Portal Designer — GeoSupply Streamlit Dashboard

Custom GeoSupply skill — Layer 0 admin portal + end-user intelligence portal

Portal Architecture

GeoSupply Portal (Streamlit)
  ├── Admin Portal (Layer 0)
  │   ├── Agent Health Monitor
  │   ├── Budget Dashboard
  │   ├── Override Manager
  │   ├── Audit Log Viewer
  │   └── System Controls
  └── Intelligence Portal (End Users)
      ├── Risk Score Dashboard
      ├── Supply Chain Map (India)
      ├── Supplier Risk Table
      ├── Weekly Report Viewer
      └── Alert Feed

Auth: JWT tokens (portal:read for users, portal:admin for admin) Framework: Streamlit >= 1.30.0 Location: src/geosupply/portal/

1. App Structure

python
1# src/geosupply/portal/app.py — main entry point
2
3import streamlit as st
4from geosupply.portal.pages import (
5    admin_health, admin_budget, admin_overrides,
6    intel_risk_map, intel_suppliers, intel_alerts
7)
8from geosupply.portal.auth import verify_jwt_token
9
10st.set_page_config(
11    page_title="GeoSupply AI",
12    page_icon="🌏",
13    layout="wide",
14    initial_sidebar_state="expanded",
15)
16
17# Auth gate
18token = st.session_state.get("jwt_token")
19if not token or not verify_jwt_token(token):
20    _show_login_page()
21    st.stop()
22
23# Role-based routing
24scopes = st.session_state.get("jwt_scopes", [])
25if "portal:admin" in scopes:
26    _render_admin_sidebar()
27else:
28    _render_user_sidebar()

2. Admin Portal Pages

Agent Health Monitor (admin_health.py)

python
1def render_health_page(health_agent):
2    st.title("🤖 Agent Health Monitor")
3
4    col1, col2, col3, col4 = st.columns(4)
5    with col1:
6        st.metric("Active Agents", "8/8", delta="0")
7    with col2:
8        st.metric("Pipeline SLA", "8.3 min", delta="-1.2 min", delta_color="normal")
9    with col3:
10        st.metric("Tasks/Hour", "142", delta="+12")
11    with col4:
12        st.metric("Error Rate", "0.7%", delta="-0.2%", delta_color="inverse")
13
14    # Agent state table
15    health_data = asyncio.run(health_agent.execute("check_all", {}))
16    df = pd.DataFrame([
17        {
18            "Agent": name,
19            "State": info["state"],
20            "Last Seen": info["last_seen"],
21            "Tasks Done": info["tasks_done"],
22            "Cost (₹)": f"₹{info['cost_inr']:.2f}",
23        }
24        for name, info in health_data["result"]["agents"].items()
25    ])
26
27    def color_state(val):
28        colors = {"IDLE": "green", "BUSY": "blue", "ERROR": "red", "RECOVERY": "orange"}
29        return f"color: {colors.get(val, 'black')}"
30
31    st.dataframe(df.style.applymap(color_state, subset=["State"]), use_container_width=True)

Budget Dashboard (admin_budget.py)

python
1def render_budget_page(budget_agent):
2    st.title("💰 Budget Dashboard")
3
4    status = asyncio.run(budget_agent.execute("get_status", {}))["result"]
5    projection = asyncio.run(budget_agent.execute("get_projection", {}))["result"]
6
7    # Gauge: monthly burn
8    monthly_pct = (status["monthly_spend_inr"] / 500) * 100
9    col1, col2 = st.columns(2)
10    with col1:
11        st.metric(
12            "Monthly Spend",
13            f"₹{status['monthly_spend_inr']:.2f}",
14            delta=f"₹{500 - status['monthly_spend_inr']:.2f} remaining",
15        )
16        st.progress(monthly_pct / 100, text=f"{monthly_pct:.1f}% of ₹500 cap")
17
18    with col2:
19        if projection["days_until_exhaustion"]:
20            st.metric("Days Until Cap", f"{projection['days_until_exhaustion']:.0f} days")
21        st.metric("Burn Rate", f"₹{projection['burn_rate_inr_per_day']:.2f}/day")
22
23    # Tier breakdown bar chart
24    tier_data = status["spend_by_tier"]
25    if tier_data:
26        st.bar_chart(tier_data)
27
28    # Alert banner
29    if status["tier3_blocked"]:
30        st.error("🚨 Tier-3 calls BLOCKED — daily critical threshold exceeded")
31    elif status["daily_spend_inr"] >= 250:
32        st.warning("⚠️ Daily spend warning threshold reached")

3. Intelligence Portal Pages

Risk Score Dashboard (intel_risk_map.py)

python
1def render_risk_map():
2    st.title("🌏 India Supply Chain Risk Map")
3
4    # Region filter
5    selected_regions = st.multiselect(
6        "Filter Regions",
7        options=list(INDIA_RISK_REGIONS.keys()),
8        default=list(INDIA_RISK_REGIONS.keys()),
9    )
10
11    # Risk score cards
12    cols = st.columns(3)
13    for i, region in enumerate(selected_regions[:6]):
14        with cols[i % 3]:
15            score = _get_region_risk_score(region)
16            color = "🟢" if score < 0.3 else "🟡" if score < 0.55 else "🔴"
17            st.metric(
18                f"{color} {region}",
19                f"{score:.0%} risk",
20                help=f"Composite risk score for {region}",
21            )
22
23    # India map (plotly choropleth)
24    st.plotly_chart(_build_india_choropleth(selected_regions), use_container_width=True)

