phoenix-observability

Everything an AI app does gets recorded step by step, on servers you run yourself rather than a vendor's — the traces, the test sets built out of them, and the scores, all in one place you own.

WHEN TO USE

SCOPE

USE FOR

Trace a slow or wrong LLM response back to the exact step that caused it.
Score a batch of model outputs for hallucination or relevance with an LLM-as-judge evaluator.
Compare two prompts or models against the same test set before shipping a change.
Watch token usage and latency on a production LLM system without a paid vendor.

NOT FOR

Choose LangSmith instead for a managed platform built around LangChain.
Choose Weights & Biases instead for deep-learning experiment tracking.
Choose MLflow instead for general ML lifecycle and model registry management.

This is a saved copy of the skill's SKILL.md instructions. It may differ from the upstream file or the version you install.

In this document (13)

Phoenix - AI Observability Platform

Open-source AI observability and evaluation platform for LLM applications with tracing, evaluation, datasets, experiments, and real-time monitoring.

When to use Phoenix

Use Phoenix when:

Debugging LLM application issues with detailed traces
Running systematic evaluations on datasets
Monitoring production LLM systems in real-time
Building experiment pipelines for prompt/model comparison
Self-hosted observability without vendor lock-in

Key features:

Tracing: OpenTelemetry-based trace collection for any LLM framework
Evaluation: LLM-as-judge evaluators for quality assessment
Datasets: Versioned test sets for regression testing
Experiments: Compare prompts, models, and configurations
Playground: Interactive prompt testing with multiple models
Open-source: Self-hosted with PostgreSQL or SQLite

Use alternatives instead:

LangSmith: Managed platform with LangChain-first integration
Weights & Biases: Deep learning experiment tracking focus
Arize Cloud: Managed Phoenix with enterprise features
MLflow: General ML lifecycle, model registry focus

Quick start

Installation

pip install arize-phoenix

# With specific backends
pip install arize-phoenix[embeddings]  # Embedding analysis
pip install arize-phoenix-otel         # OpenTelemetry config
pip install arize-phoenix-evals        # Evaluation framework
pip install arize-phoenix-client       # Lightweight REST client

Launch Phoenix server

import phoenix as px

# Launch in notebook (ThreadServer mode)
session = px.launch_app()

# View UI
session.view()  # Embedded iframe
print(session.url)  # http://localhost:6006

Command-line server (production)

# Start Phoenix server
phoenix serve

# With PostgreSQL
export PHOENIX_SQL_DATABASE_URL="postgresql://user:pass@host/db"
phoenix serve --port 6006

Basic tracing

from phoenix.otel import register
from openinference.instrumentation.openai import OpenAIInstrumentor

# Configure OpenTelemetry with Phoenix
tracer_provider = register(
    project_name="my-llm-app",
    endpoint="http://localhost:6006/v1/traces"
)

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

# All OpenAI calls are now traced
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)

Core concepts

Traces and spans

A trace represents a complete execution flow, while spans are individual operations within that trace.

from phoenix.otel import register
from opentelemetry import trace

# Setup tracing
tracer_provider = register(project_name="my-app")
tracer = trace.get_tracer(__name__)

# Create custom spans
with tracer.start_as_current_span("process_query") as span:
    span.set_attribute("input.value", query)

    # Child spans are automatically nested
    with tracer.start_as_current_span("retrieve_context"):
        context = retriever.search(query)

    with tracer.start_as_current_span("generate_response"):
        response = llm.generate(query, context)

    span.set_attribute("output.value", response)

Projects

Projects organize related traces:

import os
os.environ["PHOENIX_PROJECT_NAME"] = "production-chatbot"

# Or per-trace
from phoenix.otel import register
tracer_provider = register(project_name="experiment-v2")

Framework instrumentation

OpenAI

from phoenix.otel import register
from openinference.instrumentation.openai import OpenAIInstrumentor

tracer_provider = register()
OpenAIInstrumentor().instrument(tracer_provider=tracer_provider)

LangChain

from phoenix.otel import register
from openinference.instrumentation.langchain import LangChainInstrumentor

tracer_provider = register()
LangChainInstrumentor().instrument(tracer_provider=tracer_provider)

# All LangChain operations traced
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model="gpt-4o")
response = llm.invoke("Hello!")

