genie-flow-invoker-ollama 0.0.0.dev0

Genie Flow Invoker Ollama

Ollama Invokers

This package contains Genie Flow invokers for different Ollama invocations.

Installing the ollama invoker

pip install genie-flow-invoker-ollama

Installing Ollama

To run models locally, you'll need to install Ollama. We recommend using the official native installer for your platform:

Install Ollama

This will set up the Ollama runtime and make the ollama command available in your terminal.

Once installed, you can start a model like:

ollama run llama3

For advanced users, Ollama also provides a Docker image: Ollama for Docker which you can use in containerized environments.

Select model and query

List all available ollama models run:

ollama list

To start using the selected model, create a meta.yaml and a prompt like described here: Create LLM templates

There are three different types of invokers available:

• OllamaChatInvoker, includes dialogue history in your prompt
• OllamaGenerateInvoker, includes base64-encoded images in your prompt, using the model defined in meta.yaml
• OllamaEmbedInvoker, vectorizes text using the embedding model specified in meta.yaml

Include base64 encoded images in prompt

For images to be included, the prompt template must be structured as follows, or else, the query will only contain plain text.

prompt: |
    Some prompt text
images:
    - {{ image_as_base64 }}