Get started

The intelli chatbot function connects with the leading AI models such as GPT, Claude, Gemini and Mistral through one input. You can switch between providers, or upgrade to the latest models, without changing the code of your app.

Core Components

ChatModelInput: This class provides a unified entry to all chatbot providers. It holds the system message, the model, the message history, and model parameters like temperature and max tokens.

Chatbot: The primary class that interfaces with the AI providers. It takes the API key, the provider name, and optional settings such as a timeout, a proxy or a server URL.

Available Providers

Use the ChatProvider enum, or its string value, to select the provider:

Provider	ChatProvider	Value
OpenAI and Azure OpenAI	ChatProvider.OPENAI	openai
Anthropic	ChatProvider.ANTHROPIC	anthropic
Google Gemini	ChatProvider.GEMINI	gemini
Mistral	ChatProvider.MISTRAL	mistral
NVIDIA (DeepSeek, Llama and NIM)	ChatProvider.NVIDIA	nvidia
Self-hosted vLLM	ChatProvider.VLLM	vllm
Offline llama.cpp	ChatProvider.LLAMACPP	llamacpp
Offline Keras (Gemma, Llama, Mistral)	ChatProvider.KERAS	keras

See DeepSeek & Llama, vLLM, llama.cpp and Gemma for the provider specific setup.

Example

Imports

from intelli.model.input.chatbot_input import ChatModelInput
from intelli.function.chatbot import Chatbot, ChatProvider

Prepare the input

chat_input = ChatModelInput(system="You are a helpful assistant.", model="gpt-5.5")
chat_input.add_user_message("Explain the plot of the Inception movie in one line.")

Call the chatbot

chatbot = Chatbot(api_key=YOUR_API_KEY, provider=ChatProvider.OPENAI)
response = chatbot.chat(chat_input)
print(response[0])

chat returns a list of replies. When the model calls a tool, the item is a dictionary instead of text, see Tool calling.

Default Models

When you omit the model, intelli picks a current default for the provider:

Provider	Default model
Openai	gpt-5.5
Anthropic	claude-sonnet-5
Gemini	gemini-2.5-flash
Mistral	mistral-large-latest

Use claude-opus-5 for the larger Anthropic model. The Claude 5 family and Opus 4.7 and above dropped the sampling parameters, so intelli omits temperature for those models and they work without any change on your side.

Streaming

stream yields the reply as it is generated. It is available for openai, anthropic, nvidia, vllm and llamacpp.

chat_input = ChatModelInput("You are a helpful assistant.", model="claude-sonnet-5")
chat_input.add_user_message("Write a short poem about the sea.")

chatbot = Chatbot(YOUR_ANTHROPIC_KEY, ChatProvider.ANTHROPIC)
for chunk in chatbot.stream(chat_input):
    print(chunk, end="", flush=True)

The GPT-5 family runs on the responses API, which intelli calls through chat only. To stream from openai, use a chat completions model such as gpt-4.1, or add the :chat suffix described below.

GPT-5 Parameters

The GPT-5 family runs on the responses API, and the input accepts its parameters:

chat_input = ChatModelInput(
    system="You are a helpful assistant.",
    model="gpt-5.5",
    reasoning_effort="low",   # low, medium, high, xhigh, none
    verbosity="medium",       # low, medium, high
    max_tokens=512,           # sent as max_output_tokens
)

tool_choice is available as well, on both the openai and the anthropic paths. To force the classic chat completions endpoint for a GPT-5 deployment, add :chat to the model id, for example gpt-5.5:chat.

Chatbot Options

Pass the optional settings in options:

Option	Providers	Description
timeout	all remote providers	Request timeout in seconds, default 180.
proxy_helper	openai	Route the calls to Azure OpenAI or another proxy.
baseUrl	nvidia, vllm	The URL of your NIM or vLLM server.
model_path, model_params	llamacpp	The local GGUF file and its parameters.
model_name, model_params	keras	The Keras model and the Kaggle credentials.

chatbot = Chatbot(YOUR_API_KEY, ChatProvider.OPENAI, options={"timeout": 60})

Azure OpenAI

Set your Azure resource in ProxyHelper, then pass the deployment name as the model:

from intelli.utils.proxy_helper import ProxyHelper

proxy_helper = ProxyHelper()
proxy_helper.set_azure_openai(YOUR_AZURE_RESOURCE)

azure_bot = Chatbot(YOUR_AZURE_API_KEY, ChatProvider.OPENAI, options={"proxy_helper": proxy_helper})

chat_input = ChatModelInput("You are a helpful assistant.", model=YOUR_DEPLOYMENT_NAME)
chat_input.add_user_message("What is the capital of France?")
response = azure_bot.chat(chat_input)