feat: replace CrewAI with direct Groq call — 10x faster

fix: reduce chunk size and limit to fit free tier

fix: reduce chunks to fit free tier token limit

fix: use llama-3.1-8b-instant correct model ID

fix: use llama-4-scout highest free tier TPM

fix: use smaller model for free tier

fix: update groq model name

fix: add fastembed for qdrant

fix: use chonkie semantic extra

fix: use pydantic field for client

fix: pydantic field declaration for file_path

fix: flexible BaseTool import in app.py

fix: flexible BaseTool import for any crewai version

fix: use crewai 0.28.0 stable

fix: remove streamlit pin let HF handle it

fix: add sentence-transformers for chonkie

fix: remove crewai_tools completely

fix: remove crewai-tools, use custom serper tool

fix: match crewai and crewai-tools to 1.14.6

feat: switch to Groq LLaMA 3 free API

fix: use stable crewai 0.80.0

fix: align crewai and crewai-tools versions

fix: fix crewai-tools version conflict

fix: pin dependency versions

fix: update space metadata

fix: add HuggingFace space metadata

feat: add NeuralDocs agentic RAG app

initial commit

Duplicate from streamlit/streamlit-template-space