fix: Switch to FAISS as default and use /tmp for persist directory

fix: Add detailed logging and error handling for source ingestion

feat: Add HTTP ZIP download fallback for GitHub repos

fix: Use /tmp directory for Hugging Face compatibility

feat: Add Groq-optimized prompts, vector DB fallback, improved ignore patterns

feat: Add runtime Gemini model switching on rate limits - automatically tries next model

feat: Add Gemini 3 preview models to fallback list

feat: Add Gemini model fallback logic - tries multiple models in sequence

fix: Add ChromaDB auto-recovery from corruption

fix: Use gemini-2.5-flash as default (confirmed working)

fix: Add models/ prefix to Gemini model names

feat: Add local embeddings to bypass API rate limits

fix: Simplify __init__.py to fix import error

docs: Organize project with clear README and clean exports

chore: Remove unused files (rate_limit_config, setup.py, ingestor, cli)

feat: Add MCP + CrewAI integration with multi-mode interface

Refactor: Upgraded to Agentic Chatbot with AST & Call Graph support