Bot Sessions — Warm Studio

The health check script in colony needs a v2. Current one just checks if processes are running. I want it to also verify: MCP server connectivity, colony message round-trip latency, and disk space on the data partition. Put it at colony/scripts/health-check-v2.ts.

forge

Thinking · 3.2s

The existing health check is at colony/scripts/health-check.ts. I should read it first to understand the current approach, then extend it with the three new checks Max wants. MCP connectivity can be tested by attempting a colony_heartbeat equivalent. For latency, I can do a round-trip message send/receive. Disk space is straightforward with statfs or df parsing.

Read colony/scripts/health-check.ts ▸ Grep "health" in colony/ ▸

Read File

colony/scripts/health-check.ts

72 lines · 1.8 KB · read in 12ms

Read the existing health check. It's a simple process-alive check using pgrep for each bot name. Straightforward to extend.

I'll structure v2 with individual check functions that each return a result object, then aggregate them into a single report. Here's my plan:

1. Process checks -- keep existing behavior
2. MCP connectivity -- attempt a WebSocket handshake to the colony MCP server
3. Message latency -- send a ping through colony, measure round-trip
4. Disk space -- parse df output for the data partition

10:51 AM

Max

Perfect. For the latency check, don't actually send a colony message -- just measure the time to connect to the MCP stdio transport and get a response to a ping. Keep it self-contained.

forge

Write colony/scripts/health-check-v2.ts ▸

Write File

colony/scripts/health-check-v2.ts

148 lines written

Done. The script exports a runHealthCheck() function and also works as a CLI entrypoint. Here's the structure:

type CheckResult = {
  name: string
  status: "pass" | "warn" | "fail"
  message: string
  durationMs: number
}

type HealthReport = {
  timestamp: string
  checks: CheckResult[]
  overall: "healthy" | "degraded" | "unhealthy"
}

Each check runs independently with a 5-second timeout. The MCP connectivity test spawns the stdio transport, sends an initialize request, and measures time-to-response. Disk space warns at 80%, fails at 95%.

forge

Bash bun run colony/scripts/health-check-v2.ts ▸

Bash Output

Colony Health Check v2
======================

[PASS]  process/concierge     running (pid 48291)        2ms
[PASS]  process/forge          running (pid 48305)        1ms
[WARN]  process/historian      not running                1ms
[PASS]  process/curator        not expected               0ms

[PASS]  mcp/connectivity       connected in 127ms       127ms
[PASS]  mcp/latency            ping: 34ms                34ms

[PASS]  disk/data              62% used (186G / 300G)     8ms
[PASS]  disk/tmp               14% used (2.1G / 15G)     3ms

Overall: HEALTHY (7/8 pass, 1 warn)

All checks passing except the expected historian warning (it's offline per the bot reduction). The MCP round-trip came in at 34ms which is solid for a stdio transport.

Script is at colony/scripts/health-check-v2.ts and ready to use. Want me to wire it into the cron schedule or update the colony tools reference?

10:58 AM

Max

Looks great. Yeah, update the colony tools reference in forge's identity docs to point to v2 as well. And commit it.

forge

Edit colony-tools.md ▸ Bash git add && git commit ▸

Updated the colony tools reference to include v2 and committed both files:

feat: add health check v2 with MCP, latency, and disk checks

Co-Authored-By: forge

Two files changed: colony/scripts/health-check-v2.ts (new) and colony/identities/forge/colony-tools.md (updated reference).