This is not Gemini chat, it is google-vertexai, therefore there is no key. For real Gemini need to use in project google-genai, the key will be used there

As @fejdraus mentioned, it currently support google-vertexai. However, if you have an API key came from Google AI Studio, I think the workaround may work for you. Because it works on me:

• (Step 1) Add a line langchain-google-genai = "0.0.9" to [tool.poetry.dependencies] section in backend/app/pyproject.toml.
• (Step 2) Add a line GOOGLE_API_KEY=<your-api-key> to .env file, where the value of <your-api-key> is the api key generated on Google AI Studio
• (Step 3) Add a line from langchain_google_genai import ChatGoogleGenerativeAI at the head of python file backend/app/api/llms.py.
• (Step 4) In python file backend/app/api/llms.py, modify function def get_google_llm(): to be like:

@lru_cache(maxsize=1)
def get_google_llm():
    return ChatGoogleGenerativeAI(
        model="gemini-pro", convert_system_message_to_human=True
    )

• (Step 5) Run docker-compose build.
• (Step 6) Run docker-compose up, you should be able to run Gemini by the api key generated by the Google AI Studio.