Phoenix — Animate | AI Wisdom

← Back to topic
🎞 ANIMATE · Mode 3 of 5 · +100 XP— Arize Phoenix
📋Launch
⚙️Instrument
🔄Trace
✅Analyse
🎯 Next: Test →
Step 1 / 4
Speed
Launch Phoenix
px.launch_app() starts Phoenix UI locally on http://localhost:6006💡 Analogy
Like running a local database GUI — one command and your observability UI is ready.
Phoenix is an open-source LLM observability tool. launch_app() starts a local web UI for viewing traces and embeddings without any cloud setup.
Console Output
> Phoenix UI: http://localhost:6006
> In-memory trace store initialized
> Ready to receive spans
●phoenix_tracing.py
Launch PhoenixL9–9
1import phoenix as px
2from openinference.instrumentation.openai import OpenAIInstrumentor
3from opentelemetry import trace
4from opentelemetry.sdk.trace import TracerProvider
5from opentelemetry.sdk.trace.export import BatchSpanProcessor
6from openai import OpenAI
7 
8# Launch Phoenix UI on http://localhost:6006
9session = px.launch_app()
10 
11# Auto-instrument OpenAI calls → Phoenix traces all LLM calls
12tracer_provider = TracerProvider()
13tracer_provider.add_span_processor(
14    BatchSpanProcessor(px.get_default_span_exporter())
15)
16trace.set_tracer_provider(tracer_provider)
17OpenAIInstrumentor().instrument()
18 
19client = OpenAI()
20 
21# All OpenAI calls are now automatically traced
22def ask(question: str) -> str:
23    response = client.chat.completions.create(
24        model="gpt-4o-mini",
25        messages=[{"role": "user", "content": question}],
26    )
27    return response.choices[0].message.content
28 
29answer = ask("What is LLM observability?")
30 
31# View embeddings cluster to find bad clusters
32ds = px.Client().get_spans_dataframe("default")
33print(ds[["name", "llm.token_count.total", "latency_ms"]].head())
34print(f"Phoenix UI: {session.url}")
34 lines
Timeline — 4 Steps
Launch Phoenix
px.launch_app() starts Phoenix UI locally on http://localhost:6006💡 Analogy
Like running a local database GUI — one command and your observability UI is ready.
Phoenix is an open-source LLM observability tool. launch_app() starts a local web UI for viewing traces and embeddings without any cloud setup.
Console Output
> Phoenix UI: http://localhost:6006
> In-memory trace store initialized
> Ready to receive spans

Was this page helpful?

Discussion

px.launch_app() starts Phoenix UI locally on http://localhost:6006

💡 Analogy

Like running a local database GUI — one command and your observability UI is ready.

Phoenix is an open-source LLM observability tool. launch_app() starts a local web UI for viewing traces and embeddings without any cloud setup.

Console Output

> Phoenix UI: http://localhost:6006
> In-memory trace store initialized
> Ready to receive spans

px.launch_app() starts Phoenix UI locally on http://localhost:6006

💡 Analogy

Like running a local database GUI — one command and your observability UI is ready.

Phoenix is an open-source LLM observability tool. launch_app() starts a local web UI for viewing traces and embeddings without any cloud setup.

Console Output

> Phoenix UI: http://localhost:6006
> In-memory trace store initialized
> Ready to receive spans