How to Configure Hiring Agent to Use Google Gemini: Complete Setup Guide

Set the GEMINI_API_KEY environment variable and specify LLM_PROVIDER=gemini in your .env file to switch the Hiring Agent from the default Ollama backend to Google Gemini.

The Hiring Agent by InterviewStreet supports multiple LLM backends, including Ollama (default) and Google Gemini. To configure the agent to use Gemini, you must provide an API key and select a supported Gemini model name. This guide walks through the configuration process using the actual source code implementation.

Prerequisites

Obtain a Gemini API Key

Before configuring the Hiring Agent, you need a valid Google Gemini API key from Google AI Studio. The application expects this key to be available as the GEMINI_API_KEY environment variable.

Configuration Steps

Step 1: Configure Environment Variables

The Hiring Agent loads environment variables from .env files via python-dotenv. In main/prompt.py, lines 12-18, the code calls load_dotenv() and then reads GEMINI_API_KEY (falling back to an empty string if unset).

Create your environment file by copying the template:

cp main/.env.example .env

Edit the .env file to set the required variables:

LLM_PROVIDER=gemini
GEMINI_API_KEY=your_actual_gemini_api_key_here
DEFAULT_MODEL=gemini-2.5-pro

The prompt.py module validates these settings at runtime. Lines 66-68 retrieve the API key using os.getenv("GEMINI_API_KEY"), while lines 36-44 and 55-63 map model names like gemini-2.5-pro to the ModelProvider.GEMINI enum.

Step 2: Select a Gemini Model

In main/prompt.py, the application maintains a mapping between model names and providers. Any model name starting with gemini- (such as gemini-2.5-pro or gemini-2.5-flash) automatically resolves to the Gemini provider. Set your preferred model using the DEFAULT_MODEL environment variable.

Step 3: Install Dependencies

Ensure the Google Generative AI SDK is installed:

pip install -r requirements.txt

The requirements.txt should include the google-generativeai package required by the Gemini provider implementation.

How the Gemini Integration Works

Environment Loading in prompt.py

The configuration flow begins in main/prompt.py, where load_dotenv() processes your .env file. The module exposes the API key as a module-level variable, making it available to the provider factory. Lines 12-18 handle the dotenv loading, while lines 66-68 expose GEMINI_API_KEY to the rest of the application.

Provider Implementation in models.py

The GeminiProvider class in main/models.py (lines 13-21) handles the Gemini API connection. Its constructor receives the API key and immediately calls google.generativeai.configure(api_key=...) to initialize the client. Lines 31-38 store this configured client for subsequent chat operations.

Chat Request Handling

When processing requests, the GeminiProvider.chat() method (lines 48-66 in main/models.py) instantiates a GenerativeModel using the model name specified in your configuration. It applies generation parameters (temperature, top-p) from MODEL_PARAMETERS defined in prompt.py, then forwards messages to the Gemini API. Lines 70-78 handle the response parsing and format conversion to match the application's internal message schema.

Configuration Examples

Basic Environment Setup

# From the repository root

cp main/.env.example .env

# Edit .env to include:

# LLM_PROVIDER=gemini

# GEMINI_API_KEY=AIzaSy...

# DEFAULT_MODEL=gemini-2.5-pro

Running the Agent with Gemini

python -m main.evaluate path/to/resume.json

The script will automatically instantiate GeminiProvider using your configured API key and model.

Programmatic Configuration

If you prefer not to use a .env file, set variables before importing the agent modules:

import os

os.environ["LLM_PROVIDER"] = "gemini"
os.environ["DEFAULT_MODEL"] = "gemini-2.5-flash"
os.environ["GEMINI_API_KEY"] = "AIzaSy..."

# Import after environment setup

from main.evaluate import evaluate_resume

This works because prompt.py reads environment variables at import time.

Troubleshooting

API Key Validation

The Hiring Agent validates the presence of the Gemini API key before attempting API calls. If GEMINI_API_KEY is empty or unset, the application raises an error early in the initialization process (handled in llm_utils.py and prompt.py line 66-68). Ensure your .env file is in the correct directory and that the variable name matches exactly.

Summary

• Set GEMINI_API_KEY in your .env file or environment to authenticate with Google Gemini.
• Set LLM_PROVIDER=gemini to switch from the default Ollama backend.
• Choose a Gemini model via DEFAULT_MODEL (e.g., gemini-2.5-pro); the system automatically maps these to the Gemini provider in main/prompt.py lines 36-44.
• Install dependencies including google-generativeai to support the GeminiProvider class in main/models.py.
• Validation occurs early: The application checks for the API key before initializing the provider, preventing silent failures.

Frequently Asked Questions

What happens if I don't set GEMINI_API_KEY?

The Hiring Agent will raise a validation error during initialization. According to the source code in main/prompt.py (lines 66-68) and validation logic in llm_utils.py, the application explicitly checks for the presence of the API key before constructing the GeminiProvider instance. Without this key, the agent cannot authenticate with Google's API.

Can I use other Gemini models like gemini-2.5-flash?

Yes. The model mapping in main/prompt.py (lines 55-63) supports any model name prefixed with gemini-. Simply set DEFAULT_MODEL=gemini-2.5-flash in your .env file. The system will resolve this to ModelProvider.GEMINI and pass the model name to the GenerativeModel constructor in main/models.py (lines 48-52).

How do I switch back to Ollama?

Change the LLM_PROVIDER environment variable to ollama or simply remove the GEMINI_API_KEY variable and ensure your chosen model name is one supported by Ollama. The provider selection logic in main/prompt.py maps model names to providers; if you select a non-Gemini model, the system will instantiate the Ollama provider instead.

Where does the Hiring Agent store the API key configuration?

The application does not store the API key in code; it reads from environment variables at runtime. In main/prompt.py, lines 12-18, the load_dotenv() function reads from .env files, and lines 66-68 retrieve the key using os.getenv("GEMINI_API_KEY"). This key is then passed to the GeminiProvider constructor in main/models.py (lines 13-21) without persisting to disk.

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