GeminiWebCamApp

Gemini provides a multimodal model (gemini-pro-vision) that accepts both text and images and inputs. The GenerativeModel.generate_content API is designed handle multi-media prompts and returns a text output.

This sample will read from your webcam you can type requests to the ai model in the text field and send them when you are ready. It shows how to use Python with the new gemini large language model for

to run this.

for the main ui

python app.py

to just pay with gemini text based you can run

python gemini.py

to just run chat

python gemini_chat.py

WIP tuning models

Its broken on googles end

python create_tuned_model.py

Required

TODO: update after the beta library has been released as a pip.

pip install PyQt5
pip install opencv-python
pip install python-dotenv
pip install google-generativeai
pip install google-ai-generativelanguage

Settings

The format of the .env file is

GOOGLE_APPLICATION_CREDENTIALS is path to the service account key file only needed for tuning

API_KEY=[readacted]
TEXT_MODEL_NAME=models/gemini-pro
IMAGE_MODEL_NAME=models/gemini-pro-vision
CHAT_MODEL_NAME=gemini-pro
GOOGLE_APPLICATION_CREDENTIALS=C:\Development\FreeLance\GoogleSamples\Credentials\gemini-407220-ea106e0d696d.json

error

If you see the following error try running a vpn.

400 User location is not supported for the API use.

Text based chat conversation calls must go as user, model, user, model. This error occurs if you dont send it in that format

400 Please ensure that multiturn requests ends with a user role or a function response.