[feat] Validate and moderate chat with LLama

[pkarw]

To ensure safety standards and make sure the conversation doesn't go out of control we could periodically send the conversation to local LLama prompting for example: "Check if the conversation doesn't contain any potentially harmful suggestions or is not replacing the doctors role; it should be informational only"

Then we can use similar prompt to verify if the diagnosis set in the original documents is in a line with the suggestions and if not to suggest a second opinion or something