Deployment

TraceCraft is designed to run everywhere - from a developer laptop to high-throughput production clusters on managed cloud platforms. This section covers how to configure, deploy, and operate TraceCraft in each environment.

Deployment Options

Production Configuration →

Baseline settings for production: sampling, PII redaction, async export, health checks,

AWS AgentCore →

Deploy TraceCraft alongside AWS Bedrock AgentCore. Covers IAM roles, CloudWatch

Azure AI Foundry →

Integrate with Azure AI Foundry agents. Includes managed identity setup, Azure Monitor

GCP Vertex Agent →

Run TraceCraft with Vertex AI Agent Builder. Covers Workload Identity, Cloud Trace

Kubernetes →

Complete Kubernetes deployment guide: Helm values, ConfigMaps, Secrets, sidecar

High Throughput →

Optimize for millions of traces per day. Covers async batching, aggressive sampling,

Choosing a Deployment Model

Scenario	Recommended Guide
First production deployment	Production Configuration
Running on AWS Bedrock	AWS AgentCore
Running on Azure AI Foundry	Azure AI Foundry
Running on GCP Vertex AI	GCP Vertex Agent
Kubernetes cluster	Kubernetes
Very high trace volume	High Throughput

Quick Start: Production-Ready Config

The following configuration is a safe starting point for any production deployment. Adjust sampling_rate and exporter settings to match your environment.

import os
import tracecraft
 
tracecraft.init(
    service_name=os.getenv("SERVICE_NAME", "my-agent"),
    environment="production",
    console=False,
    otlp_endpoint=os.getenv("OTLP_ENDPOINT"),
    sampling_rate=float(os.getenv("TRACECRAFT_SAMPLING_RATE", "0.1")),
    always_keep_errors=True,
    enable_pii_redaction=True,
)

SERVICE_NAME=my-agent
OTLP_ENDPOINT=https://otlp.example.com:4317
TRACECRAFT_SAMPLING_RATE=0.1

Next Steps

Start with Production Configuration for the foundational settings, then move to the platform-specific guide that matches your infrastructure.

Cloud Platforms Production Configuration