Make LLAMA options API free

[benevolentbandwidth-nurtekin]

Research Agent currently supports paid API-based LLM providers. We also provide a free/local mode based on embeddings and heuristics.

We want to explore a stronger fully free LLM option using open-weight models (e.g., LLaMA variants or similar), without requiring paid API access.

Goals

1. Provide a no-cost LLM backend option
2. Avoid requiring external paid APIs
3. Maintain reasonable output quality