Remove GLM-4.7 model configuration and system prompts

Ensure all model response chunks are streamed without filtering

Add ability to disable thinking for specific AI models

Add native function calling and agentic loop for all models

Add function calling support and remove outdated processes