ariffazil/wealth-sse-transport-fix

WEALTH FastMCP SSE Transport Fix

Problem

wealth-organ container enters a restart loop:

Up 1 second
Exited (0) 2 seconds ago
Up 1 second
Exited (0) 2 seconds ago
...

Root Cause

monolith.py runs mcp.run() which defaults to stdio transport. When Docker starts the container without a TTY, stdin closes immediately. The stdio transport blocks waiting for input → when input never arrives, process exits → Docker restarts → loop.

Fix (one-line code change)

# Before (in monolith.py)
if __name__ == "__main__":
    mcp.run()

# After
if __name__ == "__main__":
    import os
    if os.isatty(0):
        mcp.run()                              # stdio for local dev
    else:
        mcp.run(transport="sse", show_banner=False)  # SSE for container

This makes the transport context-aware: stdio when run locally with a terminal, SSE when run in Docker.

Why SSE works in Docker

• SSE binds FastMCP to an HTTP port (default 8000)
• No stdin needed — HTTP server handles connections
• Container stays alive because the HTTP server is a long-running process
• Docker unless-stopped restart policy handles actual crashes

Healthcheck fix

Remove the broken pg_isready healthcheck from compose. FastMCP SSE doesn't have a generic /health endpoint — the container process itself is the health indicator.

# Before (BROKEN — pg_isready not in WEALTH container)
healthcheck:
  test: ["CMD", "pg_isready", "-U", "arifos"]

# After — remove healthcheck entirely (unless-stopped handles crashes)
# healthcheck: (omit entirely)

Deployment

# 1. Copy patched monolith.py to VPS
scp /local/monolith.py arifos:/root/WEALTH/internal/

# 2. On VPS: commit
cd /root/WEALTH
git add internal/monolith.py
git commit -m "feat(kernel): harden intelligence v2026.04.29 [SSE + SABAR + V2_ALIAS]"
git push origin main

# 3. Rebuild container
cd /root/compose
docker compose up -d wealth-organ

# 4. Verify stable
docker ps  # should show "Up X minutes" not cycling

How to verify the fix worked

# Check no restart loop
docker ps  # wealth-organ: "Up N minutes"

# Check SSE port is listening inside container
docker exec wealth-organ sh -lc "curl -s --connect-timeout 2 http://localhost:8000/"
# Should return something (FastMCP SSE response or 404), not "connection refused"

# Check from Caddy network
docker exec caddy sh -lc "curl -s --connect-timeout 3 http://wealth-organ:8000/"

Key Files

• /root/wealth/internal/monolith.py — source (SSE patch applied here)
• /root/compose/docker-compose.yml — WEALTH service definition
• wealth-organ — container name