app.py

# --------------------------------------------------------------
# combined_s3_sql_vector_app.py
# Full Combined: S3 SQL + Vector MCP Tool in One Gradio App
# --------------------------------------------------------------
import boto3
import pandas as pd
import duckdb
import requests
from urllib.parse import quote
from io import StringIO
import gradio as gr

# === S3 Credentials ===
ENDPOINT_URL = "https://s3.us-west-1.idrivee2.com"
ACCESS_KEY = "rNuPBAQetemqpEeBospZ"
SECRET_KEY = "BU4FccUYxzXVqiWjPSJM1CWEX1cNhBqbU9NeGidE"
BUCKET = "accusagas3"

s3 = boto3.client(
    "s3",
    endpoint_url=ENDPOINT_URL,
    aws_access_key_id=ACCESS_KEY,
    aws_secret_access_key=SECRET_KEY,
)

# --------------------------------------------------------------
# Vector MCP Tool
# --------------------------------------------------------------
def query_vector_agent_calling(user_query: str, collection_name: str) -> str:
    base_url = "https://srivatsavdamaraju-mvp-2-0-deploy-all-apis.hf.space/qdrant/search"
    encoded_collection = quote(collection_name, safe="")

    url = f"{base_url}?collection_name={encoded_collection}&mode=hybrid"

    headers = {
        "accept": "application/json",
        "Content-Type": "application/json",
    }

    payload = {"query": user_query, "top_k": 5}

    try:
        response = requests.post(url, headers=headers, json=payload, timeout=30)
        response.raise_for_status()

        data = response.json()
        results = data.get("results") or data.get("result") or []

        if not results:
            return "No relevant context found."

        output = []
        for item in results:
            text = item.get("text") or item.get("payload", {}).get("text") or str(item)
            score = item.get("score", "?")
            output.append(f"Score: {score}\n{text}\n---")

        return "\n".join(output)

    except requests.exceptions.Timeout:
        return "Vector API timeout."
    except requests.exceptions.HTTPError as e:
        return f"HTTP Error: {e.response.status_code}"
    except Exception as e:
        return f"Unexpected Error: {str(e)}"


# --------------------------------------------------------------
# SQL Query Tool (S3 → DuckDB)
# --------------------------------------------------------------
def run_sql(path: str, sql: str) -> pd.DataFrame:
    try:
        obj = s3.get_object(Bucket=BUCKET, Key=path)
        df = pd.read_csv(StringIO(obj["Body"].read().decode("utf-8")))
    except Exception as e:
        return pd.DataFrame({"error": [str(e)]})

    if df.empty:
        return pd.DataFrame({"error": ["Empty CSV"]})

    for col in df.columns:
        if any(x in col.lower() for x in ["price", "volume", "amount"]):
            df[col] = pd.to_numeric(df[col].astype(str).str.replace(r"[^\d.-]", "", regex=True), errors="coerce")

    con = duckdb.connect(":memory:")
    con.register("data", df)

    if not sql.strip().lower().startswith(("select", "with")):
        con.close()
        return pd.DataFrame({"error": ["Only SELECT allowed"]})

    try:
        result = con.execute(sql).df()
    except Exception as e:
        if "VARCHAR" in str(e):
            import re
            col = re.search(r"column ([a-zA-Z0-9_]+)", str(e))
            if col and (c := col.group(1)) in df.columns:
                sql = sql.replace(c, f"CAST({c} AS DOUBLE)")
                result = con.execute(sql).df()
            else:
                con.close()
                return pd.DataFrame({"error": [str(e)]})
        else:
            con.close()
            return pd.DataFrame({"error": [str(e)]})
    finally:
        con.close()

    return result.head(10000)


# --------------------------------------------------------------
# Combined Gradio App (with MCP enabled)
# --------------------------------------------------------------
with gr.Blocks() as app:
    gr.Markdown("# 🔥 Combined S3 SQL + Vector MCP Tool")

    gr.Markdown("## 📌 SQL Query on S3 CSV Files")
    with gr.Row():
        path = gr.Textbox(label="S3 Path", placeholder="folder/file.csv")
        sql = gr.Textbox(label="SQL Query", lines=3, placeholder="SELECT * FROM data LIMIT 10")
    btn_sql = gr.Button("Run SQL Query")
    out_sql = gr.Dataframe()

    gr.Markdown("---\n## 🔍 Vector Search MCP Tool")
    with gr.Row():
        user_query = gr.Textbox(label="Query", placeholder="Explain gold market trends")
        collection_name = gr.Textbox(label="Collection Name", placeholder="gold&silver-db")
    btn_vec = gr.Button("Run Vector Search")
    out_vec = gr.Textbox(label="Vector Output", lines=10)

    btn_sql.click(run_sql, [path, sql], out_sql)
    btn_vec.click(query_vector_agent_calling, [user_query, collection_name], out_vec)


if __name__ == "__main__":
    app.launch(server_name="0.0.0.0", server_port=7860, mcp_server=True)