Skip to content

dashboard.md

Latest commit

 

History

History
107 lines (73 loc) · 2.78 KB

dashboard.md

File metadata and controls

107 lines (73 loc) · 2.78 KB
Raw

Dashboard Guide

spanIQ V4 ships a local Streamlit dashboard — no account, no server, no cloud. It reads directly from your spaniq.db SQLite file.

Launch

pip install "spaniq[dashboard]"
spaniq dashboard
# Opens http://localhost:8501

Custom DB path or port:

spaniq dashboard --db /path/to/spaniq.db --port 8502

First launch note: Streamlit downloads frontend assets on first run (~30s). Subsequent launches are instant.

Pages

Overview

The landing page. Shows at a glance:

  • Total traces ingested — count from the timeline store
  • Components monitored — distinct pipeline components with data
  • Active alerts — total alerts in the alerts table
  • Per-component health badges — 🟢 healthy / 🟡 mild drift / 🔴 significant drift
  • Ingestion sparkline — recent trace activity

Drift Timeline

Interactive time-series charts for each metric, per component.

  • Component selector — filter to one component or view all
  • Metric selector — ResponseDrift, SemanticSimilarity, OutputStability, Consistency
  • Last N traces slider — control how much history to show (20-1000)
  • Threshold lines — dashed lines show where alerts fire
  • Auto-refresh toggle — re-queries SQLite every 10 seconds when monitoring is live

Attribution

Runs V3's changepoint/cascade analysis interactively.

  • Set Traces to analyse, PELT penalty, and Warmup traces
  • Click Run Attribution
  • See the cascade timeline chart: 🔴 root cause, 🟡 cascade, 🟢 healthy
  • Component details table shows break index and lead gap

Alert Log

Historical table of all fired alerts.

  • Filter by component, metric, and max rows
  • All columns sortable in the Streamlit dataframe
  • Export CSV button — download the full alert history

Configuration

DashboardConfig in src/spaniq/dashboard/config.py:

@dataclass
class DashboardConfig:
    db_path: str = "spaniq.db"        # SQLite file location
    refresh_interval: int = 10         # seconds between auto-refreshes
    page_title: str = "spanIQ Dashboard"
    max_rows: int = 500                # max rows per query

Architecture

The dashboard is read-only. It uses the same TimelineStore API as the V2/V3 CLI. No new storage, no new API layer, no ports.

Streamlit app (browser)
    ↓ reads
TimelineStore (SQLite) ← written by Monitor / OTelCollector

This means the dashboard works even if monitoring is stopped — it's a view of historical data.

Running alongside monitoring

# Terminal 1: collect and monitor
spaniq collect-otel --baseline my_baseline

# Terminal 2: view live data
spaniq dashboard --db spaniq.db

Enable Auto-refresh in the sidebar to see live updates as traces arrive.