path: root/streamlit_app.py

import os
from datetime import datetime, timezone

import pandas as pd
import plotly.graph_objects as go
import streamlit as st
from dotenv import load_dotenv
from streamlit_autorefresh import st_autorefresh

import aws
import persistance

load_dotenv()

st.set_page_config(page_title="Spot Crash Board", page_icon="📉", layout="wide")

st.markdown(
    """
    <style>
      .top-card {
        border: 1px solid #2a3142;
        border-radius: 14px;
        padding: 1rem 1.15rem;
        margin-bottom: 0.8rem;
        background: linear-gradient(145deg, #0d1320 0%, #121a2b 100%);
      }
      .top-label {
        color: #8fa6cc;
        font-size: 0.78rem;
        letter-spacing: 0.08em;
        text-transform: uppercase;
      }
      .top-value {
        font-size: 2.2rem;
        font-weight: 800;
        margin-top: 0.15rem;
        line-height: 1;
      }
      .top-sub {
        color: #adbacf;
        margin-top: 0.45rem;
        font-size: 0.95rem;
      }
      .mega-crash {
        border: 1px solid #7f1d1d;
        border-radius: 14px;
        background: linear-gradient(145deg, #2b0b0b 0%, #430f0f 100%);
        color: #fee2e2;
        padding: 1rem 1.15rem;
        margin-bottom: 0.8rem;
        animation: pulse 1.25s infinite;
      }
      .mega-crash strong {
        color: #fecaca;
      }
      @keyframes pulse {
        0% { box-shadow: 0 0 0 0 rgba(220, 38, 38, 0.55); }
        70% { box-shadow: 0 0 0 12px rgba(220, 38, 38, 0); }
        100% { box-shadow: 0 0 0 0 rgba(220, 38, 38, 0); }
      }
    </style>
    """,
    unsafe_allow_html=True,
)
max_pay_raw = os.getenv("MAX_PAY")
st.caption(f"i will not pay more than ${max_pay_raw} per hour")

if "simulate_mega_crash" not in st.session_state:
    st.session_state.simulate_mega_crash = False

with st.sidebar:
    st.header("Display")
    poll_seconds = st.slider("Poll interval (seconds)", min_value=2, max_value=30, value=5)
    history_limit = st.slider("Points on chart", min_value=100, max_value=5000, value=1000)
    chart_expanded = st.toggle("Chart expanded", value=True)
    show_markers = st.toggle("Show point markers", value=False)

    st.divider()
    st.caption("Test mode")

    if st.button("Trigger MEGA CRASH", type="primary", use_container_width=True):
        st.session_state.simulate_mega_crash = True

    if st.button("Clear test crash", use_container_width=True):
        st.session_state.simulate_mega_crash = False

if not max_pay_raw:
    st.error("Missing MAX_PAY environment variable. Add MAX_PAY to your .env file.")
    st.stop()

try:
    max_pay = float(max_pay_raw)
except ValueError:
    st.error(f"Invalid MAX_PAY value: {max_pay_raw}. Use a numeric value like 0.03")
    st.stop()

persistance.init_storage()
st_autorefresh(interval=poll_seconds * 1000, key="spot-pricing-poller")


@st.cache_data(ttl=1, show_spinner=False)
def fetch_spot_payload():
    return aws.get_asg_spot_pricing()


payload = fetch_spot_payload()
if payload.get("error"):
    st.error(payload["error"])
    st.stop()

latest = payload.get("latest") or {}
instance = payload.get("instance") or {}
history = payload.get("history") or []

if not history:
    if latest.get("spotPrice") is None:
        st.warning("No spot price returned from API right now.")
        st.stop()
    history = [latest]

imported_points = 0
for item in history:
    timestamp = item.get("timestamp")
    spot_price = item.get("spotPrice")

    if not timestamp or spot_price is None:
        continue

    point = {
        "polled_at": str(timestamp),
        "source_timestamp": str(timestamp),
        "asg_name": payload.get("asg_name"),
        "instance_id": instance.get("id"),
        "instance_type": instance.get("type"),
        "az": item.get("availabilityZone") or instance.get("az"),
        "spot_price": float(spot_price),
    }

    if persistance.save_spot_datapoint(point):
        imported_points += 1

rows = persistance.load_spot_datapoints(limit=history_limit)

if not rows:
    st.info("Waiting for first saved data point...")
    st.stop()

df = pd.DataFrame(rows)
df["polled_at"] = pd.to_datetime(df["polled_at"], errors="coerce", utc=True)
df["source_timestamp"] = pd.to_datetime(df["source_timestamp"], errors="coerce", utc=True)
df = df.dropna(subset=["polled_at", "spot_price"]).sort_values("polled_at")

if df.empty:
    st.info("No valid points available yet.")
    st.stop()

current_price = float(df.iloc[-1]["spot_price"])
prev_price = float(df.iloc[-2]["spot_price"]) if len(df) > 1 else current_price
visible_low = float(df["spot_price"].min())
visible_high = float(df["spot_price"].max())
price_delta = current_price - prev_price
multiplier = current_price / visible_low if visible_low > 0 else 1.0

on_demand_price = latest.get("onDemandPrice")
if on_demand_price is None:
    for item in reversed(history):
        if item.get("onDemandPrice") is not None:
            on_demand_price = item.get("onDemandPrice")
            break