Supplier Risk Table (intel_suppliers.py)

python
1def render_supplier_table():
2    st.title("🏭 Supplier Risk Assessment")
3
4    # Search + filter
5    search = st.text_input("Search supplier name")
6    action_filter = st.selectbox("Filter by recommendation", ["ALL", "MONITOR", "DUAL_SOURCE", "REPLACE", "BLOCK"])
7
8    suppliers = _get_supplier_scores(search=search, action=action_filter)
9
10    # Color-coded table
11    df = pd.DataFrame(suppliers)
12    action_colors = {
13        "MONITOR":    "background-color: #d4edda",
14        "DUAL_SOURCE":"background-color: #fff3cd",
15        "REPLACE":    "background-color: #f8d7da",
16        "BLOCK":      "background-color: #842029; color: white",
17    }
18
19    def highlight_action(row):
20        return [action_colors.get(row["recommended_action"], "")] * len(row)
21
22    st.dataframe(
23        df.style.apply(highlight_action, axis=1),
24        use_container_width=True,
25        height=400,
26    )
27
28    # Export
29    if st.button("Export to CSV"):
30        st.download_button("Download CSV", df.to_csv(index=False), "suppliers.csv")

4. Override Manager (admin_overrides.py)

python
1def render_overrides_page(override_agent):
2    st.title("⚙️ Manual Overrides")
3    st.warning("All overrides are logged, HMAC-signed, and reviewed by LoopholeHunter.")
4
5    # Recent override history
6    st.subheader("Recent Overrides")
7    overrides = asyncio.run(override_agent.execute("list_recent", {"limit": 20}))["result"]
8    st.dataframe(pd.DataFrame(overrides), use_container_width=True)
9
10    # New override (admin:portal scope required)
11    st.subheader("Submit Override")
12    with st.form("override_form"):
13        action = st.selectbox("Override Action", [
14            "force_approve_spend",
15            "retrigger_worker",
16            "clear_circuit_breaker",
17            "escalate_task_priority",
18        ])
19        reason = st.text_area("Reason (required)", max_chars=500)
20        submitted = st.form_submit_button("Submit Override")
21        if submitted:
22            if not reason.strip():
23                st.error("Reason is required for all overrides")
24            else:
25                result = asyncio.run(override_agent.execute("submit", {
26                    "action": action, "reason": reason,
27                    "admin_jwt": st.session_state["jwt_token"],
28                }))
29                st.success(f"Override submitted: {result['override_id']}")

5. Streamlit Worker (StreamlitWorker)

python
1class StreamlitWorker(BaseWorker):
2    """
3    Prepares aggregated dashboard data for Streamlit portal.
4    Tier: CPU_ONLY — just data aggregation, no LLM.
5    """
6    name = "streamlit"
7    tier = LLMTier.CPU_ONLY
8
9    async def process(self, input_data: dict) -> dict:
10        view = input_data.get("view", "risk_overview")
11        data_funcs = {
12            "risk_overview":   self._aggregate_risk_data,
13            "supplier_table":  self._aggregate_supplier_data,
14            "budget_status":   self._aggregate_budget_data,
15            "agent_health":    self._aggregate_health_data,
16            "alert_feed":      self._aggregate_alert_data,
17        }
18        if view not in data_funcs:
19            return self._error("VALIDATION", f"Unknown view: {view}", input_data)
20
21        result = await data_funcs[view](input_data)
22        return {"result": result, "meta": {"cost_inr": 0.0}}

6. Portal File Structure

src/geosupply/portal/
  ├── __init__.py
  ├── app.py              # Main Streamlit app
  ├── auth.py             # JWT verification, session management
  ├── pages/
  │   ├── admin_health.py
  │   ├── admin_budget.py
  │   ├── admin_overrides.py
  │   ├── admin_audit_log.py
  │   ├── intel_risk_map.py
  │   ├── intel_suppliers.py
  │   ├── intel_alerts.py
  │   └── intel_reports.py
  ├── components/
  │   ├── risk_gauge.py   # Reusable risk score gauge
  │   ├── india_map.py    # Plotly India choropleth
  │   └── cost_chart.py   # Budget visualization components
  └── data/
      └── india_geojson.json   # India states GeoJSON for maps

7. Portal Launch

bash
1# Run portal (development)
2PYTHONPATH=src streamlit run src/geosupply/portal/app.py \
3  --server.port 8501 \
4  --server.address 0.0.0.0

Related Skills

• budget-controller — Budget data displayed in admin portal
• product-manager — User stories driving portal features
• india-intel — India risk map data source
• supply-chain-analyst — Supplier table data source