LlamaIndex

from phoenix.otel import register
from openinference.instrumentation.llama_index import LlamaIndexInstrumentor

tracer_provider = register()
LlamaIndexInstrumentor().instrument(tracer_provider=tracer_provider)

Anthropic

from phoenix.otel import register
from openinference.instrumentation.anthropic import AnthropicInstrumentor

tracer_provider = register()
AnthropicInstrumentor().instrument(tracer_provider=tracer_provider)

Evaluation framework

Built-in evaluators

from phoenix.evals import (
    OpenAIModel,
    HallucinationEvaluator,
    RelevanceEvaluator,
    ToxicityEvaluator,
    llm_classify
)

# Setup model for evaluation
eval_model = OpenAIModel(model="gpt-4o")

# Evaluate hallucination
hallucination_eval = HallucinationEvaluator(eval_model)
results = hallucination_eval.evaluate(
    input="What is the capital of France?",
    output="The capital of France is Paris.",
    reference="Paris is the capital of France."
)

Custom evaluators

from phoenix.evals import llm_classify

# Define custom evaluation
def evaluate_helpfulness(input_text, output_text):
    template = """
    Evaluate if the response is helpful for the given question.

    Question: {input}
    Response: {output}

    Is this response helpful? Answer 'helpful' or 'not_helpful'.
    """

    result = llm_classify(
        model=eval_model,
        template=template,
        input=input_text,
        output=output_text,
        rails=["helpful", "not_helpful"]
    )
    return result

Run evaluations on dataset

from phoenix import Client
from phoenix.evals import run_evals

client = Client()

# Get spans to evaluate
spans_df = client.get_spans_dataframe(
    project_name="my-app",
    filter_condition="span_kind == 'LLM'"
)

# Run evaluations
eval_results = run_evals(
    dataframe=spans_df,
    evaluators=[
        HallucinationEvaluator(eval_model),
        RelevanceEvaluator(eval_model)
    ],
    provide_explanation=True
)

# Log results back to Phoenix
client.log_evaluations(eval_results)

Datasets and experiments

Create dataset

from phoenix import Client

client = Client()

# Create dataset
dataset = client.create_dataset(
    name="qa-test-set",
    description="QA evaluation dataset"
)

# Add examples
client.add_examples_to_dataset(
    dataset_name="qa-test-set",
    examples=[
        {
            "input": {"question": "What is Python?"},
            "output": {"answer": "A programming language"}
        },
        {
            "input": {"question": "What is ML?"},
            "output": {"answer": "Machine learning"}
        }
    ]
)

Run experiment

from phoenix import Client
from phoenix.experiments import run_experiment

client = Client()

def my_model(input_data):
    """Your model function."""
    question = input_data["question"]
    return {"answer": generate_answer(question)}

def accuracy_evaluator(input_data, output, expected):
    """Custom evaluator."""
    return {
        "score": 1.0 if expected["answer"].lower() in output["answer"].lower() else 0.0,
        "label": "correct" if expected["answer"].lower() in output["answer"].lower() else "incorrect"
    }

# Run experiment
results = run_experiment(
    dataset_name="qa-test-set",
    task=my_model,
    evaluators=[accuracy_evaluator],
    experiment_name="baseline-v1"
)

print(f"Average accuracy: {results.aggregate_metrics['accuracy']}")

Client API

Query traces and spans

from phoenix import Client

client = Client(endpoint="http://localhost:6006")

# Get spans as DataFrame
spans_df = client.get_spans_dataframe(
    project_name="my-app",
    filter_condition="span_kind == 'LLM'",
    limit=1000
)

# Get specific span
span = client.get_span(span_id="abc123")

# Get trace
trace = client.get_trace(trace_id="xyz789")

Log feedback

from phoenix import Client

client = Client()

# Log user feedback
client.log_annotation(
    span_id="abc123",
    name="user_rating",
    annotator_kind="HUMAN",
    score=0.8,
    label="helpful",
    metadata={"comment": "Good response"}
)

