fix: safely parse brainstorm_epochs value with parseInt

fix: handle None complexity data from LlamaCpp

fix: normalize llamacpp URL to handle cached values with /chat/completions

fix: remove llamacpp API key field (not required)

feat: replace llama-cpp wrapper with OpenAI client

fix: save provider selection to localStorage immediately on change

testing

feat: implement multi-anchor distillation support with dynamic UI

feat: release beta with distillation and perplexity chart

docs: update readme to prefer openrouter and llamacpp over ollama

feat: grok openrouter embeddings, rag debug fix, flux eye fix

Fix Brainstorming connection errors, standardize debug parameters, and improve code rendering and result routing.

feat: Add LlamaCpp provider, NodeChat UI, and secular QNN attributes

Fix: Handle 'Graph started' response in Brainstorming mode

fix(app): correct epoch loop condition to prevent extra pass