Skip to content

WORKAROUND_websocket.md

Latest commit

 

History

History
84 lines (54 loc) · 2.93 KB

WORKAROUND_websocket.md

File metadata and controls

84 lines (54 loc) · 2.93 KB

WebSocket Server Crash-Loop Workaround

Issue: WebSocket server crashes during initialization, preventing dashboard from receiving graph data.

Confirmed by: Felix (04:49), Victor (04:43), Iris (04:52)

Root Causes (from SYNC.md)

  1. SafeMode tripwire - Server shuts down immediately due to telemetry emission failure
  2. SubEntity loading failure - Loaded 0 subentities from FalkorDB despite thousands existing (label case mismatch?)
  3. WebSocket connection bug - control_api.py:2808 WebSocket not accepted properly

Current Status

  • ✅ FalkorDB: Running with rich data (Iris: 689 nodes, Felix: 10 nodes, others: 1000s)
  • ✅ Dashboard: Running and rendering correctly
  • ❌ WebSocket Server: Crash-looping, not binding to port 8000
  • ❌ Graph visualization: No data flowing to dashboard

Temporary Workaround: Static Snapshot View

Since the WebSocket connection is broken, we can create a static snapshot viewer that reads directly from FalkorDB without needing the WebSocket server.

Option 1: REST API Fallback

Create a simple HTTP endpoint that serves graph snapshots:

# Add to dashboard API route: app/api/consciousness/snapshot/route.ts

import { FalkorDB } from 'falkordb';

export async function GET(request: Request) {
  const db = new FalkorDB({ host: 'localhost', port: 6379 });
  const graph = db.selectGraph('mind-protocol_iris');

  const nodes = await graph.query('MATCH (n) RETURN n LIMIT 100');
  const links = await graph.query('MATCH ()-[r]->() RETURN r LIMIT 100');

  return Response.json({ nodes, links });
}

Option 2: Direct FalkorDB Query (Python Script)

Extract graph snapshot and save as JSON for dashboard to load:

python3 -c "
from falkordb import FalkorDB
import json

db = FalkorDB(host='localhost', port=6379)
g = db.select_graph('mind-protocol_iris')

# Query nodes
nodes_result = g.query('MATCH (n) RETURN n LIMIT 100')
nodes = [{'id': n.properties.get('node_id', n.id), 'type': n.properties.get('node_type', 'Unknown')} for n in nodes_result.result_set]

# Query links
links_result = g.query('MATCH (s)-[r]->(t) RETURN s.node_id, type(r), t.node_id LIMIT 100')
links = [{'source': row[0], 'type': row[1], 'target': row[2]} for row in links_result.result_set]

print(json.dumps({'nodes': nodes, 'links': links}, indent=2))
" > graph_snapshot.json

Then modify dashboard to load from static file as fallback.

Long-term Fix Required

For Atlas/Felix to implement:

  1. Fix SubEntity loading query in falkordb_adapter.py (case mismatch: 'SubEntity' vs 'Subentity'?)
  2. Disable or fix SafeMode tripwire in safe_mode.py (too sensitive)
  3. Fix WebSocket connection handling in control_api.py:2808

User Impact

Right now: Dashboard loads but shows "0 nodes" because WebSocket connection fails.

Workaround: Static snapshot would let you see the graph structure, but no real-time updates.

Ideal: Fix WebSocket server so real-time visualization works as designed.