Context only tracks assistant

[dillonroach]

narrator/narrator.py

Line 94 in 4bab104

script = script + [{"role": "assistant", "content": analysis}]

I'm not sure the best way to include user prompts in the messages history here, since you don't want to include the actual image every time, but a placeholder of some sort that shows the LLM was prompted into the given response may help consistency and avoid an 'intro' every time, I'm not certain.

Something like:
{
"role": "user",
"content": [
{"type": "text", "text": "Describe this image (user uploaded image)"},
],
},