app.py

"""
Chat-with-my-experience  –  Gradio UI
author: Arash Nicoomanesh
"""
import gradio as gr
import traceback
import logging
from datetime import datetime
import os, urllib.request, pathlib
from processing import load_or_create_index
from qa import get_qa_chain, ask_chain   # ask_chain wraps .invoke for us

# ------------------------------------------------------------------
# DOWNLOAD MODEL ON FIRST START (no CMake build)
# ------------------------------------------------------------------
MODEL_URL = "https://huggingface.co/unsloth/gemma-2-it-GGUF/resolve/main/gemma-2-2b-it.q3_k_m.gguf"
MODEL_PATH = pathlib.Path(__file__).with_name("models") / "gemma-2-2b-it.q3_k_m.gguf"

MODEL_PATH.parent.mkdir(exist_ok=True)
if not MODEL_PATH.exists():
    print("⬇️  Downloading model (1.1 GB) …")
    urllib.request.urlretrieve(MODEL_URL, MODEL_PATH)
    print("✅ Model ready.")

# point our config to the downloaded file
import config
config.Config.MODEL_PATH = str(MODEL_PATH)

# ------------------------------------------------------------------
# logging (same as before)
# ------------------------------------------------------------------
logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s %(levelname)s %(message)s",
    handlers=[
        logging.StreamHandler(),
        logging.FileHandler("app_error.log", mode="a")
    ]
)
logger = logging.getLogger(__name__)

# ------------------------------------------------------------------
# one-time startup
# ------------------------------------------------------------------
logger.info("Loading FAISS index …")
vector_store = load_or_create_index()
logger.info("Building QA chain …")
qa_chain = get_qa_chain(vector_store)

# ------------------------------------------------------------------
# original chat function (no streaming, no spinner)
# ------------------------------------------------------------------
def chat(message: str, history: list):
    if not message or not message.strip():
        return "Please type a question."

    try:
        answer = ask_chain(qa_chain, message.strip(), return_sources=False)
        logger.info("Q: %s  |  A: %s", message, answer[:100])
        return answer
    except Exception as e:
        tb = traceback.format_exc()
        logger.error("QA error for question: %s\n%s", message, tb)
        return f"⚠️ Error running QA chain: {e}  (details in server log)"

# ------------------------------------------------------------------
# Gradio UI  –  back to classic tuple format
# ------------------------------------------------------------------
demo = gr.ChatInterface(
    fn=chat,
    title="Chat with my experience – Arash Nicoomanesh",
    description="Ask anything about my career, projects, or skills.",
    examples=[
        "What is your biggest AI project?",
        "Explain your clinical triage agent",
        "What LLM optimisation techniques have you used?",
        "How would you describe your overall AI/ML expertise?"
    ],
    cache_examples=False,
    # no streaming, no type="messages" → uses classic tuples
)

if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=True,
        show_api=False,
    )