Deployment Logs Skill

Fetch and analyze deployment logs to debug issues and monitor your deployed application.

When to Use

Use this skill when the user:

• Reports that their deployed application is not working correctly
• Wants to see what errors are occurring in production
• Needs to debug a deployment failure or runtime issue
• Asks to check deployment or server logs
• Wants to monitor application behavior after deployment

How to Use

In the JavaScript notebook, call the

fetchDeploymentLogs

function:

const result = await fetchDeploymentLogs({
    afterTimestamp: (Date.now() - 3600 * 1000),  // Last hour
    message: "ERROR"
});
if (result.found) {
    console.log(result.logs);
} else {
    console.log("No deployment logs found.");
}

Parameters

• afterTimestamp

  (float, optional): Only include logs after this Unix timestamp (milliseconds)

• beforeTimestamp

  (float, optional): Only include logs before this Unix timestamp (milliseconds)

• message

  (str, optional): RE2 regular expression to filter logs by message content

• messageContext

  (dict, optional): Context configuration for regex matches with:

  • lines

    (int, required): Number of lines before/after each match (0-100)

  • limit

    (int, required): Max matches to fetch context for (1-10)

Return Value

Returns a dictionary with:

• logs

  : String containing the formatted deployment logs (empty string if none found)

• found

  : Boolean indicating whether any logs were found

• message

  : Optional message when no logs are found

Best Practices

1. Start broad, then filter: First fetch logs without filters to understand what's available, then use

   message

   regex to narrow down
2. Use timestamps for recent issues: If the user reports a recent issue, use

   afterTimestamp

   to focus on recent logs
3. Filter by log level: Use message regex like

   "ERROR"

   or

   "WARN"

   to find problematic entries
4. Look for patterns: Search for specific error messages, stack traces, or module names
5. Use context for debugging: When investigating specific errors, use

   messageContext

   to see surrounding log lines (similar to

   grep -C

   )

Examples

Fetch All Recent Logs

const result = await fetchDeploymentLogs();
console.log(result.logs);

Fetch Logs from Last Hour

const oneHourAgo = Date.now() - 3600 * 1000;  // Convert to milliseconds
const result = await fetchDeploymentLogs({ afterTimestamp: oneHourAgo });
if (result.found) {
    console.log(result.logs);
}

Find Error Logs

const result = await fetchDeploymentLogs({ message: "ERROR" });
if (result.found) {
    console.log(result.logs);
} else {
    console.log("No error logs found");
}

Find Database-Related Errors

const result = await fetchDeploymentLogs({
    message: "(?i)(database|postgres|mysql|mongo|connection.*refused)"
});
console.log(result.logs);

Find Startup Failures

const result = await fetchDeploymentLogs({
    message: "(?i)(failed|crash|exception|traceback)"
});
console.log(result.logs);

Find Errors with Context Lines

// Get 10 lines of context before/after each error (up to 3 matches)
const result = await fetchDeploymentLogs({
    message: "ERROR",
    messageContext: { lines: 10, limit: 3 }
});
if (result.found) {
    console.log(result.logs);
}

Combined Time and Pattern Filter

// Get logs from the last 30 minutes containing errors
const thirtyMinAgo = Date.now() - 1800 * 1000;
const result = await fetchDeploymentLogs({
    afterTimestamp: thirtyMinAgo,
    message: "(?i)(error|exception|failed)"
});
if (result.found) {
    console.log(result.logs);
} else {
    console.log("No matching logs in the last 30 minutes");
}

Common Log Patterns to Search For

• Any errors:

  "ERROR"

• Warnings and errors:

  "(WARN|ERROR)"

• Database issues:

  "(?i)database|postgres|mysql|connection"

• Network errors:

  "(?i)timeout|refused|unreachable"

• Authentication:

  "(?i)auth|token|unauthorized|forbidden"

• Memory issues:

  "(?i)memory|heap|oom|killed"

• Import/module errors:

  "(?i)import|module|cannot find"

Troubleshooting Tips

1. No logs returned: The deployment may not have run recently, or there may be no logs matching your filter
2. Too many logs: Use the

   message

   parameter to filter, or narrow the time range
3. Looking for specific errors: Use case-insensitive regex with

   (?i)

   prefix