src/streamlit_app.py

import streamlit as st
import pandas as pd
import altair as alt
import os
import sys
import time



BASE_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
sys.path.append(BASE_DIR)

from inference import TelemetryInferenceEngineLite

DATA_PATH = os.path.join(BASE_DIR, "assets/test.csv")

st.set_page_config(page_title="Race Telemetry", layout="wide")


@st.cache_resource
def load_engine():
    return TelemetryInferenceEngineLite()

engine = load_engine()


@st.cache_data
def load_data():
    return pd.read_csv(DATA_PATH)

FULL_DATA = load_data()


if "cursor" not in st.session_state:
    st.session_state.cursor = 0

if "telemetry" not in st.session_state:
    st.session_state.telemetry = pd.DataFrame()

st.markdown("""
<style>
.ml-label {
    font-size: 20px;
    color: #94a3b8;
    margin-bottom: 0px;
}
.ml-value {
    font-size: 38px;       
    font-weight: 500;
    line-height: 2;
}
</style>
""", unsafe_allow_html=True)


st.sidebar.title("Pit Wall Controls")

auto_refresh = st.sidebar.toggle("Auto Refresh", value=True)
refresh_interval = st.sidebar.slider("Refresh Interval (seconds)", 1, 5, 1)
batch_size = st.sidebar.selectbox("Rows per fetch", [1, 5, 10], index=0)


def fetch_rows(batch_size):
    start = st.session_state.cursor
    end = start + batch_size

    batch = FULL_DATA.iloc[start:end].copy()
    st.session_state.cursor = end

    return batch


def run_inference(df_batch):
    outputs = []

    for _, row in df_batch.iterrows():
        prediction = engine.process_row(row)

        row_dict = row.to_dict()

        row_dict["predicted_lap_time"] = prediction.get("predicted_lap_time")
        row_dict["predicted_gear"] = prediction.get("predicted_gear")
        row_dict["driving_behavior"] = prediction.get("driving_behavior")

        outputs.append(row_dict)

    return pd.DataFrame(outputs)


new_data = fetch_rows(batch_size)

if not new_data.empty:
    processed = run_inference(new_data)

    st.session_state.telemetry = pd.concat(
        [st.session_state.telemetry, processed],
        ignore_index=True
    )

df = st.session_state.telemetry

if df.empty:
    st.stop()

df["t"] = range(len(df))
latest = df.iloc[-1]


st.markdown(
    """
    <div style="text-align:center; line-height:0;">
        <h2>🏁 Race Telemetry</h2>
        <h4>Pit Wall Dashboard</h4>
    </div>
    """,
    unsafe_allow_html=True
)


left, right = st.columns([3, 1])

with left:

    with st.container(border=True):
        c1, c2, c3 = st.columns(3)

        with c1:
            st.markdown('<div class="ml-label">Predicted Lap</div>', unsafe_allow_html=True)
            st.markdown(
                f'<div class="ml-value" style="color:#38bdf8;">{latest["predicted_lap_time"]:.2f} s</div>',
                unsafe_allow_html=True
            )

        with c2:
            st.markdown('<div class="ml-label">Recommended Gear</div>', unsafe_allow_html=True)
            st.markdown(
                f'<div class="ml-value" style="color:#22c55e;">{int(latest["predicted_gear"])}</div>',
                unsafe_allow_html=True
            )

        with c3:
            st.markdown('<div class="ml-label">Driving Style</div>', unsafe_allow_html=True)
            st.markdown(
                f'<div class="ml-value" style="color:#facc15;">{latest["driving_behavior"]}</div>',
                unsafe_allow_html=True
            )

    st.markdown("<br>", unsafe_allow_html=True)

    r2c1, r2c2 = st.columns(2)

    r2c1.metric("Speed (km/h)", f"{latest['speed']:.1f}")
    r2c1.altair_chart(
        alt.Chart(df).mark_line().encode(x="t:Q", y="speed:Q"),
        use_container_width=True
    )

    r2c2.metric("Engine RPM", int(latest["current_engine_rpm"]))
    r2c2.altair_chart(
        alt.Chart(df).mark_area(opacity=0.7).encode(x="t:Q", y="current_engine_rpm:Q"),
        use_container_width=True
    )

with right:
    st.markdown("### Track")
    st.image(os.path.join(BASE_DIR, "assets/track.png"), use_container_width=True)


p1, p2, p3, p4 = st.columns(4)

df["power_kw"] = df["power"] / 1000

p1.metric("Power (kW)", f"{df['power_kw'].iloc[-1]:.1f}")
p1.altair_chart(alt.Chart(df).mark_area().encode(x="t", y="power_kw"), use_container_width=True)

p2.metric("Torque (Nm)", f"{latest['torque']:.1f}")
p2.altair_chart(alt.Chart(df).mark_line().encode(x="t", y="torque"), use_container_width=True)

p3.metric("Boost (psi)", f"{latest['boost']:.2f}")
p3.altair_chart(alt.Chart(df).mark_line().encode(x="t", y="boost"), use_container_width=True)

p4.metric("Avg Tire Temp (°C)", f"{latest['avg_tire_temp']:.1f}")
p4.altair_chart(alt.Chart(df).mark_line().encode(x="t", y="avg_tire_temp"), use_container_width=True)


attitude_chart = alt.Chart(df).transform_fold(
    ["yaw", "pitch", "roll"],
    as_=["Axis", "Value"]
).mark_line().encode(
    x="t:Q",
    y="Value:Q",
    color="Axis:N"
)

st.metric(
    "Yaw / Pitch / Roll (rad)",
    f"{latest['yaw']:.2f}, {latest['pitch']:.2f}, {latest['roll']:.2f}"
)

st.altair_chart(attitude_chart, use_container_width=True)


if auto_refresh:
    time.sleep(refresh_interval)
    st.rerun()