use preciz SDK from pip package instead of local path

fix XML tool call parser for malformed key-value pairs

handle XML-style tool calls for models without native function calling

switch to openrouter/owl-alpha model, move model name to env var

refactor: remove Google AI, switch to OpenRouter, improve service runner

seperate prompts

improved tool calling , prompt enhance

add tests

Integrate OpenRouter GLM model with RAG tool calling

Add /questions endpoint

Deploy RAG Pipeline with FastAPI and ML capabilities