refactor(api): add LLM abstraction layer to replace hardcoded model strings

feat(run): add retry mechanism with error handling to agent entry point

feat(agent): add answer extraction module and integrate into run function

chore(agent): remove verbose research complete banner from web search output

refactor(agent): split websearch agent into separate modules

chore(agent): enhance supervisor prompts and increase websearch recursion limit

fix(websearch): add recursion limit and error handling for search agent

chore: remove __pycache__ bytecode files from version control

feat(agent): implement supervisor agent with tool routing and math solver