Adding images into the 'prompting' of LLM Agent

[homiecal]

I have a workflow whereby I retrieve a certain document from GCS, and then need to answer questions on this document with respect to a user query. How do I insert this image into the context of the LLM Agent?

From my understanding:

• ArtifactService could be used but it is unclear whether the core LLM Agent (defined with instructions) has context for the loaded document?
• I could call the vertex-ai sdk within a tool and get my response this way, but this seems to negate the benefit of using LLM Agents to perform the logic.

[IAmNo1Special]

@homiecal food you ever figure this out

[boyangsvl]

You can use load_artifacts tool. Or look at its code and insert the file content yourself to the request: https://github.com/google/adk-python/blob/main/src/google/adk/tools/load_artifacts_tool.py