import gradio as gr
from rapidfuzz import process, fuzz
import pandas as pd, re

DF = pd.read_csv("./food_purine_mcp_ready_v2.csv")
STOPWORDS = {...}

def _meaningful_tokens(text: str):
    words = re.split(r"\W+", text.lower())
    return [w for w in words if w and w not in STOPWORDS]

@gr.tools(                      # NEW – turns the fn into an MCP tool
    name="purine_ingredients_fuzzy_search",
    description="Return up to `k` food-items that fuzzy-match the query and their "
                "purine data (mg/100 g)."
)
def fuzzy_search(
    query: str,                 # required → will appear in the schema
    k: int = 5,
    cutoff: int = 75
) -> list[dict]:
    """
    Parameters
    ----------
    query : str
        Ingredient or dish name to look up.
    k : int, optional
        Maximum rows to return (default = 5).
    cutoff : int, optional
        Minimal RapidFuzz token-set ratio (default = 75).

    Returns
    -------
    list[dict]
        Each dict contains 'food', 'purines_mg_per_100g', and 'classification'.
    """

    # direct substring hits (best precision, no scoring needed)
    mask = DF["food"].str.contains(query, case=False, regex=False)
    direct = DF[mask]
    if not direct.empty:
        return direct.head(k).to_dict(orient="records")

    # token filter
    q_tokens = set(_meaningful_tokens(query))
    if not q_tokens:
        return []  # nothing to match on

    candidates = [
        food for food in DF["food"]
        if q_tokens & set(_meaningful_tokens(food))
    ]

    # fuzzy rank
    matches = process.extract(
        query, candidates,
        scorer=fuzz.token_set_ratio,
        score_cutoff=cutoff,
        limit=k
    )
    rows = DF[DF["food"].isin([m[0] for m in matches])]
    return rows.to_dict(orient="records")

# ---------- Minimal UI (optional) -------------------------------------
with gr.Blocks(title="Purine DB MCP") as demo:
    gr.Markdown("## Purine Lookup Tools (MCP-enabled)")
    with gr.Tab("Fuzzy search"):
        in2 = gr.Textbox(label="Fuzzy term")
        out2 = gr.JSON()
        in2.submit(fuzzy_search, in2, out2)

# ---------- Launch ----------------------------------------------------
demo.launch(mcp_server=True)