Streamlit Dashboards

TL;DR

Skills that store data (SQLite, JSON logs) can include a Streamlit dashboard for visualization. The dashboard uses synchronous sqlite3 + pandas for read-only access, avoiding conflicts with the skill’s async runtime.

When to add a dashboard

Add a dashboard when your skill accumulates data over time — audit logs, scan results, monitoring metrics. The dashboard provides a visual interface for exploring that data.

Directory structure


my-skill/
├── skill/
├── src/my_skill/
├── dashboard/
│   ├── app.py              # Streamlit entry point
│   ├── db_reader.py        # Sync data access
│   ├── charts.py           # Reusable chart components
│   └── requirements.txt    # streamlit, pandas
└── data/                   # Runtime data (gitignored)
    └── my_skill.db

The async problem

Skills use async libraries (httpx, Tortoise ORM) for their runtime. Streamlit is synchronous and runs in a separate process. Sharing an async database connection is not possible.

Solution: use sqlite3 + pandas for read-only access in the dashboard. The skill writes to SQLite via Tortoise ORM; the dashboard reads via sqlite3.

Dashboard entry point

dashboard/app.py:


import streamlit as st
import pandas as pd
from db_reader import load_data
 
st.set_page_config(page_title="My Skill Dashboard", layout="wide")
 
# Guard: stop if no data
df = load_data()
if df.empty:
    st.info("No data yet. Run the skill first.")
    st.stop()
 
# Sidebar filters
with st.sidebar:
    status = st.selectbox("Status", ["all"] + df["status"].unique().tolist())
 
if status != "all":
    df = df[df["status"] == status]
 
# Main content
st.title("My Skill Results")
st.dataframe(df, use_container_width=True)

st.set_page_config() must be the first Streamlit call. Place it before any other st.* function.

Database reader

dashboard/db_reader.py:


import sqlite3
from pathlib import Path
import pandas as pd
 
DB_PATH = Path(__file__).resolve().parent.parent / "data" / "my_skill.db"
 
def load_data() -> pd.DataFrame:
    if not DB_PATH.exists():
        return pd.DataFrame()
    conn = sqlite3.connect(str(DB_PATH))
    df = pd.read_sql_query("SELECT * FROM results ORDER BY created_at DESC", conn)
    conn.close()
    return df

Derive DB_PATH from __file__ — never hardcode absolute paths.

Dashboard dependencies

dashboard/requirements.txt:


streamlit>=1.40.0
pandas>=2.0.0

Running the dashboard


pip install -r dashboard/requirements.txt
streamlit run dashboard/app.py

Opens at http://localhost:8501. Use --server.port 8502 for a custom port.

Auto-refresh

For monitoring dashboards that need live updates:


import time
 
placeholder = st.empty()
 
while True:
    df = load_data()
    with placeholder.container():
        st.dataframe(df, use_container_width=True)
    time.sleep(30)
    st.rerun()

Entry point in pyproject.toml

Optionally add a script entry point for convenience:


[project.scripts]
my-skill = "my_skill._skill:main"
my-skill-dashboard = "streamlit:run dashboard/app.py"