How to Set the Google AI Studio API Key for Gemini in the Hiring Agent

The Hiring Agent project reads your Gemini API key from the GEMINI_API_KEY environment variable and automatically initializes the GeminiProvider class when this variable is present.

The InterviewStreet Hiring Agent supports Google's Gemini models through the Google AI Studio API. To enable this integration, you must configure the Google AI Studio API key using environment variables that the application reads at startup. This guide shows you how to set up the key in the interviewstreet/hiring-agent repository and verify that the system correctly routes LLM requests to the Gemini provider.

Copy the Environment Configuration Template

Start by creating your local environment file from the provided template. The repository includes an .env.example file that contains placeholder entries for all supported API keys.

cp .env.example .env

The example file already contains the GEMINI_API_KEY placeholder entry that you will populate with your actual key.

Add Your Google AI Studio API Key

Edit the .env file and replace the placeholder value with your actual API key from Google AI Studio. You can obtain your key from the Google AI Studio API keys page.

GEMINI_API_KEY=your_actual_api_key_here

This variable is defined in prompt.py (lines 66-68), which loads the configuration using load_dotenv() at application startup.

Select the Gemini Provider

By default, the Hiring Agent uses the Ollama backend for local inference. To switch to Gemini, set the LLM_PROVIDER environment variable to gemini and specify your preferred model.

LLM_PROVIDER=gemini
DEFAULT_MODEL=gemini-2.5-pro

Add these entries to the same .env file alongside your API key. The DEFAULT_MODEL can be any supported Gemini model identifier available in your Google AI Studio account.

How the Application Validates the Key

The initialization logic in llm_utils.py (lines 50-60) checks for the presence of GEMINI_API_KEY when you select the Gemini provider. If the key is missing, the system logs a warning and falls back to Ollama. If present, it creates a GeminiProvider instance.

# From llm_utils.py

if model_provider == ModelProvider.GEMINI:
    if not GEMINI_API_KEY:
        logger.warning("⚠️ Gemini API key not found. Falling back to Ollama.")
    else:
        logger.info(f"🔄 Using Google Gemini API provider with model {model_name}")
        provider = GeminiProvider(api_key=GEMINI_API_KEY)

The GeminiProvider class, implemented in models.py (lines 13-22), wraps the official Google Gemini client and handles all downstream API interactions.

Verify the Integration

After starting the application, check the logs to confirm the Gemini provider is active. You should see the message: "Using Google Gemini API provider with model [model_name]".

You can also verify the configuration programmatically by inspecting the environment variables loaded in prompt.py:

from prompt import GEMINI_API_KEY, PROVIDER

print("Provider:", PROVIDER)               # → gemini

print("Gemini API key set:", bool(GEMINI_API_KEY))  # → True

Complete Configuration Example

Here is a minimal .env configuration that enables Gemini with the 2.5 Pro model:

# Core LLM Configuration

LLM_PROVIDER=gemini
DEFAULT_MODEL=gemini-2.5-pro

# API Keys

GEMINI_API_KEY=sk-abcdef1234567890abcdefg

Direct Provider Initialization

For testing or custom scripts, you can initialize the provider directly without relying on the automatic configuration flow:

from llm_utils import initialize_llm_provider

provider = initialize_llm_provider("gemini-2.5-pro")
response = provider.chat(
    model="gemini-2.5-pro",
    messages=[{"role": "user", "content": "Hello, Gemini!"}]
)
print(response["message"]["content"])

This bypasses the environment check and uses the specific model you pass as an argument.

Summary

• Environment Variable: The Hiring Agent requires GEMINI_API_KEY to authenticate with Google AI Studio.
• Configuration File: Copy .env.example to .env and populate the key value.
• Provider Selection: Set LLM_PROVIDER=gemini to override the default Ollama backend.
• Validation: Check application logs for "Using Google Gemini API provider" to confirm successful setup.
• Fallback Behavior: If the key is missing, the system automatically falls back to Ollama with a warning logged.

Frequently Asked Questions

What happens if I don't set the GEMINI_API_KEY?

If the GEMINI_API_KEY environment variable is not set and you have selected LLM_PROVIDER=gemini, the application logs a warning and falls back to the Ollama backend according to the logic in llm_utils.py (lines 50-60). Your requests will be processed locally rather than through the Google AI Studio API.

Can I use different Gemini models with the same API key?

Yes. The GEMINI_API_KEY authenticates your account, while the specific model is determined by the DEFAULT_MODEL environment variable or the model parameter passed to the provider. You can use gemini-2.5-pro, gemini-1.5-pro, or any other model available in your Google AI Studio account.

Where is the Gemini API key stored in the code?

The prompt.py file defines the GEMINI_API_KEY constant by reading it from the environment (lines 66-68). This value is then passed to the GeminiProvider constructor in llm_utils.py when initializing the provider. The key is never hardcoded in the source files and should always be provided via environment variables.

Do I need to restart the application after changing the .env file?

Yes. The Hiring Agent loads environment variables at startup when load_dotenv() executes in prompt.py. Changes to the .env file require a restart to take effect, as the configuration is not hot-reloaded during runtime.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}