers for whatever GGUF you loaded — the model id is free-form"56tasks:7  classify:8    ◇infer:9      model: llamacpp/default10      max_tokens: 2011      temperature: 012      prompt: |13        One word, bug or feature:14        "clicking save twice creates two invoices"1516outputs:17  label: ${{ tasks.classify.output }}18

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 server brew install llama.cpp
2Serve the GGUF you have llama-server -m model.gguf
3The wire model id is free-form. The server answers for whatever it loaded.

Pick it for when you want the metal: quantized weights, CPU and GPU offload, nothing else running. Open source (MIT) · the upstream lives at ggml-org/llama.cpp. Then prove the seat: nika doctor names every provider this binary can reach.

swap it

4 more local seats

Moving off llama.cpp 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: llamacpp/default model: mistral/mistral-large-latestthe whole migration · nothing else moves

cross-references

the words that meet it

[]

where it lives

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