SUSE AI Ollama API Examples

Installation

To set up the project, follow these steps:

1. Create a virtual environment:

   python3 -m venv venv

2. Activate the virtual environment:

   source venv/bin/activate

3. Install the required packages:

   pip install -r requirements.txt

Setting Up the Environment

Before running the application, you need to create a .env file to configure the necessary environment variables.

1. Create a .env file in the root of your project directory.

2. Add the following content to the .env file:

   OLLAMA_BASE_URL=http://localhost:11434
   DEFAULT_MODEL=llama2:7b
   OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
   

   You can change the value to point to your specific Ollama server if it is running on a different host or port.

Usage

You can run the Ollama chat application using the following command:

python3 ./ollama-chat.py --prompt "Explain to me what Linux is."

Arguments:

• -h, --help: Show this help message and exit.
• -p, --prompt: The prompt to send to Ollama (required).
• -b, --base-url: The base URL for the Ollama server (default from .env).
• -m, --model: The model to use (default from .env).

Example

To run the application with a specific model and prompt, use:

python3 ./ollama-chat.py --prompt "What is quantum computing?"

SUSE Observability OpenTelemetry Collector

Configure the OTEL_EXPORTER_OTLP_ENDPOINT in the .env file to point to your SUSE Observability OpenTelemetry Collector. See https://docs.stackstate.com/open-telemetry/collector and https://docs.stackstate.com/open-telemetry/languages/sdk-exporter-config for more information.

Sandbox OpenTelemetry Collector with Docker

Use the following command to start a sandbox collector in Docker. This will start the OpenTelemetry Collector and Jaeger. The application will send traces to the collector, which will then forward them to Jaeger for visualization. To view the traces, access the Jaeger UI at http://localhost:16686 in your web browser.

cd docker && docker-compose up

Running the Application with OpenTelemetry Collector

Run the application with the following command to instrument the application and send traces to the collector:

opentelemetry-instrument --traces_exporter console python3 ollama-chat.py --prompt "What is quantum computing?"

For more detailed logs, you can use the following commands to view the live logs from the collector:

# live logs
docker-compose logs -f otel-collector

# all logs
docker-compose logs otel-collector

# last minute
docker-compose logs --since 1m otel-collector