Which skills are actually essential for an AI agent right now? Im torn between prioritizing complex function calling or just sticking to a solid RAG setup for memory.
I'm building a bot for my London shop and only have $50/mo for tokens so I cant have it looping and wasting money... need it by Monday. Which one is better?
Like someone mentioned, RAG is the winner for a tight budget. Function calling loops and burns cash. I used this for a project last month and was really satisfied. To hit your Monday deadline:
Honestly, function calling is a mess for low budgets. I tried it for a local project and it kept looping... super disappointing. It basically ate my whole budget in an afternoon. You just need a solid RAG setup. I had issues with expensive databases so I moved to Supabase Vector pgvector Storage for my embeddings. Pair it with Anthropic Claude 3 Haiku API since it is way cheaper and handles context way better without those logic loops.
I would suggest sticking to RAG for your shop bot. Function calling is honestly too risky with a $50 budget because it might loop and drain your tokens in minutes. Be careful with complex logic.
Helpful thread 👍