class LLM::OllamaAdapter

Overview

Adapter for Ollama (LLM::Ollama) with optional context reuse.

Included Modules

Defined in:

llm/adapter.cr

Constructors

Class Method Summary

Instance Method Summary

Instance methods inherited from module LLM::Adapter

close : Nil close, request(prompt : String, format : String = "json") : String request, request_messages(messages : Messages, format : String = "json") : String request_messages, request_messages_with_tools(messages : Messages, _tools : String) : String request_messages_with_tools, request_with_context(system : String | Nil, user : String, format : String = "json", cache_key : String | Nil = nil) : String request_with_context, supports_context? : Bool supports_context?, supports_native_tool_calling? : Bool supports_native_tool_calling?

Constructor Detail

def self.new(client : LLM::Ollama) #

[View source]

Class Method Detail

def self.flatten_messages(messages : Messages) : Tuple(String | Nil, String) #

Promoted to a class-level pure function so the flattening rule (system messages joined with \n\n, non-system/non-user roles dropped, nil system when no system messages were present) is unit-testable without standing up a real Ollama client.


[View source]

Instance Method Detail

def client : LLM::Ollama #

[View source]
def request(prompt : String, format : String = "json") : String #
Description copied from module LLM::Adapter

Send a single prompt and get a response as a String.


[View source]
def request_messages(messages : Messages, format : String = "json") : String #
Description copied from module LLM::Adapter

Send chat-style messages (system/user) and get a response as a String.


[View source]
def request_with_context(system : String | Nil, user : String, format : String = "json", cache_key : String | Nil = nil) : String #
Description copied from module LLM::Adapter

Context-aware request. Adapters that support provider-side context can reuse it using a cache_key. Default implementation falls back to request_messages without context reuse.


[View source]
def supports_context? : Bool #
Description copied from module LLM::Adapter

Whether this adapter supports server-side KV context reuse across calls.


[View source]