@dreki-gg/pi-datadog

Datadog log and RUM search tools for pi — query production logs and Real User Monitoring sessions with project-aware context.

Setup

1. Install

pi install npm:@dreki-gg/pi-datadog

2. Set credentials

Export your Datadog API and Application keys as environment variables:

export DD_API_KEY="your-datadog-api-key"
export DD_APP_KEY="your-datadog-app-key"

Tip: Add these to your shell profile (~/.zshrc, ~/.bashrc) or use a .env manager. Never commit credentials to version control.

You need an Application Key (not just an API key) because log search uses Datadog's read endpoints.

3. Configure your project (optional)

Create .pi/datadog.json in your project root to set defaults:

{
  "service": "my-api",
  "env": "production",
  "site": "datadoghq.com",
  "defaultTags": ["team:backend"],
  "defaultTimeRange": "1h",
  "rumApplicationId": "abcd-1234",
  "rumService": "web-frontend"
}

Field	Description	Default
service	Default service name for queries	(none)
env	Default environment	(none)
site	Datadog site	datadoghq.com
defaultTags	Tags auto-appended to every query	[]
defaultTimeRange	Default lookback window	1h
rumApplicationId	Default RUM application id (@application.id) for RUM searches	(none)
rumService	Default service for RUM searches (falls back to service)	(none)

Supported sites: datadoghq.com (US1), us3.datadoghq.com, us5.datadoghq.com, datadoghq.eu (EU).

Usage

Natural language

Just ask pi to search your logs:

> Show me the errors in production from the last 30 minutes
> What's causing the 500 errors on the payments service?
> Find logs with "timeout" in staging from the past 24 hours

The agent uses your .pi/datadog.json defaults automatically — you don't need to specify service or environment unless you want to override them.

You can also ask about front-end / RUM sessions:

> Show me RUM sessions from the last hour
> Find RUM errors on the checkout view in production
> What views did users hit in the last 30 minutes?

Tool: datadog_logs_search

The agent calls this tool with Datadog query syntax:

Parameter	Type	Description
query	string	Required. Datadog log query (e.g. status:error, @http.status_code:500)
from	string	Start time — relative (15m, 1h, 7d) or ISO 8601
to	string	End time — relative, ISO 8601, or now
limit	number	Max results (1–100, default 25)
sort	string	newest or oldest
service	string	Override project default service
env	string	Override project default environment

The agent receives a compact inline digest (counts, status breakdown, services). The full, untruncated log entries are written to a temp file whose path is in the response — the agent reads that file with the read tool to inspect actual content.

Tool: datadog_rum_search

Searches Datadog RUM events (sessions, views, actions, front-end errors) with RUM query syntax. Defaults to session events (@type:session) when the query omits @type.

Parameter	Type	Description
query	string	Required. RUM query (e.g. @type:session, @type:error, @view.url_path:/checkout). Leave empty to list recent sessions
from	string	Start time — relative (15m, 1h, 7d) or ISO 8601
to	string	End time — relative, ISO 8601, or now
limit	number	Max results (1–100, default 25)
sort	string	newest or oldest
service	string	Override project default RUM service
env	string	Override project default environment
applicationId	string	Override project default RUM application id (@application.id)

Like the logs tool, it returns a compact digest inline (counts, event-type breakdown, services) and writes the full events to a rum-*.md temp file the agent can read.

Command: /datadog

Check your configuration and connection status:

/datadog

Shows: credential status (set/missing), project config (service, env, site, tags, RUM application, RUM service).

Query Syntax Examples

status:error                          # All errors
service:my-api status:error           # Errors for a specific service
@http.status_code:500                 # 500 errors
"connection refused"                  # Full-text search
service:my-api env:production @duration:>5000  # Slow requests

See the Datadog Log Search Syntax docs for the full reference.

RUM query examples

@type:session                          # User sessions (the default)
@type:error                            # Front-end errors
@type:view @view.url_path:/checkout    # Views on the checkout page
@type:session @session.type:user       # Real user sessions (exclude synthetics)
@application.id:abcd-1234 @type:action # Actions in a specific RUM application

See the RUM Search Syntax docs for the full reference.

How It Works

1. The extension loads .pi/datadog.json from your project on session start
2. When the agent calls datadog_logs_search, it merges your query with project defaults (service, env, tags)
3. Results are formatted as markdown with status icons, truncated messages, and key attributes
4. The agent receives structured metadata (status breakdown, services found) to summarize intelligently

License

MIT