AutoGem

Agent Framework for Gemini Pro

Running Tests

To run the tests for AutoGem, navigate to the root directory of the project and execute the following command:

For Python:

python -m unittest discover tests

For detailed documentation on system architecture, setup, and usage, refer to our documentation.

Using the GeminiClient Class

The GeminiClient class provides an interface to the Google Gemini Pro Models API and Google Gemini Vision Pro Models API. Here's how you can use it:

Setup

Before using the GeminiClient, ensure you have installed the necessary dependencies:

pip install google-cloud-gemini-pro-models google-cloud-gemini-vision-pro-models

You must also configure your Google Cloud authentication by setting the GOOGLE_APPLICATION_CREDENTIALS environment variable to the path of your service account key file:

export GOOGLE_APPLICATION_CREDENTIALS="/path/to/your/service-account-file.json"

Initializing the Client

To initialize the GeminiClient, provide your Google Cloud API key:

from src.gemini_client import GeminiClient

client = GeminiClient(api_key='YOUR_API_KEY')

Making Prediction Requests

To make prediction requests, use the predict_with_gemini_pro_models or predict_with_gemini_vision_pro_models methods. Provide the model name and the path to the image you wish to classify:

# Predict with Gemini Pro Models
results = client.predict_with_gemini_pro_models('model_name', 'path/to/image.jpg')

# Predict with Gemini Vision Pro Models
results = client.predict_with_gemini_vision_pro_models('model_name', 'path/to/image.jpg')

Interpreting the Results

The methods return a list of dictionaries, each representing a prediction result. Here's how to interpret these results:

for result in results:
    print(f"Category: {result['category']}, Confidence: {result['confidence']}, Bounding Box: {result['bounding_box']}")

Known Issues and Limitations

For detailed documentation on system architecture, setup, and usage, refer to our documentation.

• The current implementation does not support streaming predictions.
• Only prediction requests with single instances (images) are supported; batch predictions are not yet implemented.
• The API may impose limits on the number of requests per minute or other usage restrictions.