update model repo to FallacyHunter org

fix: use groq client directly for explain

fix: switch explain to Groq free tier

fix: switch to zephyr-7b-beta with chat_completion

fix: use text_generation instead of chat_completion for Mistral

initial deploy