@plurnk/plurnk-providers-ollama

Ollama provider for plurnk-service. Routes ollama/{model[/registry]} aliases through Ollama's OpenAI-compatible chat-completions endpoint.

install

npm install @plurnk/plurnk-providers-ollama

Requires Node ≥ 25 (native TypeScript).

use

plurnk-service constructs the provider via the static fromEnv factory (SPEC §3). Direct construction is also supported.

import Ollama from "@plurnk/plurnk-providers-ollama";

const provider = await Ollama.fromEnv(process.env, "qwenzel:latest");

const result = await provider.generate({
    messages: [
        { role: "system", content: "You are a plurnk agent." },
        { role: "user",   content: "What is the capital of France?" },
    ],
});

env

No fallback defaults — required vars throw at fromEnv if missing or unparseable. Defaults belong in plurnk-service's .env.example cascade, not in library code.

| Variable | Required | Notes | |---|---|---| | OLLAMA_BASE_URL | yes | Ollama server URL (e.g. http://localhost:11434). The official OLLAMA_HOST (a bare host:port) is also accepted; OLLAMA_BASE_URL wins | | PLURNK_PROVIDERS_REASONING_BUDGET | yes | Universal reasoning budget (SPEC §4); 0 disables, > 0 toggles think: true on the request body | | PLURNK_FETCH_TIMEOUT | yes | Universal fetch timeout in ms (SPEC §4) | | PLURNK_PROVIDER_RETRY_ATTEMPTS | yes | Transient-failure retry budget (SPEC §4): 0 disables; N retries on 429/5xx/timeout/network with exponential backoff, honoring Retry-After. |

context size

Dynamic, resolved at fromEnv time via POST /api/show. The sibling iterates model_info for any key matching *.context_length (Ollama keys it per-family: qwen35.context_length, llama.context_length, etc.). Throws if no such key is present.

tokenization & pricing

• countTokens: family-dispatched. Llama-family models (llama / llama2 / llama3 / mistral / mixtral, detected via /api/show details.family) tokenize through llama-tokenizer-js — sync, pure JS, drop-in. Everything else (qwen, gemma, phi, deepseek, etc.) falls back to the chars/4 heuristic until per-family tokenizers land in pass-3. The dispatch decision is made once at fromEnv time and frozen on the instance.
• costFor: returns 0. Local Ollama models are free to operate; pico-dollar cost rollups always sum to zero.

license

MIT.