@fdekit/provider-ollama

Purpose

@fdekit/provider-ollama provides local Ollama config helpers and a runtime adapter for FDEKit. It calls the local Ollama chat API and parses the next FDEKit loop step.

Use it when an FDEKit deployment should run against a local model server instead of a hosted provider.

Who should use this package

• Deployment authors running local or offline model experiments.
• Runtime integrators who want to pass an Ollama adapter through a provider registry.
• Contributors maintaining provider request/response parsing and resilience behavior.

Choose @fdekit/provider-mock for deterministic no-model tests and hosted provider packages for cloud model APIs.

5-minute quick example

ollama pull llama3.1:8b
ollama serve

import { defineDeployment } from '@fdekit/core';
import { localOllamaProvider } from '@fdekit/provider-ollama';

export default defineDeployment({
  name: 'ollama-support-triage',
  environment: 'local',
  providers: {
    localOllama: localOllamaProvider({
      model: process.env.FDEKIT_MODEL || 'llama3.1:8b',
      apiBaseUrl: process.env.OLLAMA_BASE_URL,
    }),
  },
  agents: {
    // ...
  },
});

Public API surface

Import from the package root:

import { localOllamaProvider, ollamaProvider, createOllamaProvider } from '@fdekit/provider-ollama';
import type { OllamaProviderOptions, OllamaRuntimeOptions } from '@fdekit/provider-ollama';

The provider family is summarized in the public API index: Public API Reference.

Stability/backward-compat notes

@fdekit/provider-ollama is public but pre-1.0. Its root exports are localOllamaProvider, ollamaProvider, createOllamaProvider, OllamaProviderOptions, and OllamaRuntimeOptions.

Subpath imports are internal. Config defaults, env var names, and the FDEKit provider-step JSON contract are the compatibility-sensitive parts.

See also

• Provider contracts: @fdekit/core
• Runtime execution: @fdekit/runtime
• Provider family: mock, OpenAI, Anthropic, Google
• CLI workflow: fdekit