This guide explains how to use Datadog with your Pipecat Cloud agents for logging, traces, and custom metrics.
This guide is a work in progress and may have some sharp edges. However, it has been vetted for basic usage.

​
Prerequisites

  • Have a Datadog account and API key.
  • Use at least version 0.0.7 of the Pipecat Cloud base image:
# dailyco/pipecat-base:latest should also work
FROM dailyco/pipecat-base:0.0.7

​
Add Datadog API key as a secret

If you’re using keeping your secrets in an .env file as recommended, add your Datadog API key to that file: 
DD_API_KEY=<your-datadog-api-key>
Then update your secret set from that file: 
pcc secrets set <your-agent-secrets> --file .env
See Secrets for more details.

​
Create datadog.yaml

This file contains the Datadog Agent configuration. The configuration shown below assumes you want all Datadog features covered in this guide (logging, traces, and custom metrics). 
site: <your-datadog-site> # e.g. us5.datadoghq.com
hostname: pipecat.daily.co # (or whatever would be useful for you to see in the Datadog dashboards)

# Include the following line only if you need logging (see "Enabling Logging")
logs_enabled: true

# Uncomment the below lines if you DON'T need traces (see "Enabling Traces")
# apm_config:
#   enabled: false

# Uncomment the below lines if you DON'T need custom metrics (see "Enabling Custom Metrics")
# use_dogstatsd: false

kubelet_tls_verify: false
autoconfig_exclude_features:
  - kubernetes
  - orchestratorexplorer

​
Create pre-app.sh

This script runs before the agent launcher runs. The file shown below assumes you want all Datadog features covered in this guide (logging, traces, and custom metrics). 
#!/bin/bash

# Write the Datadog API key to the config file
echo "api_key: $DD_API_KEY" >> /etc/datadog-agent/datadog.yaml

# Start the Datadog Agent service
service datadog-agent start || true

# Include the below line only if you need traces (see "Enabling Traces")
# HACK: here we restart the Datadog Trace Agent specifically, since we know it failed to start above
service datadog-agent-trace restart || true

## DEBUGGING

# Check the status of the Datadog Agent subsystems for logging, custom metrics, and traces
# service datadog-agent status
# service datadog-agent-trace status
# datadog-agent status

# Display the Datadog agent log
# cat /var/log/datadog/agent.log
You may see Starting Datadog Agent: datadog-agent failed! in your logs, triggered by service datadog-agent start. You can safely ignore this! As long as service datadog-agent status and service datadog-agent-trace status indicate that both of those services end up running, then we should have no trouble sending logs, traces, and custom metrics to Datadog.

​
Update Dockerfile with the Datadog Agent

The snippet shown below assumes you want all Datadog features covered in this guide (logging, traces, and custom metrics). 
# Install the Datadog Agent
# (note that you can literally use "dummy"—or anything else—as the API key)
RUN apt-get -y update; apt-get -y install curl
RUN DD_API_KEY=dummy \
    DD_INSTALL_ONLY=true \
    bash -c "$(curl -L https://install.datadoghq.com/scripts/install_script_agent7.sh)"

# Configure Datadog Agent
COPY ./datadog.yaml /etc/datadog-agent
# Include the below line if you need logging (see "Enabling Logging")
COPY ./python.d /etc/datadog-agent/conf.d/python.d

# Include the below lines only if you need traces or custom metrics (see "Enabling Traces" or "Enabling Custom Metrics")
# Prepare file system for the Datadog Agent to use for Unix Sockets
RUN mkdir -p /var/run/datadog/
RUN chown dd-agent:dd-agent /var/run/datadog/

# Start Datadog Agent (move pre-app script to expected place where it'll be invoked by Pipecat Cloud base image)
COPY --chmod=755 ./pre-app.sh /app/

​
Enabling logging

​
Enable logging in datadog.yaml

Ensure the following is present in your datadog.yaml: 
logs_enabled: true

​
Create python.d/conf.yaml

Configure log collection for your agent: 
# Use whatever agent name would be useful for you to see in the Datadog dashboard
logs:
  - type: file
    path: /var/log/<my-agent-name>/datadog.log
    service: <my-agent-name>
    source: python
    sourcecategory: sourcecode

​
Update Dockerfile for logging

Ensure the following is present in your Dockerfile: 
COPY ./python.d /etc/datadog-agent/conf.d/python.d
This configures the Datadog Agent to collect logs from your agent.

​
Configure logger in bot.py

Add a Datadog sink to your logger. This sink will write log lines into a file that your Datadog Agent will read from, in a Datadog-friendly format. 
import json

# Format log lines so that:
# - Datadog can parse and display log levels, messages, and timestamps
# - You can search Datadog logs by session_id
def datadog_format(record):
    format_string_safe_message = (
        json.dumps(record["message"]).replace("{", "{{").replace("}", "}}")
    )
    session_id = record["extra"].get("session_id", "")
    return (
        "{{"
        '"timestamp":"{time:YYYY-MM-DD HH:mm:ss.SSS}",'
        '"levelname":"{level}",'
        '"name":"{name}",'
        '"function":"{function}",'
        '"line":{line},'
        f'"message":{format_string_safe_message},'
        f'"session_id": "{session_id}"'
        "}}\n"
    )

async def bot(args: DailySessionArguments):
    # Add Datadog-specific sink
    logger.add(
        "/var/log/<my-agent-name>/datadog.log", # Use same path as in python.d/conf.yaml
        rotation="10 MB",
        retention="7 days",
        format=datadog_format,
    )
    # ...

​
Enabling traces

​
Enable traces in datadog.yaml

Ensure apm_config.enabled is NOT set to false in your datadog.yaml: 
# DELETE or COMMENT OUT the following lines, if they're present
# apm_config:
#  enabled: false

​
Start the Datadog Trace Agent in pre-app.sh

Ensure the following is present in your pre-app.sh: 
service datadog-agent-trace restart || true

​
Prepare file system in Dockerfile

Ensure the following is present in your Dockerfile: 
RUN mkdir -p /var/run/datadog/
RUN chown dd-agent:dd-agent /var/run/datadog/

​
Add ddtrace Python library

In your requirements.txt: 
ddtrace

​
Record traces in bot.py

from ddtrace import tracer

@tracer.wrap()
def my_function_call():
    # ...

​
Enabling custom metrics

​
Enable custom metrics in datadog.yaml

Ensure use_dogstatsd is NOT set to false in your datadog.yaml: 
# DELETE or COMMENT OUT the following line, if it's present
# use_dogstatsd: false

​
Prepare file system in Dockerfile

Ensure the following is present in your Dockerfile: 
RUN mkdir -p /var/run/datadog/
RUN chown dd-agent:dd-agent /var/run/datadog/

​
Add datadog Python library

In your requirements.txt: 
datadog

​
Record custom metrics in bot.py

from datadog import statsd

def my_function_call():
    statsd.increment("myagent.my_function_call")