Codeqai enables semantic search and conversational interaction with your codebase directly from the CLI. It generates fine-tuning datasets in Alpaca, conversational, instruction, or completion formats to train custom code-focused LLMs. The tool maintains a FAISS vector store synchronized with code changes, supports fully local execution via llama.cpp and Ollama, and integrates with OpenAI, Azure OpenAI, and Anthropic APIs. It uses Treesitter for code parsing and langchain for orchestration.