Ollama Integration LLM Default Temperature Setting

mikey60 · April 22, 2025, 8:47pm

Since the HA Ollama integration does not have a configuration option for the LLM model temperature, I was wondering what is the default temperature setting. Or does it default to the value in the Modelfile for the LLM configured in the Ollama integration. IMO a configuration option for a LLM temperature value would be useful in the Ollama integration.

super-qua · April 23, 2025, 5:26pm

it seems that no temperature is set explicitly, at least in the chat message

github.com/home-assistant/core

homeassistant/components/ollama/conversation.py

65db3c116


      
          for _iteration in range(MAX_TOOL_ITERATIONS):
              try:
                  response_generator = await client.chat(
                      model=model,
                      # Make a copy of the messages because we mutate the list later
                      messages=list(message_history.messages),
                      tools=tools,
                      stream=True,
                      # keep_alive requires specifying unit. In this case, seconds
                      keep_alive=f"{settings.get(CONF_KEEP_ALIVE, DEFAULT_KEEP_ALIVE)}s",
                      options={CONF_NUM_CTX: settings.get(CONF_NUM_CTX, DEFAULT_NUM_CTX)},
                  )
              except (ollama.RequestError, ollama.ResponseError) as err:
                  _LOGGER.error("Unexpected error talking to Ollama server: %s", err)
                  raise HomeAssistantError(
                      f"Sorry, I had a problem talking to the Ollama server: {err}"
                  ) from err
          
              message_history.messages.extend(
                  [
                      _convert_content(content)

but that might be a good enhancement actually

mikey60 · April 23, 2025, 6:09pm

Thanks for the response. When I get a chance I will post something in the feature request.