Phoenix Cloud

​

What is Phoenix Cloud?

​

Benefits of Phoenix Cloud

​

Getting Started

import phoenix as px

# Configure Phoenix Cloud
client = px.Client(
    endpoint="https://app.phoenix.arize.com",
    api_key="YOUR_API_KEY"
)

from openinference.instrumentation.openai import OpenAIInstrumentor
from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
from opentelemetry.sdk.trace import TracerProvider
from opentelemetry.sdk.trace.export import SimpleSpanProcessor

# Configure OpenTelemetry
tracer_provider = TracerProvider()
tracer_provider.add_span_processor(
    SimpleSpanProcessor(
        OTLPSpanExporter(
            endpoint="https://app.phoenix.arize.com/v1/traces",
            headers={"api_key": "YOUR_API_KEY"}
        )
    )
)

# Instrument OpenAI
OpenAIInstrumentor().instrument(tracer_provider=tracer_provider)

​

Features

​

Tracing and Observability

• Full OpenTelemetry Support: Ingest traces via OTLP (HTTP and gRPC)
• LLM-Specific Instrumentation: Automatic instrumentation for OpenAI, Anthropic, LangChain, LlamaIndex, and more
• Real-Time Monitoring: Live trace ingestion with sub-second latency
• Advanced Search: Filter and search traces by any attribute
• Cost Tracking: Automatic token usage and cost calculation

​

Evaluation and Experimentation

• LLM Evals: Run evaluations using Phoenix’s built-in evaluators or custom logic
• Datasets: Version-controlled datasets for experimentation
• Experiments: Track prompt and model variations
• A/B Testing: Compare performance across different configurations

​

Prompt Management

• Version Control: Track prompt changes over time
• Playground: Test prompts with different models and parameters
• Deployment: Deploy prompts from the UI or API
• Templates: Reusable prompt templates with variables

​

Team Collaboration

• Workspaces: Isolate projects by team or environment
• Role-Based Access: Admin, Member, and Viewer roles
• Shared Dashboards: Collaborate on trace analysis
• Comments and Annotations: Add context to traces

​

Pricing

​

Free Tier

• Traces: Up to 100,000 spans/month
• Retention: 7 days
• Users: Unlimited
• Projects: Unlimited
• Support: Community (Slack)

​

Pro Tier

• Traces: 1M spans/month included, $0.10 per additional 100K spans
• Retention: 30 days (configurable)
• SSO: Google, GitHub, Microsoft
• Support: Email support with 24-hour SLA
• Advanced Features: Custom evaluators, datasets, experiments

​

Enterprise Tier

• Traces: Custom volume
• Retention: Custom (up to 365 days)
• SSO: SAML, OIDC, LDAP
• Support: Dedicated support with 4-hour SLA
• Advanced Features: All features included
• Compliance: SOC 2, HIPAA, GDPR
• SLA: 99.9% uptime guarantee
• Custom Deployment: Option for dedicated infrastructure

​

Security and Compliance

​

Data Security

• Encryption in Transit: TLS 1.3 for all data in transit
• Encryption at Rest: AES-256 encryption for all stored data
• Network Isolation: VPC isolation with private subnets
• Access Controls: Role-based access control (RBAC)
• Audit Logs: Complete audit trail of all actions

​

Compliance

• SOC 2 Type II: Annual audit completed
• GDPR: EU data residency available
• HIPAA: Available on Enterprise plans
• ISO 27001: In progress

​

Authentication

• SSO: SAML 2.0, OIDC, Google, GitHub, Microsoft
• MFA: Multi-factor authentication required for admins
• API Keys: Scoped API keys with expiration
• Session Management: Automatic session timeout

​

Data Retention and Deletion

• Configurable Retention: Set retention policies by project
• Automatic Deletion: Data deleted after retention period
• Right to Deletion: Delete user data on request
• Export: Export all traces in OpenTelemetry format

​

Regions and Data Residency

• US East (Virginia): Default region for North America
• EU West (Ireland): For EU customers requiring GDPR compliance

​

Integration Examples

​

Python SDK

import phoenix as px
from phoenix.trace import SpanEvaluations

# Configure client
client = px.Client(
    endpoint="https://app.phoenix.arize.com",
    api_key="YOUR_API_KEY"
)

# Query traces
traces = client.query_spans(
    project_name="my-project",
    start_time="2024-01-01",
    end_time="2024-01-31"
)

# Run evaluations
evaluations = client.run_evals(
    evaluators=["hallucination", "toxicity"],
    dataset_name="production-traces"
)

​

OpenTelemetry Collector

receivers:
  otlp:
    protocols:
      grpc:
        endpoint: 0.0.0.0:4317
      http:
        endpoint: 0.0.0.0:4318

exporters:
  otlphttp:
    endpoint: https://app.phoenix.arize.com
    headers:
      api_key: YOUR_API_KEY

service:
  pipelines:
    traces:
      receivers: [otlp]
      exporters: [otlphttp]

​

cURL Examples

# Send traces via HTTP
curl -X POST https://app.phoenix.arize.com/v1/traces \
  -H "Content-Type: application/json" \
  -H "api_key: YOUR_API_KEY" \
  -d @traces.json

# Query projects
curl -H "Authorization: Bearer YOUR_API_KEY" \
  https://app.phoenix.arize.com/api/v1/projects

​

Migration from Self-Hosted

from phoenix.db import get_traces

traces = get_traces(
    project_name="my-project",
    start_time="2024-01-01"
)

traces.to_parquet("traces.parquet")

import phoenix as px

client = px.Client(
    endpoint="https://app.phoenix.arize.com",
    api_key="YOUR_API_KEY"
)

client.upload_traces("traces.parquet", project_name="my-project")

# Before (self-hosted)
endpoint = "http://localhost:6006"

# After (Phoenix Cloud)
endpoint = "https://app.phoenix.arize.com"
headers = {"api_key": "YOUR_API_KEY"}

​

Monitoring and Alerts

​

System Status

​

Alerts

• Trace volume thresholds
• Error rate increases
• Evaluation failures
• Cost thresholds

​

Metrics Dashboard

• Traces ingested
• Storage used
• API calls
• Active users
• Cost breakdown

​

Support

​

Community Support (Free Tier)

• Slack: Join the Phoenix Slack community
• GitHub: File issues at github.com/Arize-ai/phoenix
• Documentation: docs.arize.com/phoenix

​

Email Support (Pro Tier)

• Response Time: 24-hour SLA
• Email: [email protected]
• Priority: Normal

​

Dedicated Support (Enterprise Tier)

• Response Time: 4-hour SLA
• Channels: Email, Slack Connect, Phone
• Priority: High
• Success Manager: Dedicated customer success manager
• Training: Onboarding and training sessions

​

FAQ

​

Next Steps