VantagePeers Docs

Error Monitoring

Proactive error detection across Convex deployments with automatic GitHub issue creation.

Error Monitoring

VantagePeers proactively monitors your Convex deployments for errors. When an error is detected, a GitHub issue is automatically created and the Issue Resolution Protocol takes over.

How It Works

A cron job runs every 5 minutes, polling each monitored deployment's error logs via the Convex REST API. New errors are deduplicated by function name + error message, and unique errors trigger automatic GitHub issue creation.

Cron (every 5 min)
       |
       v
Poll each deployment's error logs
       |
       v
New error detected? ──No──> Skip (dedup by hash)
       |
      Yes
       v
Create GitHub issue with stack trace
       |
       v
IRP webhook triggers (auto-comment + 14-task mission)

Adding a Deployment

Register a deployment to monitor via MCP:

// add_deployment
{
  "name": "myreeldream",
  "deploymentUrl": "https://calm-gerbil-63.convex.cloud",
  "deployKeyEnvVar": "DEPLOY_KEY_MYREELDREAM",
  "githubRepo": "myreeldream-ai/MyShortReel-beta",
  "orchestrator": "dave"
}

Then set the deploy key as a Convex environment variable:

npx convex env set DEPLOY_KEY_MYREELDREAM=your-admin-deploy-key

Deduplication

Errors are deduplicated using a hash of functionName + errorMessage. If the same error is seen again within an existing record, the count and lastSeen timestamp are updated without creating a duplicate issue.

MCP Tools

ToolDescription
add_deploymentRegister a Convex deployment to monitor
remove_deploymentStop monitoring a deployment
list_errorsList detected errors, optionally filtered by deployment
get_errorGet full error details including stack trace

Integration with IRP

When the error monitor creates a GitHub issue, the existing webhook pipeline handles the rest:

  1. Issue created with [Auto] prefix and bug + auto-detected labels
  2. Webhook triggers IRP: auto-comment + 14-task mission
  3. Assigned orchestrator receives the mission and begins resolution

This means errors are detected and assigned before users report them.

On this page