[PARKED] feature: Image Description With AI

[panaC]

Based on https://github.com/edrlab/thorium-reader/pull/new/feat/image-click

[panaC]

OPENAI_API_KEY=xxxxx npm run start:dev

[panaC]

OPENAI_API_KEY=xxxxx MISTRAL_API_KEY npm run start:dev

First question from openai=gpt-40-mini

second question "french translation" with mistralAI-pixtral-12b

we can mix AI providers as long as message history is preserved

[panaC]

You can enable the feature under _AI_FEATURE_ and IS_DEV flag

declare const __AI_FEATURE__: boolean;
export const _AI_FEATURE = __AI_FEATURE__;
export const IS_AI_FEATURE = IS_DEV && _AI_FEATURE;

[danielweck]

maybe useful resource?
https://github.com/sul-rds/avila/blob/main/prompts/menuez_qwen2.5-VL.yaml
from:
https://www.choice360.org/libtech-insight/creating-alt-text-for-digital-collections-with-ai-a-case-study/

[danielweck]

let's not merge this branch to develop yet, the additional translated labels will really confuse contributors, and there are non-localized strings still hard coded (system prompt etc.)

[danielweck]

...I am also reviewing the build process with terser and define webpack plugins for dead code elimination, which makes sense for a feature flag

[panaC]

let's not merge this branch to develop yet, the additional translated labels will really confuse contributors, and there are non-localized strings still hard coded (system prompt etc.)

I think too, let's merge this PR after the Th3.2 release