Currently we feed the screenshots if certain keywords are detected, and if a game is detected then we always send the additional context from ChromaDB to Gemini. Instead of manually calling functions add tool calling for said functions so that they can be called by the LLM itself. Also add the following improvements:
- Allow more flexibility for gemini to ask for a certain number of screenshots and from a specified range of dates
- Allow more flexibility for for grabbing context from the vector db where it can ask for context multiple times and amount of context it needs.
- Plus modularity to expand functionality