ollama (slug.thirdparty)

slug.thirdparty.ollama

slug.ollama — Ollama LLM client

Sends chat requests to a locally running Ollama instance via the /api/chat endpoint and returns the model’s reply as a string.

Configuration

Quick start

val { ollamaChat } = import("slug.ollama")

val history = [
  { role: "user", content: "What is Slug?" }
]

val reply = ollamaChat(history)
println(reply)

Conversation history

history is a list of { role, content } maps in Ollama’s message format. Roles are "user", "assistant", and "system". Build up history by appending each turn:

val history1 = history :+ { role: "assistant", content: reply }
val history2 = history1 :+ { role: "user", content: "Tell me more" }
val reply2   = ollamaChat(history2)

System prompt

Pass systemPrompt to prepend a system message before every request. The system message is injected fresh on each call and never stored in history, so it doesn’t grow with the conversation.

ollamaChat(history, systemPrompt: "You are a helpful Slug expert.")

@effects(‘net’)

TOC

• ollamaChat(history, systemPrompt, model)

Functions

ollamaChat(history, systemPrompt, model)

fn slug.thirdparty.ollama#ollamaChat(@list history, @str systemPrompt = nil, @str model = cfg(ollama_model, llama3)) -> ?

sends a chat request to Ollama and returns the model’s reply text.

history is a list of { role, content } maps representing the conversation so far. Roles are "user" and "assistant". If systemPrompt is provided it is prepended as a "system" message on every call without being stored in history.

Returns nil on error (details printed to stdout).

@effects(‘net’)

Effects: net