if on_demand_price is not None:
    on_demand_price = float(on_demand_price)
    savings_per_hour = on_demand_price - current_price
    savings_percent = (savings_per_hour / on_demand_price) * 100 if on_demand_price > 0 else None
else:
    savings_per_hour = None
    savings_percent = None

peak_all_time = persistance.get_peak_spot_price()
first_breach = persistance.get_first_breach(max_pay)
historical_mega_crash = first_breach is not None
mega_crash = historical_mega_crash or st.session_state.simulate_mega_crash

if mega_crash:
    if historical_mega_crash:
        breach_time = first_breach.get("polled_at", "unknown")
        breach_price = float(first_breach.get("spot_price", 0.0))
        message = (
            '<strong>MEGA CRASH:</strong> MAX_PAY has been exceeded in your saved history. '
            f'First breach at <strong>{breach_time}</strong> '
            f'with <strong>${breach_price:.5f}/hr</strong> '
            f'(MAX_PAY = <strong>${max_pay:.5f}</strong>).'
        )
    else:
        message = (
            '<strong>MEGA CRASH (TEST):</strong> Simulated crash from sidebar button. '
            f'MAX_PAY is <strong>${max_pay:.5f}</strong>.'
        )

    st.markdown(f'<div class="mega-crash">{message}</div>', unsafe_allow_html=True)

trend_icon = "▲" if price_delta >= 0 else "▼"
trend_color = "#f59e0b" if price_delta >= 0 else "#22c55e"
if mega_crash:
    trend_color = "#ef4444"

if savings_percent is None or savings_per_hour is None:
    savings_text = "Savings vs on-demand: n/a"
elif savings_per_hour >= 0:
    savings_text = f"Savings vs on-demand: {savings_percent:.2f}% (${savings_per_hour:.5f}/hr)"
else:
    savings_text = f"Spot above on-demand by {abs(savings_percent):.2f}% (${abs(savings_per_hour):.5f}/hr)"

st.markdown(
    f"""
    <div class="top-card">
      <div class="top-label">Crash Multiplier</div>
      <div class="top-value" style="color: {trend_color};">{multiplier:.2f}x {trend_icon}</div>
      <div class="top-sub">
        Current: <strong>${current_price:.5f}/hr</strong> &nbsp; • &nbsp;
        On-demand: <strong>{f'${on_demand_price:.5f}/hr' if on_demand_price is not None else 'n/a'}</strong> &nbsp; • &nbsp;
        {savings_text} &nbsp; • &nbsp;
        MAX_PAY: <strong>${max_pay:.5f}</strong>
      </div>
    </div>
    """,
    unsafe_allow_html=True,
)

c1, c2, c3, c4, c5, c6, c7 = st.columns(7)
c1.metric("Instance", instance.get("type", "unknown"))
c2.metric("AZ", instance.get("az", "unknown"))
c3.metric("Spot ($/hr)", f"{current_price:.5f}", f"{price_delta:+.5f}")
c4.metric("On-demand ($/hr)", f"{on_demand_price:.5f}" if on_demand_price is not None else "n/a")

if savings_percent is not None and savings_per_hour is not None:
    c5.metric("Savings (%)", f"{savings_percent:.2f}%", f"${savings_per_hour:+.5f}/hr")
else:
    c5.metric("Savings (%)", "n/a")

c6.metric("24h estimate ($)", f"{current_price * 24:.2f}")
c7.metric("All-time peak ($/hr)", f"{(peak_all_time or current_price):.5f}")

with st.expander("Ticker chart", expanded=chart_expanded):
    line_color = "#ef4444" if mega_crash else ("#f59e0b" if price_delta >= 0 else "#22c55e")
    marker_mode = "lines+markers" if show_markers else "lines"

    fig = go.Figure()
    fig.add_trace(
        go.Scatter(
            x=df["polled_at"],
            y=df["spot_price"],
            mode=marker_mode,
            line={"width": 3, "color": line_color, "shape": "spline"},
            marker={"size": 5, "color": line_color},
            fill="tozeroy",
            fillcolor="rgba(245, 158, 11, 0.13)" if not mega_crash else "rgba(239, 68, 68, 0.18)",
            name="Spot price",
        )
    )

    fig.add_hline(
        y=max_pay,
        line_dash="dash",
        line_color="#ef4444",
        annotation_text=f"MAX_PAY ${max_pay:.5f}",
        annotation_position="top right",
        annotation_font_color="#ef4444",
    )

    fig.update_layout(
        template="plotly_dark",
        margin={"l": 10, "r": 10, "t": 10, "b": 10},
        hovermode="x unified",
        height=430,
        xaxis={
            "title": "Time (UTC)",
            "tickformat": "%H:%M:%S",
            "hoverformat": "%Y-%m-%d %H:%M:%S",
            "showgrid": False,
            "rangeslider": {"visible": False},
        },
        yaxis={
            "title": "Spot price ($/hr)",
            "tickprefix": "$",
            "tickformat": ".5f",
            "gridcolor": "#2c3448",
        },
        showlegend=False,
    )

    st.plotly_chart(fig, width="stretch")

status = f"Imported {imported_points} new historical points" if imported_points else "No new historical points"
st.caption(f"{status} • Last refresh {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}")

with st.expander("Recent datapoints", expanded=False):
    st.dataframe(df.tail(150).sort_values("polled_at", ascending=False), width="stretch")