Dashboard

Real-time overview across hotspots. Reporter safety first.

Active Streams

0

0 with high escalation

High-Escalation Alerts

0

0 APPROACH • 0 RETREAT

Recent Tips

0

Across current hotspots

Hotspot Pulse

0.0

Avg escalation (active)

Current Hotspots • Recent Protest Streams

Recent High-Value Alerts

Quick Actions

Live Monitor

Public event streams and coordination monitoring • Real-time intel + reporter safety flagging

Active Sources

→ Full Sources Directory (per-journalist profiles).
Video platforms: Overview • YouTube (primary) • Twitch • Facebook • Instagram • TikTok

Flagging here updates the shared Alerts and Dashboard in real time (preview simulation).

Intel Library

RSS news, tips, verifications and other intel. For historical livestreams see the dedicated Archive page.

Past Livestream Archive

All historical protest livestream recordings for reference and knowledge. Browse the full collection of archived streams (distinct from the current live Dashboard catalog and Monitor). Click any card to view the video player, historical transcript, and summary/analysis directly on this page.

0

archived streams

These are reference copies of past livestreams. Video playback uses the same stable player as live streams. Data comes from research/past_streams.json (curated historical entries). Add more entries and re-run generate-data.py to expand the archive.

Tools

On-demand builders for records, tips, permits, and verification. All tie into the shared intel + alerts system.

Submit a Tip

Permits Helper

FOIA Toolkit

Fact Check

Alerts Center

Settings

Active Hotspots

All changes here are simulated in this preview. In the real site they would persist and affect discovery + tools. The button above triggers the same merge+re-render as the top bar (equivalent to running generate-data.py then refreshing data in-app).

Map & Timeline

Simple Geo View (Escalation Heat)

Pins colored by escalation. Click for focus.

Event Timeline (All Sources)

Appendix: Escalation Framework

Purpose

The APPROACH / MONITOR / RETREAT statuses (and associated ESCALATION_LEVEL) provide a consistent, rapid decision framework for reporter safety and intelligence value. They are produced by a standardized "v2 premium" analysis protocol applied to livestream transcripts, chat messages, RSS items, tips, and other incoming data.

Escalation Level (1-10)

1 — Calm / monitoring only

5 — Notable activity

7 — Escalating tensions, arrests, or violence

10 — Immediate danger or major event

Levels are assigned by the analysis model based on content, context, crowd dynamics, authority presence, and risk signals.

Recommended Action

APPROACH High-value intel for reporters. Situation warrants moving closer (with appropriate caution and team coordination) for better observation, documentation, or engagement.

MONITOR Routine or low-risk observation. Continue watching from a safe distance; no immediate need to advance or withdraw.

RETREAT Safety risk to on-ground team. Prioritize distance, extraction routes, or disengagement. Reporter safety is the primary consideration.

Supporting Elements

KEY_RISK_FLAGS: Comma-separated specific indicators (e.g. "police advancing, projectiles, crowd density >80, new armed group"). These heavily influence the action choice.
Full structured analysis: Every output also includes key facts, hypotheses, contradictions, ranked priorities, sources, and uncertainties. The escalation/action is the final "machine-readable" summary for alerting and flagging.
Auto-alerting: Backend tools trigger alerts and notifications when escalation ≥ 7 or when the action is APPROACH or RETREAT.

Human + Automated Use

Automated analysis from the monitoring tools (livestream listener, chat monitor, RSS aggregator, etc.) produces these values to feed the hub. In the Monitor view, reporters can manually apply APPROACH / MONITOR / RETREAT flags to any active stream using the large buttons. Manual flags update alerts in real time and are preserved across data refreshes where possible.

This framework prioritizes reporter safety first, followed by actionable intelligence value for the broader team.

The exact prompt language used in the analysis tools is based on the definitions above. Thresholds and examples are tuned for protest / public event environments.