llm.md

Now we only support api-based LLM. You can also use ollama to wrap a local LLM as openai-api style.

You can setup a llm by:

from miniagent.llm import ChatGPT, LLMInput

api_key = "xxx"
llm = ChatGPT(api_key)
message = LLMInput(role="user", content="Hey!")
llm.invoke(message)

If you want to set base_url, http_client, and so on. You can input them as kwargs.