@namzu/ollama

Ollama LLM provider for @namzu/sdk. Local-first LLM support — run Llama, Mistral, Phi, and 100+ open models on your own hardware via the official ollama client, conformant to the LLMProvider contract.

Prerequisites

Requires the Ollama daemon (https://ollama.com) running locally or at a configured host. Install Ollama, pull a model (ollama pull llama3.2), and ensure the daemon is reachable (default http://localhost:11434).

Install

pnpm add @namzu/sdk @namzu/ollama

@namzu/ollama declares @namzu/sdk as a peer dependency. Install both.

Usage

import { ProviderRegistry } from '@namzu/sdk'
import { registerOllama } from '@namzu/ollama'

// Register once at app startup.
registerOllama()

// Fully typed via module augmentation: ProviderConfigRegistry['ollama'].
const { provider, capabilities } = ProviderRegistry.create({
  type: 'ollama',
  host: 'http://localhost:11434', // optional; defaults to OLLAMA_HOST env or localhost
  model: 'llama3.2',              // default model when none is specified per-call
})

const response = await provider.chat({
  model: 'llama3.2',
  messages: [{ role: 'user', content: 'Hello' }],
})

Streaming:

for await (const chunk of provider.chatStream({
  model: 'llama3.2',
  messages: [{ role: 'user', content: 'Tell me a story' }],
})) {
  if (chunk.delta.content) process.stdout.write(chunk.delta.content)
}

Configuration

| Option | Type | Default | Description | | --------- | ------------------- | --------------------------------------------- | ---------------------------------------------- | | host | string | OLLAMA_HOST env var or http://localhost:11434 | Ollama daemon base URL. | | model | string | — | Default model when params.model is omitted. | | fetch | typeof fetch | global fetch | Custom fetch (e.g. to inject auth headers). | | timeout | number (ms) | — | Reserved; not currently enforced by the client. |

Capabilities

import { OLLAMA_CAPABILITIES } from '@namzu/ollama'

// {
//   supportsTools: false,            // varies by model — conservative default
//   supportsStreaming: true,
//   supportsFunctionCalling: false,
// }

Ollama's tool-use support varies by model. The default is false; if you target a tool-capable model (e.g. llama3.1, mistral-nemo), pass { replace: true } with custom capabilities, or consume tool output yourself at the call site.

Observability

This package ships without observability hooks in 0.1.x. OpenTelemetry span emission and structured logging are roadmapped for the forthcoming @namzu/telemetry package — a separate opt-in dependency that wraps any LLMProvider to emit GenAI semantic-convention spans.

License

FSL-1.1-MIT. Same as @namzu/sdk.