app.py

import requests, pandas as pd
from transformers import pipeline
from datetime import datetime, timedelta
import gradio as gr

# ============================================================
# 1️⃣ FRED API — экономические данные США1
# ============================================================
API_KEY = "YOUR_FRED_API_KEY"  # зарегистрируй на https://fred.stlouisfed.org/
FRED_URL = "https://api.stlouisfed.org/fred/series/observations"

INDICATORS = {
    "GDP": "GDP",
    "Inflation (CPI)": "CPIAUCSL",
    "Unemployment Rate": "UNRATE",
    "Interest Rate (Fed Funds)": "FEDFUNDS"
}

def fetch_fred_data(series_id, start="2024-01-01", end=None):
    if end is None:
        end = datetime.now().strftime("%Y-%m-%d")
    params = {
        "series_id": series_id,
        "api_key": API_KEY,
        "file_type": "json",
        "observation_start": start,
        "observation_end": end
    }
    r = requests.get(FRED_URL, params=params)
    data = r.json().get("observations", [])
    df = pd.DataFrame(data)
    if not df.empty:
        df["value"] = pd.to_numeric(df["value"], errors="coerce")
    return df

# ============================================================
# 2️⃣ Генерация сводки (LLM)
# ============================================================
generator = pipeline("text2text-generation", model="google/flan-t5-base")

def generate_summary(topic):
    sid = INDICATORS.get(topic)
    df = fetch_fred_data(sid)
    if df.empty:
        return "⚠️ No data available for this indicator."
    last = df.tail(5)
    trend = last["value"].pct_change().mean() * 100
    context = f"Recent values of {topic}:\n{last[['date','value']].to_string(index=False)}\n\nAverage change: {trend:.2f}%"
    prompt = f"Provide an analytical summary of the trend:\n{context}"
    summary = generator(prompt, max_new_tokens=150)[0]["generated_text"]
    return summary, df

# ============================================================
# 3️⃣ Экспорт CSV → Power BI
# ============================================================
def export_csv(df, topic):
    filename = f"powerbi_{topic.replace(' ', '_').lower()}.csv"
    df.to_csv(filename, index=False)
    return filename

# ============================================================
# 4️⃣ Gradio UI
# ============================================================
with gr.Blocks(title="🏦 RAG Financial Analytics → Power BI") as app:
    gr.Markdown("## 🏦 Financial RAG: FRED API → LLM → Power BI")

    topic = gr.Dropdown(list(INDICATORS.keys()), label="Select Indicator", value="Inflation (CPI)")
    out_summary = gr.Textbox(label="Generated Summary", lines=8)
    out_file = gr.File(label="Power BI Export")

    def run_pipeline(topic):
        summary, df = generate_summary(topic)
        file_path = export_csv(df, topic)
        return summary, file_path

    gr.Button("Generate & Export").click(run_pipeline, inputs=topic, outputs=[out_summary, out_file])

app.launch()