Feature: Prompt & response saving

[jvm123]

Prompt & response logging capability was added to controller.py, evaluator.py and database.py.
The resulting data is saved in the programs/<id>.json.

The visualization UI was improved for easy prompt & response viewing.

Internals

• database.py:
  • A new log_prompt() method stores prompts & their optional response in memory.
  • The save_program() method saves all prompts & responses for a program to the ["prompts"] key in the programs json file.
• evaluator.py
  • _llm_evaluate() stores its generated prompts and the LLMEnsemble responses
• controller.py
  • run() stores prompts and the LLMEnsemble responses
• config.py provides a Database: log_prompts setting

Usage

The config.yaml thereby now offers a setting to turn the prompt logging feature on and off, default is on. Accordingly, nothing has to be done to make use of this feature.

database:
    log_prompts: bool = True

Other included features

• Visualizer UI
  • Show list item number graphically
  • When user pans away and the graph appears to be empty, show a "Recenter" button

Other included fixes

• Fixed $ make help
• Visualizer UI:
  • Performance graph did not automatically refresh when new data were available
  • Graceful handling of duplicate program IDs in the UI

---

Note: This PR includes PR #68.