This app presents two types of agents: 'Tool Calling' and 'ReAct' using LangChain. GPT models from OpenAI and Gemini models from Google are supported, but only the 'ReAct' type of agent is implemented for Gemini models.

• For GPT models such as 'gpt-4o', your OpenAI API key is needed. You can obtain an API key from https://platform.openai.com/account/api-keys.

• For Gemini models such as 'gemini-1.5-pro', your Google API key is needed. You can obtain an API key from https://levelup.gitconnected.com/api-tutorial-how-to-use-bing-web-search-api-in-python-4165d5592a7e.

• For internet searches, obtain your Bing Subscription Key here or Google CSE ID here.

• Temperature can be set by the user.

• Voice recognition and Text-To-Speech (TTS) functionalities are supported using OpenAI functions, and therefore are enabled when you use GPT models from OpenAI.

• Recording of the user's voice is stopped when there is no input for 3 seconds.

• In addition to the search tools from Bing or Google, ArXiv, Wikipedia, Retrieval (RAG), and pythonREPL are supported. (PythonREPL from LangChain is still experimental, so caution is needed.)

• Tracing LLM messages is possible using LangSmith if you download the source code and run it on your machine or server. For this, you need a LangChain API key that can be obtained here.

• When running the code on your machine or server, you can use st.secrets to keep and fetch your API keys as environments variables. For such secrets management, see this page.