Using prompto with Gemini
For prompts to Gemini API, you can simply add a line in your experiment (.jsonl) file where you specify the api to be gemini. See the models doc for some details on the environment variables you need to set.
Note that the Gemini API is different to the Vertex AI API. For Vertex AI API, see the vertexai example.
We provide an example experiment file in data/input/gemini-example.jsonl. You can run it with the following command (assuming that your working directory is the current directory of this notebook, i.e. examples/gemini):
Multimodal prompting
Multimodal prompting is available with the Gemini API. We provide an example notebook in the Multimodal prompting with Vertex AI notebook and example experiment file in data/input/gemini-multimodal-example.jsonl. You can run it with the following command:
Environment variables
To run the experiment, you will need to set the following environment variables first:
You can also use an .env file to save these environment variables without needing to export them globally in the terminal:
By default, the prompto_run_experiment command will look for an .env file in the current directory. If you want to use a different .env file, you can specify it with the --env flag.
Also see the gemini.ipynb notebook for a more detailed walkthrough on the how to set the environment variables and run the experiment and the different types of prompts you can run.
Do note that when you run the experiment, the input file (data/input/gemini-example.jsonl) will be moved to the output directory (timestamped for when you run the experiment).