LLM Setup

You can use Puppet Master with one or two LLMs. If using a single LLM, the model must be able to do function calling.

One quick way to get started is to use OpenAI with gpt-4o-mini for both function calling and chat. You can also use Ollama and a single LLM model.

Here's an example of using two different LLM models:

OpenAI gpt-4o-mini

Used for doing function calling Example: "Catbot, rez a prim, make ia pyramid and color it purple. Name it rootprim."

Ollama dolphin-mistral

Used for uncensored chat. If your agents determine no function calling is needed, Puppet Master will use Ollama dolphin-mistral for general uncensored chat. Example: "Hello, Catbot."

Since no function calling is needed for any of the bot's agents to respond to "Hello", we pass the user text to Ollama which replies with the bot's normal chat prompt instead of any of the agents' custom prompts.

Another example of using two LLMs would be OpenAI gpt-4o-mini for function calling and gpt-5.1 for general chat.

pm.conf

[inference]
function_calling = openai
chat = ollama

[ollama]
#ollama_function_calling_model = llama3.2
ollama_chat_model = dolphin-mistral
ollama_host = 127.0.0.1
ollama_port = 11434

[openai]
openai_api_key =
#openai_chat_model = gpt4o
openai_function_calling_model = gpt-4o-mini
openai_base_url = https://api.openai.com/v1/