Annotation Controls

YPhotoSharing supports several optional annotation streams that enrich the simulation with photo-specific metadata and interaction history.

Supported Toggles

Place these keys under simulation in client_config.json:

{
  "simulation": {
    "enable_emotion_annotation": true,
    "enable_sentiment": true,
    "enable_memory_annotations": true
  }
}

That same block can be combined with the logging flags that expose the debugging streams used by the client:

{
  "logging": {
    "enable_action_log": true,
    "enable_prompt_log": true,
    "enable_llm_usage_log": true
  }
}

enable_emotion_annotation: stores the dominant visual emotion for new photos in photo_emotions
enable_sentiment: stores a numeric sentiment score for photos and comments
enable_memory_annotations: writes memory events and makes the client load memory context into comments and DMs

Keep enable_emotion_annotation on when you want the platform to preserve visual labels attached to generated photos.
Keep enable_sentiment on if you want downstream analytics or moderation to exploit text polarity.
Keep enable_memory_annotations on if you want agents to condition comments and direct messages on prior interactions.

logging.enable_action_log enables per-client action summaries
logging.enable_prompt_log enables per-client prompt traces
logging.enable_llm_usage_log enables per-client LLM usage accounting
stress_reward.enabled enables the affective feedback loop
simulation.opinion_dynamics.enabled enables bounded-confidence or discrete LLM opinion updates

These toggles are orthogonal. You can enable them independently depending on how much annotation detail you want in an experiment.