FluxMeter Python SDK

Send AI token usage events to FluxMeter for real-time aggregation and billing.

Website: fluxmeter.dev · GitHub · API reference

Install

pip install fluxmeter

Quick Start (3 lines)

from fluxmeter import FluxMeter

# Lite (HTTP) — no Kafka
meter = FluxMeter(api_url="http://localhost:8000")
meter.track("cust_123", "gpt-4o", input_tokens=500, output_tokens=150)

# Full (Kafka + WAL)
meter = FluxMeter(kafka_brokers="localhost:9094")

Wrap (path activation)

from openai import OpenAI
from fluxmeter import FluxMeter, wrap, BudgetExceededError, StreamKilledError

meter = FluxMeter(api_url="http://localhost:8000")
client = wrap(OpenAI(), meter, customer_id="cust_123", fail_open=True)
try:
    client.chat.completions.create(model="gpt-4o-mini", messages=[...])
except BudgetExceededError:
    pass  # provider never called
# stream=True → StreamKilledError if est cost exceeds reserve

Gateway (zero-code ingest)

Point OpenAI base_url at FluxMeter Gateway (http://localhost:8080/v1) with header X-FluxMeter-Customer-Id — no SDK track_* needed. See docs/gateway.md.

OpenAI Integration

import time
from openai import OpenAI
from fluxmeter import FluxMeter

client = OpenAI()
meter = FluxMeter(kafka_brokers="localhost:9094", environment="production")

start = time.time()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)
latency = int((time.time() - start) * 1000)

# One line to meter the usage
meter.track_openai("cust_123", response, latency_ms=latency)

Anthropic Integration

import anthropic
from fluxmeter import FluxMeter

client = anthropic.Anthropic()
meter = FluxMeter(kafka_brokers="localhost:9094")

response = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
)

meter.track_anthropic("cust_123", response)

Manual Tracking (any provider)

meter.track(
    customer_id="cust_123",
    model_id="gemini-1.5-pro",
    provider="google",
    input_tokens=2000,
    output_tokens=500,
    request_id="req_abc123",
    span_id="span_7f3a",          # link to your tracing
    session_id="sess_456",        # group by conversation
    latency_ms=890,
    environment="production",
    metadata={"feature": "code-review", "team": "platform"},
)

Query usage (HTTP API)

Metering is ingest-only in the SDK. Read usage via the FluxMeter API (see API reference):

import httpx

API = "http://localhost:8000"
headers = {"X-API-Key": "your-key"}

# Lifetime cumulative
httpx.get(f"{API}/usage/customer/cust_123", headers=headers).json()

# Monthly / daily (UTC calendar buckets)
httpx.get(f"{API}/usage/customer/cust_123/period/2026-07", headers=headers).json()
httpx.get(f"{API}/usage/customer/cust_123/day/2026-07-05", headers=headers).json()

# Agent task (set parent_span_id on track)
httpx.get(f"{API}/usage/span/span_agent_42", headers=headers).json()

# Project / conversation (set session_id; lite HTTP ingest path)
httpx.get(f"{API}/usage/session/sess_456", headers=headers).json()

Configuration

meter = FluxMeter(
    kafka_brokers="kafka1:9092,kafka2:9092",  # Kafka cluster
    topic="token-events",                       # Topic name (default)
    environment="production",                   # Applied to all events
    producer_config={                           # Extra Kafka producer config
        "security.protocol": "SASL_SSL",
        "sasl.mechanisms": "PLAIN",
        "sasl.username": "...",
        "sasl.password": "...",
    },
)

How It Works

Your App  →  meter.track(...)  →  Kafka  →  Flink (real-time aggregation)  →  Redis
                                                                                 ↓
                                                                           Grafana / API

Events are batched and compressed (lz4) before sending. The SDK flushes automatically on process exit.

Requirements

• Python 3.9+
• confluent-kafka (librdkafka-based, high performance)
• FluxMeter infrastructure running (Kafka + Flink + Redis)