Integrate image understanding pipeline

[dolfim-ibm]

As part of the enrichment pipeline we want to leverage multi-modal vision models for the analysis of images in documents.

For example:

• Charts
• UML diagrams
• and more

An initial prototype is in #25, which will be re-implemented on top of the stronger v2 pipelines.

Runtime

The system will support:

1. Prompting a model served as API, e.g. using the openai vision api
2. Launching a local model, e.g. using vllm