Export data

# Export to pandas
df = client.get_spans_dataframe(project_name="my-app")

# Export traces
traces = client.list_traces(project_name="my-app")

Production deployment

Docker

docker run -p 6006:6006 arizephoenix/phoenix:latest

With PostgreSQL

# Set database URL
export PHOENIX_SQL_DATABASE_URL="postgresql://user:pass@host:5432/phoenix"

# Start server
phoenix serve --host 0.0.0.0 --port 6006

Environment variables

Variable	Description	Default
`PHOENIX_PORT`	HTTP server port	`6006`
`PHOENIX_HOST`	Server bind address	`127.0.0.1`
`PHOENIX_GRPC_PORT`	gRPC/OTLP port	`4317`
`PHOENIX_SQL_DATABASE_URL`	Database connection	SQLite temp
`PHOENIX_WORKING_DIR`	Data storage directory	OS temp
`PHOENIX_ENABLE_AUTH`	Enable authentication	`false`
`PHOENIX_SECRET`	JWT signing secret	Required if auth enabled

With authentication

export PHOENIX_ENABLE_AUTH=true
export PHOENIX_SECRET="your-secret-key-min-32-chars"
export PHOENIX_ADMIN_SECRET="admin-bootstrap-token"

phoenix serve

Best practices

Use projects: Separate traces by environment (dev/staging/prod)
Add metadata: Include user IDs, session IDs for debugging
Evaluate regularly: Run automated evaluations in CI/CD
Version datasets: Track test set changes over time
Monitor costs: Track token usage via Phoenix dashboards
Self-host: Use PostgreSQL for production deployments

Common issues

Traces not appearing:

from phoenix.otel import register

# Verify endpoint
tracer_provider = register(
    project_name="my-app",
    endpoint="http://localhost:6006/v1/traces"  # Correct endpoint
)

# Force flush
from opentelemetry import trace
trace.get_tracer_provider().force_flush()

High memory in notebook:

# Close session when done
session = px.launch_app()
# ... do work ...
session.close()
px.close_app()

Database connection issues:

# Verify PostgreSQL connection
psql $PHOENIX_SQL_DATABASE_URL -c "SELECT 1"

# Check Phoenix logs
phoenix serve --log-level debug

References

Advanced Usage - Custom evaluators, experiments, production setup
Troubleshooting - Common issues, debugging, performance

Resources

Documentation: https://docs.arize.com/phoenix
Repository: https://github.com/Arize-ai/phoenix
Docker Hub: https://hub.docker.com/r/arizephoenix/phoenix
Version: 12.0.0+
License: Apache 2.0

This skill ships inside the AI Research Skills plugin.

Run this outside your AI tool, using Node.js 22.20.0 or newer. Skills CLI 1.7.2 skips skill selection; prompts cancel installation without an interactive terminal. If choosing Project when asked about installation scope, open your project's folder before running this.

Install checklist

Check selected tools if a picker appears; Claude Code may be unchecked under Additional agents. ↑/↓ moves, Space toggles selections, Enter confirms.
Without a tool picker, check Installing to: for your intended tool. Stop if it names a different tool.
If asked, choose Symlink (Recommended) for shared files or Copy to all agents for separate copies; review the summary before confirming installation.
Install the find-skills skill? defaults to Yes, adding another skill for all projects. Choose No if you don't want that extra installation.

Skills CLI 1.7.2 sends install data to skills.sh by default, including the source, skill names, target tools and file details; GitHub sources are reported only when GitHub confirms the repository is public, while other remote sources can also be reported. Confirmed-public GitHub sources also trigger a security-audit request. Before installing, set the environment variable DO_NOT_TRACK=1 or DISABLE_TELEMETRY=1 to disable both.

npx skills add orchestra-research/ai-research-skills --skill "phoenix-observability" --full-depth

After installing for Claude Code, open it in the installation project if you chose Project; Global makes the skill available across projects on this computer. Once Claude Code loads the skill, its description helps Claude decide when to use it.

If installation created a top-level skills folder that did not exist when your Claude Code session started, run /reload-skills in that session. Run it again after later changes in that folder.

What's inside

Files included with this skill.

View on GitHub